Valuing Green Infrastructure Case Study of Kali Gandaki Watershed, Nepal natural capital P R O J E C T © 2019 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Attribution: Please cite the work as follows: “World Bank. 2019. Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal. ©World Bank.” Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. [Cover photo: Aleksei Kazachok/Shutterstock.com] Valuing Green Infrastructure Case Study of Kali Gandaki Watershed, Nepal © Trancedrumer/Shutterstock.com CONTENTS ACKNOWLEDGEMENTS 1 EXECUTIVE SUMMARY 2 1. INTRODUCTION 14 1.1. The Problem 14 1.2. Purpose and overview of this study 16 2. STUDY AREA 19 2.1. Socio-economic characteristics 19 2.2. Watershed characteristics 22 2.3. Baseline sediment monitoring 23 3. METHODS AND TOOLS 27 3.1. Overview of assessment approach 27 3.2. Developing a sediment budget 29 3.2.1. Sheet and rill erosion 29 3.2.2. Glacial erosion 30 3.2.3. Mass movements: Landslides and rockfalls 31 3.2.4. Road induced erosion 35 3.2.5. Sediment transport in river and fluvial erosion 38 3.3. Modeling benefits of watershed management 38 3.3.1. Activities & costs 39 3.3.2. Impacts on hillslope erosion 42 3.3.3. Impacts on landslide-related risks 42 3.3.4. Impacts on carbon storage 46 3.3.5. Hydropower benefits 48 3.3.6. Local (on-site) benefits 53 3.4. Where to intervene? Prioritizing watershed management activities and locations 55 3.5. Summary of data requirements 57 4. RESULTS 61 4.1. Baseline conditions 61 4.1.1. Sediment budget 61 4.1.2. Landslide risk 65 4.2. Economic values of watershed management 66 4.2.1. The value of an optimal portfolio 66 4.2.2. Value of sediment reduction to Kali Gandaki A 67 4.2.3. Value of reduction in landslide risk 70 4.2.4. Value of carbon storage 71 4.2.5. Local (on-site) benefits 71 4.2.6. The costs of degradation 71 iii Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal 4.3. Prioritizing watershed management activities 71 4.3.1. Evaluating individual activities 72 4.3.2. Intervention portfolios 72 4.4. Changing conditions, changing values 77 5. CONCLUSIONS & RECOMMENDATIONS 79 5.1. Conclusions 79 5.2. Capacity, data, and technical needs 82 5.3. Future work 83 6. REFERENCES 84  FIGURES Figure 2.1: Study area - Kali Gandaki watershed, with location of hydropower facility 20 Figure 2.2: Department of Survey land use/land cover map from year 2000, also showing the locations of settlements 21 Figure 2.3: Example of agricultural terracing on a steep slope in the Kali Gandaki watershed 21 Figure 2.4: The lower Kali Gandaki watershed, where many settlements are connected by a network of rural roads 22 Figure 2.5: Example of rural road construction, cut into steep hillsides without mitigation measures, increasing the chance of erosion and landslides 22 Figure 2.6: Topography of the Kali Gandaki watershed, location of gauging stations and their respective drainage area 24 Figure 2.7: Total load and contribution of sub-watersheds draining to KGA 26 Figure 3.1: Workflow used in this study to evaluate watershed management activities, value their impacts, prioritize intervention locations and estimate the benefit/cost at different levels of investment 28 Figure 3.2: Results of the stochastic connectivity of landslides and runout tool for an area on the middle Kali Gandaki River 32 Figure 3.3: Transport capacity for medium sand (0.5 mm diameter) (left panel) and actual sediment transport for the same grain size (right panel) 39 Figure 3.4: Power generation and river flow at KGA 48 Figure 3.5: Sand concentration entering turbines by month 49 Figure 3.6: Conceptual representation of operating the desanding basins to minimize the sum of costs and damages 50 Figure 3.7: Daily electricity demand 52 Figure 4.1: Comparison of modeled and observed load from the multi-model suite, including the calibrated mass-movement/landslide model 62 Figure 4.2: Sediment load from each sub-watershed to the streams (left) and the processes dominating sediment load in each sub-watershed (right) 63 Figure 4.3: Sediment load from each process in the sediment budget and each sub-watershed to the streams 64 Figure 4.4: Buildings (left) and roads (right) at risk, binned by the failure probability of the landslide/runout they are located on 65 Figure 4.5: The multiple values of watershed management 66 Figure 4.6: Benefit/cost ratio of modeled portfolios (blue points), showing high and low boundaries on estimates (lines) 67 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  iv Figure 4.7: Additional sediment export that could result from abandonment of watershed management activities and existing soil conservation structures (e.g. terraces) in cultivated areas in the lower watershed 72 Figure 4.8: Modeled sediment reduction by sub-watershed, with full implementation of different management practices 73 Figure 4.9: Intervention portfolios optimized for two competing objectives (left column: downstream sediment for hydropower and right column: local erosion reduction) and two budget levels (US $5M and $20M), for comparison 74 Figure 4.10: Trade-off curve showing performance of 1000 optimal scenarios (US $20M budget) in terms of their reduction in downstream sediment and local erosion control 75 Figure 4.11: Agreement map for US $5M scenario 76 TABLES Table 2.1: Land uses and their total areas found in the Kali Gandaki watershed 21 Table 2.2: Gauging stations used in this study and overview of sub-watershed characteristics 25 Table 3.1: Interventions modeled in this study, examples of practices normally included in such programs, and rules for where on the landscape each activity was modeled 41 Table 3.2: Cost summary for interventions modeled 42 Table 3.3: Landslide classes, mitigation approach assumed and parameters impacted 43 Table 3.4: Costs to implement various watershed management activities and benefits to landholders implied based on reported cost sharing 54 Table 3.5: Objectives used to prioritize watershed management activities and locations 55 Table 3.6: Sources and descriptions of data used in this study 57 Table 4.1: Total mean sediment load and sediment yield (per unit area) for major land uses in the study area 63 Table 4.2: Breakdown of values of investment in watershed management and benefit/cost analysis, for budgets ranging from US $500,000 to $50M 68 BOXES Box 1.1: Sedimentation issues and approaches in the Kali Gandaki watershed 17 Box 2.1: Methods used for converting observed sediment data to longer-term sediment load 25 Box 3.1: Methods used for modeling hillslope erosion 30 Box 3.2: Methods used for modeling glacial erosion 31 Box 3.3: Methods used for modeling landslide locations and probabilities 33 Box 3.4: Different approaches for large-scale landslide hazard assessments 34 Box 3.5: Methods used for modeling erosion from road surfaces 36 Box 3.6: Methods used for modeling sediment delivery from road cuts 36 Box 3.7: Road construction practices and standards in Nepal 37 Box 3.8: Methods used for calculating the value of a structure at risk from destruction in a landslide 45 Box 3.9: Methods used for calculating expected costs of road repairs from landslides 45 v Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal ACRONYMS DEM Digital elevation model DoFSC Department of Forests and Soil Conservation, Ministry of Forests and Environment, Government of Nepal GoN Government of Nepal InVEST Integrated Valuation of Ecosystem Services and Tradeoffs KGA Kali Gandaki A Hydropower Plant LSO Landslide object (a group of connected pixels likely to form a landslide) LULC Land use/land cover NEA Nepal Electricity Authority NPR Nepalese rupee NPV Net present value OSM Open Street Map PES Payments for ecosystem services ROOT Restoration Opportunities Optimization Tool RUSLE Revised universal soil loss equation SCC Social cost of carbon SDR Sediment delivery ratio SDU Spatial decision unit SOC Soil organic carbon UNEP United Nations Environment Programme USAID United States Agency for International Development USD United States dollars VSL Value of a statistical life WOCAT World Overview of Conservation Approaches and Technologies Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  vi ACKNOWLEDGEMENTS This report was prepared by a team led by Urvashi Narain stakeholder consultation meetings in October 2017 and with the core team composed of Adrian L. Vogl, Rafael January 2019 in Kathmandu. These include staff from Schmitt, David Simpson, Ben Bryant, and Stacie Wolny the Nepal Electricity Authority, Soil Conservation and from Stanford University’s Natural Capital Project, Kumud Watershed Management office, Parbat, Department Raj Kafle and Rajendra Bhandari from Kathmandu of Hydrology and Meteorology, Department of Mines University, and Jiang Ru, Pyush Dogra, Annu Rajbhandari, and Geology, Department of Roads, Water and Energy and Ahmad Aslam from the World Bank (WB). The team Commission Secretariat, Ministry of Agriculture and would like to thank Suzata Karki, Dejina Shrestha, Abha Livestock Development, Kathmandu University, PANI/ Bhattarai, Prasansha Shrestha and Stuti Shakya from USAID and World Wildlife Fund. Kathmandu University for assisting with the field sampling and data collection. Contributions were also received from Constructive comments on the report were received from the David Freyberg of Stanford University and Andrea Kutter following peer reviewers: from the World Bank - Nagaraja and Marcelo Arcerbi of the World Bank. The team also Rao Harshadeep, Grant Milne, Rikard Liden, Neha Vyas, thanks Gregory Morris of GLM Engineering and Govinda Stuart Alexander Fraser, and Stefano Pagiola, and Stephen Timilsina and Rabin Shrestha of the World Bank for Polasky from the University of Minnesota. helpful correspondence and conversations. The authors would also like to thank Latha Sridhar for The team is very grateful for the support and overall guidance production management. The manuscript was edited by Simi of Dr. Ram Parsad Lamsal, Director General, and Dr. Prem Mishra and designed by Roots Advertising. Any remaining Paudel, Under Secretary, Department of Forests and Soil errors or omissions are the authors’ own. Conservation, Ministry of Forests and Environment. The team also thanks Hartwig Schafer (Regional Vice President, The team gratefully acknowledges the financial support World Bank), Robert Saum (Director, Regional Integration, provided for this report by the Korean Green Growth Trust South Asia Region), Idah Z. Pswarayi-Riddihough and Fund (KGGTF). The KGGTF is a partnership between the Qimiao Fan (current and former Nepal Country Directors, World Bank Group (WBG) and the Republic of Korea. As an World Bank), Faris Hadad-Zervos (Nepal Country Manager, implementation-focused trust fund that prioritizes the World World Bank), Karin Kemper (Global Director, World Bank’s inclusive green growth objectives, this partnership Bank), John Roome (Regional Director, World Bank) and supports countries in their sustainable growth strategies and Magda Lovei (Practice Manager, World Bank) for their investments. In addition to funding, the KGGTF enhances encouragement and support. other WB tools and value-added services by providing access to technical experts and facilitating dialogue between The team would like to extend special thanks to the many practitioners of green growth policies and investments. The experts and stakeholders who provided assistance with KGGTF strengthens and expands the World Bank’s climate data and knowledge of local conditions and who attended smart investment portfolio. 1 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal EXECUTIVE SUMMARY THE MULTIPLE BENEFITS OF acquire and transport water from other places. Hedgerows and plantings may also be harvested for food, fodder, WATERSHED MANAGEMENT or other products. They may also serve as windbreaks, Watersheds are an appropriate and effective unit for providing protection against the elements, and detaining or managing ecological assets, given the interconnected nature diverting floodwaters. of economic activities and their impacts within a watershed, locally and regionally, upstream and downstream. Watersheds These physical effects translate into economic and societal are increasingly recognized as a critical form of green benefits. The implications of higher soil fertility and more infrastructure that provides a flow of economic benefits. In reliable water availability is that more food (or other crops) mountainous countries like Nepal, watershed management can be produced with fewer purchased inputs and/or farm can contribute to important development goals and increase labor. The latter consideration may be particularly significant resilience to climate change. from a broader societal perspective. Farm households may need to haul less water, fertilizer, and fodder, or spend less Watershed management refers to a wide variety of practices time herding their livestock in search of fodder (Pandit, that fall under the umbrella of “investment in green Shrestha, and Bhattarai 2014). This may have substantial infrastructure”, such as slope correction using terracing, equity, as well as simply productive benefits, to the extent planting hedgerows and cover crops, using crop residues, that women, children, the elderly, or other disadvantaged cover crops, and mulches, trenching and bunding, re- and groups engage in these tasks. afforestation, and revision of grazing practices. Minimizing the loss of soil and downstream sedimentation is one of Water supply the most visible and immediate benefits of watershed There is abundant evidence that healthy watersheds provide management, the positive impact of which can be felt a suite of hydrologic ecosystem services, i.e., the benefits to across many sectors of the economy, including agriculture, people produced by ecosystem effects on freshwater systems hydropower, and water. These practices also help to regulate (Brauman et al. 2007). These hydrologic services include water flows, stabilize soils, maintain soil fertility, improve soil water purification, seasonal flow regulation, flood mitigation, water holding capacity, regulate water quality in downstream habitat protection, and provision of water-related cultural rivers, mitigate shallow to medium depth landslides, and services (Brauman et al. 2007; Postel and Thompson 2005). sequester carbon. They generate other on-site benefits to As water moves through a landscape, the physical conditions landholders such as fuelwood and fodder for livestock. The that affect its flow and recycling are affected by the condition multiple benefits of watershed management therefore accrue and structure of vegetation cover. Watershed management not only to the agriculture, energy, and water sectors, but also is therefore an important strategy for societies looking to have implications for disaster risk reduction, transportation, meet the needs of growing populations for clean and reliable and climate change mitigation. water supplies. Agriculture and rural development The water quality benefits of watershed management Soils store nutrients on which crops depend, so preventing interventions (such as retention of sediment and other nutrient loss both increases their availability to crops pollutants) are unambiguous, and there is strong evidence for being grown and reduces the necessity of applying other the importance of preserving natural vegetation to maintain fertilizers. A number of studies have demonstrated that existing hydrologic regulation services (Brauman et al. 2007). terraces, hedgerows, reduced tillage, and other practices However, the impacts on seasonal water flows and flood that prevent soil erosion also prevent soil nutrient losses, mitigation due to land management interventions such as thereby improving crop yields (see, e .g., Atreya et al. 2008; reforestation, afforestation, and best management practices Das and Bauer 2012). Similarly, porous and absorbent commonly adopted in croplands and rangelands varies with soils retain more moisture, having the effect of making local conditions, and the mechanisms are still hotly debated more water available to crops, and reducing the need to in the literature (Dennedy-Frank 2018; Filoso et al. 2017). Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  2 Energy Climate change mitigation Hydropower is a major source for strategic energy Managing watersheds through interventions that involve development in the Himalayan region, and a key sector the planting of trees (such as agroforestry), improving to promote sustainable economies. However, the efficient vegetation cover and soil health can increase both above- operation of hydropower is hindered by excessive and below-ground carbon pools as well as soil organic sedimentation that reduces the lifespan of reservoirs carbon. Sequestering more carbon in landscapes is a clear by decreasing storage capacity, while increasing short- win for watershed management activities, and also provides term operations costs and reducing generation efficiency. opportunities for co-financing from existing climate Hydropower therefore relies heavily on ecosystem services mitigation programs. More intense weather patterns due from watersheds and the sector has already begun to to climate changes have the potential to increase existing recognize the need for managing sediment production from problems of sedimentation even further, affecting, in landscapes as an integrated part of a sediment management turn, development outcomes for multiple sectors. Greater strategy (Annandale, Morris, and Karki 2016). investment in resource management, through integrated and targeted programs of watershed management, has the Roads potential to address these challenges. Well-managed watersheds can also contribute to maintaining Understanding the multiple benefits of watershed infrastructure, particularly roads, by reducing risks from management and how these benefits accrue to different erosion, landslides, and flooding (Mandle et al. 2016). Well- sectors is fundamental to designing effective programs that anchored vegetation above roads can reduce the risk of maximize return on investment. However, many of the landslides that cut off the flow of goods and people and economic benefits are hidden, as these watershed services result in significant costs for repairs. Preserving upstream are not transacted in the market, which leads in turn to catchments can mitigate flood risk, thereby reducing risk of under-investment in watershed management. In order to road washout. Understanding and managing the benefits of efficiently and sustainably manage these important assets, watershed management for roads and other transportation it is critical to quantify and value the many services that infrastructure can reduce costs by, for example, reducing watershed management can provide. the need for more costly engineering solutions to manage sediment and other risks. OBJECTIVES Disaster management and resilience The objectives of this study are to Landslides are both a major source of sediment in 1. Develop methodologies to value a range of ecosystem mountainous catchments and a major risk to life, property, services that come from watershed management, and and other assets that are located on unstable slopes. to demonstrate their application in the Kali Gandaki Landslides impose numerous social, environmental, and watershed to help create evidence on the value of green economic costs on affected areas, such as loss of life and infrastructure. property, damage to infrastructure, and economic impacts 2. Develop tools and demonstrate landscape-scale methods associated with loss of connectivity, particularly in remote to help practitioners target watershed management areas with limited road networks. interventions to improve effectiveness and reduce project costs. The maintenance and improvement of vegetation cover can help to stabilize slopes, slough off rain before it infiltrates, Since sediment retention is one of the most immediate and channel water away from vulnerable slopes, and increase soil visible impacts of watershed management activities, this strength (Collison, Anderson, and Lloyd 1995; Vanacker et study focused primarily on benefits that result from avoided al. 2003). Reducing the risk of landslides through watershed erosion and sedimentation and looked secondarily at some of management – where appropriate – can have downstream the co-benefits arising from activities that are used to control benefits, by reducing the amount of sediment reaching sediment. While proper watershed management is essential rivers, as well as local benefits, by avoiding loss of life and to maintaining water flow and quality, the quantification damages to infrastructure. and, particularly, valuation of these benefits requires greater detail of data than was available in this study. Those aspects are, therefore, left for future work. 3 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal STUDY AREA capacity in the reservoir (e.g. storage below the level of the lowest outlet, designed to trap excess sediment) was already This study focuses on the watershed area that drains to filled by the time the plant was operational due to the small the Kali Gandaki A Hydropower Plant (KGA), located reservoir volume and large monsoon sediments, and live just below the confluence of the Kali Gandaki and Aadhi storage (storage above the lowest outlet) has also declined rivers (see Figure 1). KGA, operated by the Nepal Electricity over KGA’s operation (Morris 2014). Authority (NEA) and built at a cost of about US $350 M (ADB 2012), is currently the largest power plant in Nepal The Department of Forests and Soil Conservation (DoFSC), with an installed capacity of 144 MW. Since it became Ministry of Forests and Environment, Government of Nepal, operational in 2002, the plant has experienced multiple has been investing in watershed management (referred issues caused by sedimentation, including turbine erosion to as “catchment area treatment”) activities for decades. due to the abrasion from inflowing sediment combined with These investments typically involve practices to prevent cavitation, leading to frequent repairs (an overhaul every 3 erosion (such as cover cropping or inter-cropping with fruit years) and unplanned shutdowns. In addition, dead storage trees in cultivated lands), to reduce overland flow, promote  Figure - 1: Study area - Kali Gandaki watershed, with location of hydropower facility Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  4 infiltration, and prevent or mitigate landslide damages infrastructure approaches for watershed management, (such as terracing, contour trenches, or tree planting), and and evaluate their impacts. The economic benefits of a to capture sediment in runoff (such as hedgerows or check targeted program of watershed investments are evaluated, dams). However, the DoFSC programs are focused on a and the results are used to develop cost-effective watershed single priority sub-watershed at a time, interventions are management investment portfolios to achieve multiple highly localized, and not targeted to maximize the flow of ecosystem service objectives. Ecosystem services benefits ecosystem services. analyzed included the following According to the 2011 national census, there are Downstream benefits, with a focus on reduced sediment approximately 590,000 people living in the watershed area arriving at KGA: (Government of Nepal 2013). Cultivation is the main source • Reductions in damage to equipment, efficiency loss, and of income for residents, and most agricultural activity occurs need for repairs in the southern foothills of the watershed on very steep • Reduced costs of desanding and preventative measures slopes, with the mean gradient of farmland being 41%, • Maintenance of storage capacity for peaking and little farmland on slopes with less than 5%. The steep slopes and high precipitation require that most croplands Local benefits, arising from the reduction in landslide risk: are converted to an elaborate system of terraces to control • Avoided lives lost erosion and manage water on the hillslopes. Labor migration • Avoided cost of replacing structures away from the hills is common (Jaquet, Kohler, and Schwilch • Avoided cost of road repairs 2019), often leaving behind terraces that are abandoned, and which may be more prone to erosion. Fragile geology, Global benefits: naturally high levels of erosion and mass movements make • Carbon sequestration from improving or preserving the issues of erosion and sedimentation of high priority and vegetation cover and enhancing soil carbon mean that this area is particularly vulnerable to impacts of land management. METHODS Establish baseline sediment budget: A set of newly ASSESSMENT APPROACH collected field data on sediment concentrations was used The study presents a systematic approach to assess where, with an existing InVEST model for erosion and sediment in what quantity, and through what processes sediment transport, along with novel approaches to estimate the is being generated in the Kali Gandaki watershed, contribution of roads, landslides, and glacial erosion to total identify plausible interventions through investing in green sediment loads.  Figure - 2: Workflow used in this study to evaluate watershed management activities, value their impacts, prioritize intervention locations and estimate the benefit: cost at different levels of investment Establish Plausible Evaluate Assess Prioritize baseline ranges of impacts of marginal & develop sediment intervention activities value optimal budget impacts portfolios Physical models: Literature-based Physical models: Economic analysis: Effectiveness • Erosion estimates • Erosion & • Energy • Landslides sedimentation • Landslide risk Benefit/cost of • Roads Stakeholder • Landslides • On-site benefits investment • Glaciers consultation • Carbon • Carbon 5 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal To assist in developing the sediment budget, monitoring Identify plausible interventions and range of data at multiple locations in the watershed were collected impacts: A combination of literature review and stakeholder over a period of one year. These data sets, along with flow consultation was used to select activities that are feasible and data obtained from the Nepal Department of Hydrology suitable to local conditions, and to provide estimates of their and Meteorology, were used to develop a 5-year record of costs and effectiveness. Activities modeled include terrace sediment contributions from the different sub-watersheds of improvements, soil and water conservation practices (such the Kali Gandaki (Figure 3). Models were calibrated to this as hedgerows, cover crops, agroforestry), reclamation of sediment record, and the resulting model performance was degraded forests and rangelands, and landslide mitigation acceptable (Figure 4). practices (such as revegetating denuded slopes and slope  Figure - 3: Topography of the Kali Gandaki watershed, location of gauging stations and their respective drainage area Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  6 correction). While roads are included as a source of sediment generated in the watershed and impacted by watershed in much of the study area, we do not model any interventions management interventions were adjusted to account that affect road erosion, as the engineering solutions required for the balance of long-term sediment deposition and to manage sediment from roads were outside the scope of what re-mobilization in stream channels and deposition in are normally considered “green infrastructure” interventions. KGA’s reservoir. Costs for interventions modeled range from about US $880 per hectare for rangeland rehabilitation to over $39,000 per Assess marginal value of sediment reductions: A hectare for mitigation of landslide-prone areas. combination of micro-economic modeling, spatial overlays, and qualitative methods was employed to evaluate the value Evaluate potential impacts of activities: The impacts of implementing watershed management practices that of activities on ecosystem services were evaluated using reduce sediment and landslide risk, provide local benefits to the biophysical models mentioned above, to determine landholders, and store carbon. the location-specific benefits of activities in every possible location. For each type of watershed management activity, Reductions in sediment to the KGA facility were valued impacts on hillslope erosion and landslide risks were based on the avoided cost of damages, as well as measures evaluated. In landslide-prone areas that were treated, the to prevent damage and avoided loss of reservoir storage for impacts of treatment on the overall risk to lives, buildings, generating electricity during peak times. This study does and roads was evaluated using data on the locations of not attempt to quantify avoided cost of damages to the infrastructure in the study area. Estimates of sediment Modi Khola (14.8 MW) and the Lower Modi 1 (10 MW)  Figure - 4: Comparison of observed sediment load and results from the calibrated multi-model suite. Colors indicate the total sediment contribution from each component of the sediment budget (glaciers, hillslope erosion/SDR, landslides, and road erosion). Error bars indicate ± 1 standard deviation in observed loads X106 12 Glaciers SDR 10 Landslides Roads Observations 8 Standard deviation Sediment load [t/yr] 6 4 2 0 som l ani eni at aki u yap alg top dib nd Jom ngh Na liga Ta Mo Ma Ka 7 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal hydropower facilities, located upstream of KGA on the RESULTS Modi tributary. The impacts of landslide mitigation were valued based on the avoided loss of lives (using a value of Results for watershed management portfolios ranging from statistical life approach), avoided loss of structures (using US $500,000 to $50M show that such programs can have a average rental rates in rural areas), and avoided costs of significant, positive impact across many sectors. The benefits road repair (using average road repair costs). The impacts are driven largely by local benefits and the value of avoided on carbon storage were valued using the social cost of lives lost in landslides, with the next highest beneficiary carbon; the benefits to local landholders (such as improved being downstream hydropower (Figure 5). At the $500,000 soil fertility, crop production, and water regulation) were budget level, each $1 invested yields $4.38 in benefits, but valued based on the average cost-share reported for similar this ratio drops as budgets are increased. However, even programs and locations. with an investment of $50M, the program still has a positive benefit: cost ratio, even without considering the carbon Prioritize intervention scenarios: Estimates of sequestration benefit. implementation costs and modeled effectiveness of each activity/location were used to identify optimal portfolios Figure 6 shows the benefit: cost ratio of the modeled of interventions at different budget levels. Objectives for portfolios of interventions, including high and low bounds prioritizing activities, the associated beneficiaries, and on the estimated total benefits. These bounds are based the valuation approaches employed are summarized in on potential values for each benefit stream using a range Table 1. of parameter estimates in the economic valuation models. Table - 1: Objectives used to prioritize watershed management activities and locations activities and locations Objective Unit Beneficiary Valuation approach On-farm benefits of soil Tons of sediment/ Local landholders Revealed preference based retention yr on reported cost-share from similar programs Avoided sediment reaching Tons of sediment/ KGA hydropower plant Avoided damage Kaligandaki reservoir yr Avoided costs of desanding Peaking capacity maintained Avoided lives lost from USD People at risk from landslides Value of statistical life landslides Avoided damages to structures USD Communities at risk from Rental rate landslide damages to structures Avoided repairs to roads USD Dept of Roads, VDCs and Avoided repair costs communities at risk from landslide damages to roads Added carbon storage Metric tons National (e.g. REDD+ Social cost of carbon in 2020 program), Global Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  8  Figure - 5: The multiple values of watershed management. The benefits are driven largely by local benefits and the value of avoided lives lost in landslides, with the next highest beneficiary being downstream hydropower (KGA) $70 $60 USD, Millions $50 Carbon $40 Lives Roads $30 Structures $20 Hydropower Landholder benefits $10 $ $ 0.5M$ 1.0M$ 2.0M$ 3.0M$ 5.0M$ 7.0M$ 10.0M$ 20.0M$ 50.0M Budget  Figure - 6: Benefit/cost ratio of modeled portfolios (blue points), showing high and low boundaries on estimates (lines). High and low bounds are based on potential values for each benefit stream based on a range of parameter estimates as explained in the main report text. These ranges should be considered illustrative and are not to be interpreted as confidence intervals 9 8 7 6 Benefit/Cost 5 4 3 2 1 0 0.5 1 2 3 5 7 10 20 50 Budget (millions USD) 9 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal These ranges illustrate that the positive economic benefit of major benefit streams, was employed using watershed- of watershed management interventions is relatively robust and region-specific data to rigorously evaluate these to model assumptions but should not be interpreted as benefits. In this way, our study goes beyond the often-used confidence intervals. approach of simply transferring area-based estimates of the value of watershed benefits from one region to another, Some of the largest benefits valued in this study are those and represents a proof-of-concept for how such approaches relating to landslide risk reduction. Figure 7 shows the results may be applied in other contexts. of the landslide risk model for an area in the central Kali Gandaki watershed, overlaid with data on the locations of Conservative assumptions were applied throughout structures and roads. Reddish areas show the parts of the the economic analysis; even so, the results show that hillslope that are prone to failure. The different shades of the aggregated benefits of such a program can greatly red indicate that different connected parts of a hillslope will outweigh the costs. The benefits to cost ratio is highest fail with different probabilities. The brown cells indicate at lower investment levels and decreases to 1.2 with a modeled runout paths that begin at the downslope end of US $50M investment. There is both a physical limit each landslide. Runout paths are colored in different shades and a feasibility limit as to how much can be achieved of brown, according to the failure probability of the landslide with watershed management alone, using the types from which they originate. Note that many houses and roads of practices evaluated in this study. But as part of a are located in areas of medium risk of being impacted by a comprehensive sediment strategy that includes land landslide or resulting runout. Overlaying the infrastructure management improvements, structural sediment at risk allows for an estimation of the total infrastructure at mitigation approaches, reclamation of degraded lands, risk in the watershed, and the value of reducing that risk and best practices for road engineering, our results show through watershed management activities. The combined that a data-driven and targeted program of watershed values of avoided lives lost, avoided damages to structures, management can contribute greatly to a broader social and avoided road repairs, comprise between 25 and 75% benefit through real and significant economic gains of the total value of benefits from the modeled watershed to society. management interventions (depending on the budget level). The results highlight the importance of considering multiple Finally, watershed management activities can be prioritized benefit streams and sources of value to make the case that based on different objectives, which will impact where investments in watershed services are sound. With the investments should be focused. Figure 8 illustrates the exception of the benefits from landslide mitigation1, no potential trade-off between prioritizing activities for local one sector receives enough benefits to justify 100% of the versus downstream benefits. Those portfolio maps show that investment cost, and in some cases targeting investments to when downstream sediment is the primary focus, reducing benefit one sector will reduce the benefits accrued to other sediment through mitigating mass movement in landslides sectors. Mapping and quantifying the sources of sediment along the main stem and tributary channels are frequently and benefit pathways will help policymakers to design the preferred options. However, when local erosion is equitable programs that distribute the costs of sediment the main concern, the focus shifts more toward terrace management across different actors who receive benefits, improvement, grazing land and forest rehabilitation in the and that address conservation and development goals as well middle hills area. as the need for sustainable energy and rural development. As with any study that relies on physically-based models and CONCLUSIONS extrapolates landscape-scale effects from local data, there This study presents a novel attempt to generate a are uncertainties inherent in the analysis. Every attempt has comprehensive valuation of the multiple benefits that been made to use the best available data, vetted through a can result from implementing a watershed management stakeholder engagement process. Errors in the underlying program to control erosion and sedimentation in the data on topography, historical climate, streamflow and Kali Gandaki watershed. A physically-based modeling sediment concentrations, and uncertainties about the costs approach, in combination with micro-economic modeling and characteristics of watershed management practices The total benefits from reducing landslide risks (value of avoided lives lost, avoided loss of structures and avoided road repairs) is greater than the cost 1. of implementation only up to a budget of about US $5M. Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  10  Figure - 7: Results of the landslides risk and runout modeling for an area on the middle Kali Gandaki River (small cutout for location). Red colors indicate landslide probability, and brown colors indicate runout probability. The darkest colors show areas with the highest probability of failure. Green squares are structures and black lines are roads. This overlay reveals that many of these infrastructures are located in areas of medium landslide risk and/or high probability of being impacted by landslide runout 11 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  Figure - 8: Intervention portfolios at a budget of US $20M optimized for two competing objectives (left column: downstream sediment for hydropower and right column: local erosion reduction). Note that different activities and sub-watersheds are chosen for implementation to meet the different objectives in the two scenarios US $20M portfolio, optimized to reduce sediment downstream US $20M portfolio, optimized to reduce local erosion as implemented in specific and varying locations on the watersheds is therefore a policy challenge, and one that can ground means that the results of this study should be taken be informed by the types of information provided by this as demonstrative, rather than definitive. However, this study: e.g., where watershed management practices provide study overall is conservative in its assumptions and thus greatest overall economic benefits and how these benefits provides evidence that watershed management can have accrue to different sectors. Such a systematic approach positive economic benefits that greatly exceed the costs of allows for further engagement with different sectors to align its implementation. interests and leverage resources. It is worth noting that the benefits accruing to landholders are The agriculture, forestry, and water sectors can use a large fraction of the total benefits. The assertion that better this valuation methodology to make a case for why watershed land-management practices might provide such benefits to management programs are good investments. Understanding the landholders implementing them may beg the question of and quantifying the benefits that accrue to different sectors why they have not already been adopted. Reasons for this can enables the design of more efficient and robust payments for include lack of access to capital, lack of information, or the ecosystem services (PES) schemes and can leverage investment fact that the cost of implementing improvements may equal from multiple actors. The transportation and disaster or exceed the private marginal benefit of their adoption. risk management sectors can apply the landscape-scale Aligning the incentives for landholders with broader societal hazard mapping developed in this study to estimate the goals for improving the value of ecosystem services from exposure of assets such as roads, at a finer spatial resolution Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  12 than is currently available from landscape-scale screening Overall, the methods and data resulting from this study analyses. Further, the prioritization tools can be used to demonstrate why effective and efficient targeting is key to identify areas of particular risk that may require a higher achieving the greatest benefits at the lowest costs. Across standard of impact assessment and/or consideration of all of these sectors, the use of watershed-scale tools to cumulative (rather than project-specific) impacts on ecosystem evaluate and integrate the multiple benefits of watershed services. The hydropower sector can use the valuation management into sectoral and cross-sectoral policy and and prioritization methodologies to design PES schemes planning can be used as a strategic tool to build resilience that more effectively control sediment from watersheds. The as climate change impacts are increasingly felt. Further, tools also have relevance for environmental and social the stakeholder-driven process employed here allows for safeguards, by providing a data-driven and systematic way more durable and sustainable solutions, and the science- to incorporate ecosystem services impacts into environment based, landscape-level assessment uncovers the underlying management plans and to identify mitigation opportunities drivers of problems instead of focusing only the individual, to offset project impacts to ecosystem services. localized results of such problems. 13 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal 1. INTRODUCTION © David Cutler 1.1. THE PROBLEM planting hedgerows and cover crops, using crop residues, cover crops, and mulches, trenching and bunding, re- and Watersheds are an appropriate and effective unit for afforestation, and revision of grazing practices. Minimizing managing ecological assets, given the interconnected nature the loss of soil and downstream sedimentation is one of of economic activities and their impacts within a watershed, the most visible and immediate benefits of watershed locally and regionally, upstream and downstream. Watersheds management the positive impact of which can be felt are increasingly recognized as a critical form of green across many sectors of the economy, including agriculture, infrastructure that provides a flow of economic benefits. In hydropower, and water. These practices also help to mountainous countries like Nepal, watershed management regulate water flows, stabilize soils, maintain soil fertility, can contribute to important development goals and increase improve soil water holding capacity, regulate water quality resilience to climate change. in downstream rivers, mitigate shallow to medium depth landslides, and sequester carbon. They generate other on- Watershed management refers to a wide variety of practices site benefits to landholders such as fuelwood and fodder for that fall under the umbrella of “investment in green livestock. The multiple benefits of watershed management infrastructure”, such as slope correction using terracing, therefore accrue not only to the agriculture, energy, and Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  14 water sectors, but also have implications for disaster risk habitat protection, and provision of water-related cultural reduction, transportation, and climate change mitigation, services (Brauman et al. 2007; Postel and Thompson 2005). as described below. As water moves through a landscape, the physical conditions that affect its flow and recycling are affected by the condition Agriculture and rural development and structure of vegetation cover. Watershed management is therefore an important strategy for societies looking to Soils store nutrients on which crops depend, so preventing meet the needs of growing populations for clean and reliable nutrient loss both increases their availability to crops being water supplies. grown and reduces the necessity of applying other fertilizers. A number of studies have demonstrated that terraces, The water quality benefits of watershed management hedgerows, reduced tillage, and other practices that prevent interventions (such as retention of sediment and other soil erosion also prevent soil nutrient losses (see, e .g., Atreya pollutants) are unambiguous, and there is strong evidence for et al. 2008; Das and Bauer 2012). Agricultural models then the importance of preserving natural vegetation to maintain predict that future crop yields will be improved as a result existing hydrologic regulation services (Brauman et al. 2007). (Das and Bauer 2012). However, the impacts on seasonal water flows and flood mitigation due to land management interventions such as Similarly, porous and absorbent soils retain more moisture, reforestation, afforestation, and best management practices having the effect of making more water available to crops, commonly adopted in croplands and rangelands varies with particularly between rainfalls, and reducing the need to local conditions, and the mechanisms are still hotly debated acquire and transport water from other places. Some in the literature (Dennedy-Frank 2018; Filoso et al. 2017). measures that are put in place to retain soils also have beneficial side-effects. Hedgerows and plantings may be harvested for food, fodder, or other products. They may Energy also serve as windbreaks, providing protection against the Hydropower is a major source for strategic energy elements, and detaining or diverting floodwaters. Another development in the Himalayan region, and a key sector reason for building terraces, as well as hedgerows, bunds, to promote sustainable economies. However, the efficient and other measures that may have the effect of establishing operation of hydropower is hindered by excessive terrace-like features over time, is that they provide a more sedimentation that reduces the lifespan of reservoirs by level surface that is more easily worked than a sloping one decreasing storage capacity, while increasing short-term (Thapa and Paudel 2002; Bhattarai 2018) operations costs and reducing generation efficiency. For reservoir-based hydropower projects, excessive sedimentation These physical effects translate into economic and societal causes a loss of storage capacity and reduces the effective benefits. The implications of higher soil fertility and more lifespan of the reservoir or increases operations costs by reliable water availability is that more food (or other crops) requiring expensive dredging to be carried out. Run-of- can be produced with fewer purchased inputs and/or farm river, or diversion, hydropower projects are also common in labor. The latter consideration may be particularly significant this region. These projects face increased wear and tear of from a broader societal perspective. Farm households may electro-mechanical as well as structural components when need to haul less water, fertilizer, and fodder, or spend less incoming sediment levels are too high. Hydropower therefore time herding their livestock in search of fodder (Pandit, relies heavily on ecosystem services from watersheds and the Shrestha, and Bhattarai 2014). This may have substantial sector has already begun to recognize the need for managing equity, as well as simply productive benefits, to the extent sediment production from landscapes as an integrated part that women, children, the elderly, or other disadvantaged of a sediment management strategy (Annandale, Morris, groups engage in these tasks. and Karki 2016). Water supply Roads There is abundant evidence that healthy watersheds provide Well-managed watersheds can also contribute to maintaining a suite of hydrologic ecosystem services i.e., the benefits to infrastructure, particularly roads, by reducing risks from people produced by ecosystem effects on freshwater systems erosion, landslides, and flooding (Mandle et al. 2016). Well- (Brauman et al. 2007). These hydrologic services include anchored vegetation above roads can reduce the risk of water purification, seasonal flow regulation, flood mitigation, landslides that cut off the flow of goods and people and 15 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal result in significant costs for repairs. Preserving upstream under-investment in watershed management. In order to catchments can mitigate flood risk, thereby reducing risk of efficiently and sustainably manage these important assets, road washout. Understanding and managing the benefits of it is critical to quantify and value the many services that watershed management for roads and other transportation watershed management can provide. infrastructure can reduce costs by, for example, reducing the need for more costly engineering solutions to manage This study presents a novel approach to comprehensively sediment and other risks. value a variety of benefits that can be achieved with a watershed management program aimed to reduce erosion Disaster management and resilience and sediment loss. The analysis shows that even under conservative assumptions, the benefits of a data-driven and Landslides are both a major source of sediment in targeted program of watershed management can outweigh mountainous catchments and a major risk to life, property, the costs. and other assets that are located on unstable slopes. Landslides impose numerous social, environmental, and economic costs on affected areas, such as loss of life and 1.2. PURPOSE AND OVERVIEW OF property, damage to infrastructure, and economic impacts associated with loss of connectivity, particularly in remote THIS STUDY areas with limited road networks. The objectives of this study are to The maintenance and improvement of vegetation cover can 1. Develop methodologies to value a range of ecosystem help to stabilize slopes, slough off rain before it infiltrates, services that come from watershed management, and channel water away from vulnerable slopes, and increase soil to demonstrate their application in the Kali Gandaki strength (Collison, Anderson, and Lloyd 1995; Vanacker et watershed to help create evidence on the value of green al. 2003). Reducing the risk of landslides through watershed infrastructure. management – where appropriate – can have downstream 2. Develop tools and demonstrate landscape-scale methods benefits, by reducing the amount of sediment reaching to help practitioners target watershed management rivers, as well as local benefits, by avoiding loss of life and interventions to improve effectiveness and reduce damages to infrastructure. project costs. Climate change mitigation Since sediment retention is one of the most immediate and Managing watersheds through interventions that involve visible impacts of watershed management activities, this the planting of trees (such as agroforestry), improving study focused primarily on benefits that result from avoided vegetation cover and soil health can increase both above- erosion and sedimentation and looked secondarily at some of and below-ground carbon pools as well as soil organic the co-benefits arising from activities that are used to control carbon. Sequestering more carbon in landscapes is a clear sediment. While proper watershed management is essential win for watershed management activities, and also provides to maintaining water flow and quality, the quantification opportunities for co-financing from existing climate and, particularly, valuation of these benefits requires greater mitigation programs. More intense weather patterns due detail of data than was available in this study. Those aspects to climate changes have the potential to increase existing are, therefore, left for future work. problems of sedimentation even further, affecting, in turn, development outcomes for multiple sectors. Greater This report begins with a systematic, watershed-level investment in resource management, through integrated assessment of sediment sources in the Kali Gandaki and targeted programs of watershed management, has the catchment area. We use newly available data and models to potential to address these challenges. consider not only sources of sediment but also how sediment moves across the landscape and in rivers and is finally Understanding the multiple benefits of watershed deposited into the reservoir. We present a novel method for management and how these benefits accrue to different assessing sediment contribution from landslides, and the sectors is fundamental to designing effective programs that potential for watershed management activities to mitigate maximize return on investment. However, many of the the risk of landslides to lives, roads and built structures. economic benefits are hidden, as these watershed services We then assess the potential for watershed management are not transacted in the market, which leads in turn to to mitigate sediment sources and present an economic Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  16 valuation of the benefits of such activities in terms of carbon management can have positive economic benefits that sequestration, energy generation, reservoir storage capacity, greatly exceed the costs of its implementation. operations and maintenance costs, the avoided loss of lives and the avoided costs of replacing structures and repairing The rest of the report is organized as follows: chapter 2 roads due to landslide risk mitigation. Further, we discuss the provides information about the Kali Gandaki watershed. potential for additional co-benefits that can accrue locally, Next, in chapter 3, methods are presented to (1) develop a such as improved soil fertility and soil moisture, local water detailed sediment budget for the watershed; (2) model benefits regulation, and crop productivity. and values of watershed management to a suite of ecosystem services; and (3) prioritize where, in a watershed, activities Every attempt has been made to use the best available should be focused to maximize impacts while minimizing cost. data, vetted through a stakeholder engagement process. Chapter 4 presents the results of the analysis, in terms of both As with every modeling study, necessary assumptions and physical changes brought about by implementing watershed simplifications are made to enable analysis at a watershed management as well as the economic benefits that accrue to scale. Errors in the underlying data on topography, different sectors at a range of budget levels. The implications historical climate, streamflow and sediment concentrations, of prioritizing activities to achieve different objectives are and uncertainties about the costs and characteristics of discussed with some illustrative examples. Finally, chapter watershed management practices as implemented in specific 5 draws out the main findings of the study and makes and varying locations on the ground means that the results recommendations as to how the findings can be used by the of this study should be taken as demonstrative, rather than different benefiting sectors – agriculture, roads, hydropower, definitive. However, this study is overall conservative in its disaster management – and outlines future work to improve assumptions and, thus, provides evidence that watershed the data and technical basis of these estimates. Box 1.1: Sedimentation issues and approaches in the Kali Gandaki watershed The Kaligandaki A Hydropower Plant (KGA), operated by the Nepal Electricity Authority (NEA) and built at a cost of about US $350 M (ADB 2012), is the largest power plant in Nepal with an installed capacity of 144 MW. Since its opening in 2002, KGA has been the largest single generator in the country, providing a quarter or more of the power generated by NEA assets (NEA 2013; NEA Annual Reports 2009–18). The steep-sided gorge of the Kali Gandaki river afforded a good site for locating a hydroelectric facility (World Bank 2013), but the need to control sediment was recognized from the earliest planning stages (ADB 2012). KGA’s design incorporated two large desanding basins that were intended to remove most of the coarse, abrasive sediment that could damage turbines and other equipment (ADB 2012; IHA 2017). However, KGA has suffered greater losses in terms of damage to equipment, loss of efficiency, and more frequently required maintenance than had been anticipated (ADB 2012; World Bank 2013; Morris 2014). Since it became operational in 2002, the plant has experienced multiple issues caused by sedimentation, including turbine erosion due to the abrasion from inflowing sediment combined with cavitation, leading to frequent repairs (an overhaul every 3 years) and unplanned shutdowns. In addition, dead storage capacity in the reservoir (e.g. storage below the level of the lowest outlet, designed to trap excess sediment) was already filled by the time the plant was operational due to the small reservoir volume and large monsoon sediments. Sediment accumulation affects operations by restricting the plant’s ability to meet peak demand. KGA was designed with over 3 million cubic meters of live storage volume in the reservoir behind the dam (Morris 2014). From roughly June through October, monsoon rains and melting snow and ice from the high Himalayas generate water flow in the Kali Gandaki much greater than needed to generate at full power. From November until May, however, the flow declines to a relative trickle. As each kilowatt hour is more valuable during periods of peak, as opposed to off-peak, demand, the water in storage is used to generate more power during those hours of the day when it is most valuable. 17 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal Box 1.1 (Contd.): Sedimentation issues and approaches in the Kali Gandaki watershed Sedimentation also results in loss of reservoir volume (Word Bank 2013; Morris 2014; more generally, see Førsund 2009). NEA officials have estimated that, at times, as much as half of designed peaking capacity may have been lost. Capacity losses may be reduced and, in some instances, reversed by management measures such as maintaining lower water heights, and hence a lower residence time in the reservoir, during high flow periods, or flushing by opening the floodgates (Morris 2014; World Bank 2013). Such measures, however, also impose costs. A lower operating level implies loss of hydraulic head and, consequently, lower power production; flushing implies forgone power production during the period that water is diverted through the floodgates. In an effort to address the issue of sedimentation, the World Bank initiated a US $30M project in 2013 to revamp the civil and electro-mechanical works at the plant, as well as to provide technical assistance and build capacity. As part of that project, funding was provided to consider whether and how changes in land management in the watershed might affect sediment delivery (World Bank 2018b). That study made recommendations for priority locations to invest in various sediment management practices, including land management interventions, structural interventions, and mitigating impacts from road construction. The current study goes a step further in developing a rigorous cost-benefit analysis to understand the economic value of sediment reductions to KGA, among other benefits. The Department of Forests and Soil Conservation (DoFSC), Ministry of Forests and Environment, Government of Nepal, has been investing in watershed management (referred to as “catchment area treatment”) activities for decades. Management practices employed in the study area (Figure 2.1) typically involve practices to prevent erosion (such as cover cropping or inter-cropping with fruit trees in cultivated lands), to reduce overland flow, promote infiltration, and prevent or mitigate landslide damages (such as terracing, contour trenches, or tree planting), and to capture sediment in runoff (such as hedgerows or check dams). However, the DoFSC programs are focused on a single priority sub-watershed at a time, interventions are highly localized, and not targeted to maximize the flow of ecosystem services. The watershed management personnel of DoFSC and the District Soil Conservation Offices have a deep knowledge of on- the-ground issues with sediment management, and they have detailed norms for designing and implementing watershed management interventions to address specific problems. The recently formed Gandaki Basin Management Centre aims to provide a centralized knowledge platform for data, guidance, and best practices on watershed management in this region, but they currently lack a systematic approach to model and assess impacts of potential activities at the scale of this study. Such a landscape-scale approach to assessment and targeting would make the best use of the in-depth knowledge that does exist on how to design and implement effective interventions. Further, the DoFSC’s current expenditure (less than US $100,000 allocated each year) is extremely small compared to the scale of the problem (over 30 million tons of sediment coming down the Kali Gandaki River, on average, each year), as is demonstrated later in this report. Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  18 2. STUDY AREA © Martin M303/Shutterstock.com 2.1. SOCIO-ECONOMIC to around 1200 m) but also occurs up to 2000 m on the southern slopes of the Himalayas. CHARACTERISTICS This study focuses on the watershed area that drains to the Higher altitudes show a transition to forest, bushland, and Kali Gandaki A Hydropower Plant (KGA), located just grassland that is used for grazing and collection of fuel below the confluence of the Kali Gandaki and Aadhi rivers wood. Agriculture takes place on very steep slopes, with the (Figure 2.1). mean gradient of farmland being 41%, and little farmland on slopes with less than 5%. The steep slopes and high The distribution of settlements, livelihoods and infrastructure precipitation require that most croplands are converted to an in the Kali Gandaki watershed reflects the geographic elaborate system of terraces to control erosion and manage variability of the area. Cultivation is the main source of water on the hillslopes, as shown in Figure 2.3. income for residents and most agricultural activity occurs in the southern foothills, where a majority of the roads On the northern side of the Himalayas, agricultural activity and villages are also located (Figure 2.2). Agriculture is is limited to small pockets of farming on the alluvial plains the dominant land use for the lower elevation range (500 of the Mustang plateau as a result of very high altitude and 19 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  Figure - 2.1: Study area - Kali Gandaki watershed, with location of hydropower facility low rainfall. The valley bottom of the Mustang Plateau is Labor migration away from the hills is common (Jaquet, covered in grassland, supporting grazing in many areas, Kohler, and Schwilch 2019), often leaving behind terraces which is a primary source of income in the high mountains that are abandoned, which may be more prone to erosion. (Aryal, Maraseni, and Cockfield 2014). This grassland Naturally high levels of erosion in the Himalayas are reaches up to about 4500 m. compounded by a lack of integrated spatial planning and development, leading to widespread degradation of According to the 2011 national census, there are forest cover and loss of fertile soils. Extreme topography approximately 590,000 people living in the watershed area and climate also contribute to erosion where terraces are (Government of Nepal 2013). Most settlements are located managed in a sub-optimal way. In addition, rural road- at lower elevations to the south, and there are a few small building has increased, expanding transportation options villages in the Mustang area. Villages in the lower watershed and accessibility. But often these roads are built hastily, are connected with a dense network of roads, while several using cut and throw practices, on steep slopes and without major highways follow the course of the Kali Gandaki river stabilization methods (Figure 2.5), and this is reported to be (Figure 2.4). another major source for sediment through thrown soil and resulting landslides (Shrestha 2009). Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  20  Figure - 2.2: Department of Survey land use/land cover map from year 2000, also showing the locations of settlements Table - 2.1: Land uses and their total areas found in the  Figure - 2.3: Example of agricultural terracing on a steep Kali Gandaki watershed slope in the Kali Gandaki watershed Land use Area (ha) Percent of total Barren/Cliff 268,100 35.4 Sand 9,900 1.3 Built up 300 < 1.0 Bush 36,700 4.8 Cultivation 106,100 14.0 Forest 155,300 20.5 Glacier 17,300 2.3 Grass 156,700 20.7 Orchard/Nursery 1,100 < 1.0 Snow 1,100 < 1.0 Waterbody 4,300 < 1.0 © David Cutler TOTAL 756,900 100% 21 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  Figure - 2.4: The lower Kali Gandaki watershed, where many settlements are connected by a network of rural roads 2.2. WATERSHED CHARACTERISTICS  Figure - 2.5: Example of rural road construction, cut into steep hillsides without mitigation measures, The watershed of KGA covers 7600 km2 and is characterized increasing the chance of erosion and by a very high spatial variability of geology, climate, and landslides. altitude (ranging from 8144 m - 525 m) resulting in variable geomorphic and hydrologic processes. The Kali Gandaki river originates from the Mustang Plateau on the Chinese- Nepali border (Figure 2.1). The river flows southwards through the Mustang Plateau, which is characterized by its high altitude (> 4000 m) and very low precipitation. At the southern end of the Mustang Plateau, the Kali Gandaki cuts through the main range of the Himalayas, in between the Dhaulagiri and Annapurna massifs, forming a deep and narrow gorge between the two rapidly uplifting mountain ranges. This part of the watershed is herein referred to as Upper Kali Gandaki. On the southern, and specifically south-eastern slopes of the Himalayan main range, referred to as Middle Kali Gandaki, very high annual rainfalls are observed reaching up to 5000 mm/yr. From these southern slopes, the river, referred to as Middle © David Cutler Kali Gandaki from here on, flows through lesser Himalayas. Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  22 The Middle Kali Gandaki receives one major tributary, the Khola and lower Kali Gandaki) are all in the range of 2000 Myagdi River, which drains from the Dhaulagiri massif – 3000 t/km2/yr. This means that a square kilometer of to the west. The lower Kali Gandaki River is delineated land in the Upper or Middle Kali Gandaki will in average upstream by the confluence of the Middle Kali Gandaki produce five times more sediment than a square kilometer with the Modi River, which drains the Annapurna massif to of land in the Mustang Area. The spatial heterogeneity in the east, and downstream by the KGA reservoir, located just sediment yield is the results of differences in geology, uplift below the confluence of the Kali Gandaki and Aadhi Rivers. rates, and precipitation between sub-watersheds. Hence, six distinct sub-watersheds can be delineated within the Kali Gandaki watershed, each with a gauging station at These findings also highlight the need for spatially distributed its outlet (see Figure 2.6 and Table 2.2). sediment measurements to determine sediment origins in the watershed, and the need for adopting a landscape-scale perspective on sediment management supported by such 2.3. BASELINE SEDIMENT measurements. For example, most of the sediment load MONITORING arriving in KGA is derived from areas that are relatively far away from KGA (Figure 2.6 and Figure 2.7). Information To model where and how sediment is generated, sediment on sediment yield cannot easily be inferred nor generalized monitoring data at multiple locations throughout the from readily available (e.g., global) data. For example, the watershed are needed. Such data sets are not available sub-watersheds of Modi and Myagdi Rivers cover a similar in Nepal, and therefore the study included a sediment area and elevation range than the middle Kali Gandaki monitoring campaign. Specifically, sediment delivery but contribute less than half of the sediment of the middle from the sub-watersheds was monitored by a team from Kali Gandaki. Kathmandu University over a period of one year in 2018 – 2019 (Kafle and Bhandari 2019). Samples of suspended Results also allow for estimation of the total sediment sediment, i.e., sand, silt and clay, were taken every two delivery to KGA. The sampling results show that the total weeks at the gauging stations in each sub- watershed, except load of suspended sediment, i.e., sand and finer, arriving for the station at the reservoir of KGA itself (Figure 2.6 at KGA is 31.7 ± 4.9 Mt/yr, similar to what is reported and Appendix 1). Monitoring the bedload of gravel and based in other studies (Struck et al. 2015). The current coarser fractions was not part of the sampling campaign, sampling only considered fine sediment transported in so estimates given below are for fine sediments (sand and suspension (i.e., sand and finer). The Kali Gandaki River finer). Results allow to determine the sediment load from also transports a significant amount of coarse material the sub-watersheds of the Mustang Plateau, the Upper Kali (pebbles, cobbles, boulders) as bed load, which is likely Gandaki, the middle Kali Gandaki, and the Myagdi Khola in the range of 10 – 20 % of the fine load (Turowski, and Modi Khola tributaries. The characteristics of the area Rickenmann, and Dadson 2010). Unmonitored bed load draining to each gauging station are shown in Table 2.2, and thereby adds another 3.1 to 6.2 Mt/yr to the sediment the sampling locations and sub-watersheds are identified on budget at KGA. What cannot be determined from the the map in Figure 2.6. We used a rating curve approach to derived suspended load measurements is which processes, develop a longer-term sediment budget for the Kali Gandaki e.g., landslides, road erosion, or glaciers, generated the watershed covering 2009-2015 (see Box 2.1). This approach sediment loads in different areas. is useful to understand if the results derived for 2018-2019 can be generalized for a longer time period. The team also analyzed the mineral composition of suspended sediment at selected stations (Kafle and Bhandari 2019) which In terms of total contribution, the biggest fraction of can give some hints about sediment generating processes sediment originates from the Upper Kali Gandaki, upstream and their impacts on equipment. Sediment composition was of Jomsom (7.9 Mt/yr), followed by the Middle Kali Gandaki analyzed with regard to four minerals: Quartz, Feldspar, (6.8 Mt/yr) and the Mustang Plateau (6.6 Mt/yr; Figure Muscovite, and Tourmaline. Amongst these minerals, Quartz 2.7). The tributaries and the lower Kali Gandaki watershed (Mohs hardness 7) and Tourmaline (Mohs Hardness 7 – 7.5) each contribute only around 2 – 3 Mt/yr. It should be are hardest and Muscovite (Mohs hardness 2 – 3) is softest. noted that the yield (sediment generation per drainage area) All except Muscovite are harder than chrome-nickel steel from the Upper Kaligandaki and the Middle Kaligandaki commonly used for turbine parts (Mohs hardness around 4) is extremely high (around 10,000 t/km2/yr). The yields for (Felix et al. 2016). The other, harder minerals will be highly the remaining sub-watersheds (i.e., Myagdi Khola and Modi abrasive on the softer turbine material. 23 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  Figure - 2.6: Topography of the Kali Gandaki watershed, location of gauging stations and their respective drainage area. Note that no sampling was taken at Kali Gandaki Reservoir during this period, rather values were interpolated from upstream observations Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  24 Table - 2.2: Gauging stations used in this study and overview of sub-watershed characteristics Gauging station Sub-watershed Drainage Elevation: Elevation: Glacier- Mean annual area Mean Standard covered Area precipitation [km2]+ [m]+ deviation [% of total]* [mm]# [m]+ Jomsom Mustang 3165 4786 845 7.9 426.0 Tatopani Upper Kali Gandaki 782 3946 1251 9.8 1191.1 Manghalghat Myagdi Khola 1095 3357 1509 12.6 2260.3 Modi Beni Middle Kali Gandaki 840 2405 1092 1.5 2076.1 Nayapul Modi Khola 648 3192 1645 11.8 2988.0 Kaligandaki Lower Kali Gandaki 1034 1245 365 0.0 2394.7 Reservoir (not covered by current sampling campaign) + Derived from the DEM (30 m resolution) * ICMOD glacier dataset # Interpolated from DHM rain gauges using Kriging Box 2.1: Methods used for converting observed sediment data to longer-term sediment load Methods & Tools 2.1: Rating curves for long term sediment budgets. Sediment rating curves were fitted to the observations at each sampling location to convert sediment observations, which cover only a single year, to a baseline annual sediment load that is representative for a longer time period. These rating curves of the form CS = a * Qb where CS is the sediment concentration in gram/m3, Q is the discharge in m3/s and a and b are location-specific scaling parameters. Rating curves relate water discharge to sediment concentration, which is useful in areas like the Kali Gandaki watershed, where a much longer-term record is available for discharge than for sediment. Rating curves can then be used to reconstruct past, unmonitored sediment concentrations and loads from discharge observations, as long as there is sufficient confidence that the processes linking the generation and transport of sediment have not changed significantly over the time horizon on which the rating curve method is applied. Using this approach, the total load reaching KGA can be attributed to the various sub-watersheds. Figure 3.2 reports the total annual sediment load estimated at each station, as well as how much the corresponding sub-watershed adds to the total sediment load, over a roughly 5-year time period for which flow data were available. For more details on input data and limitations of this approach, see Appendix 1. This analysis helps to understand possible differences and This is because glaciers can scour hard bed-rock creating fine similarities in sediment generating processes between sub- sediment that is difficult to remove in desanders and is very watersheds. Sediment from different sources (in terms abrasive. In Kali Gandaki, parts of the watershed are glaciated of location and process) can have different impacts on and glacial erosions cannot be controlled by watershed hydropower. For example, sediment derived from glacial management. A large sediment contribution from glaciers erosion is often particularly damaging to hydropower plants. would hence limit opportunities for such interventions. 25 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  Figure - 2.7: Total load and contribution of sub-watersheds draining to KGA. The total load describes the total amount of fine sediment transported at the outlet of each sub-watershed. The added load indicates how much of that load originates within a sub-watershed. Loads are calculated using a rating curve approach using results from the 2018 – 2019 sediment monitoring campaign by Kathmandu University and past discharge data from 2009 – 2015 (See Box 3.1). Error bars indicate the standard deviation of total and added load over this time period. Note that there were no measurements for the Lower Kali Gandaki, and therefore data are derived from various data sources 4.00E+07 3.50E+07 3.00E+07 Load [t/yr] 2.50E+07 2.00E+07 1.50E+07 1.00E+07 5.00E+06 0.00E+00 Mustang Upper Myagdi River Modi River Middle Lower Kaligandaki Kaligandaki Kaligandaki Total load [t/yr] Added load [t/yr] However, the analysis concluded that even sediment from is not significantly different from the others. This also little-glaciated sub-watersheds consists mostly of very hard implies that managing sediment from the non-glaciated minerals (only around 10% of the sediment consists of areas, and mitigating processes such as landslides, might be Muscovite). The mineral composition of sediment from an effective strategy for reducing the load of hard sediment Myagdi and Modi (the sub-watersheds with most glaciers) to KGA. Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  26 3. METHODS AND TOOLS © David Cutler 3.1. OVERVIEW OF ASSESSMENT Through a series of consultative workshops with stakeholders from six departments representing four government APPROACH ministries, the cross-cutting Water and Energy Commission The study presents a systematic approach to assess where, Secretariat, as well as NGOs, consultants, and researchers, in what quantity, and through what processes sediment is we identified the following ecosystem service benefits of generated in the Kali Gandaki watershed, identify plausible watershed management of importance in the Kali Gandaki interventions through investing in green infrastructure watershed: approaches for watershed management, and evaluate their impacts. Activities and locations based on the potential for Downstream benefits, arising from reduced sediment achieving multiple ecosystem service objectives are then arriving at KGA: prioritized, and finally the economic benefits of a targeted • Reductions in damage to equipment, efficiency loss, and program of watershed investments are evaluated, to develop need for repairs a cost-effective watershed management investment portfolio • Reduced costs of desanding and preventative measures (see Figure 3.1 for general workflow). • Maintenance of storage capacity for peaking. 27 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  Figure - 3.1: Workflow used in this study to evaluate watershed management activities, value their impacts, prioritize intervention locations and estimate the benefit: cost at different levels of investment Establish Plausible Evaluate Assess Prioritize baseline ranges of impacts of marginal & develop sediment intervention activities value optimal budget impacts portfolios Physical models: Literature-based Physical models: Economic analysis: Effectiveness • Erosion estimates • Erosion & • Energy • Landslides sedimentation • Landslide risk Benefit/cost of • Roads Stakeholder • Landslides • On-site benefits investment • Glaciers consultation • Carbon • Carbon Local benefits, arising from the reduction in landslide risk: transport, along with novel approaches to estimate the • Avoided lives lost contribution of roads, landslides, and glacial erosion to • Avoided cost of replacing structures total sediment loads. • Avoided cost of road repairs. • Identify plausible interventions and range of impacts: A combination of literature review and Global benefits: stakeholder consultation was used to select activities that • Carbon sequestration from improving or preserving agencies are currently engaged in and provide estimates vegetation cover and enhancing soil carbon of their effectiveness. • Evaluate potential impacts of activities: The Other benefits that were mentioned in our stakeholder impacts of activities on ecosystem services of interest consultations, but not quantified in this study, include water were evaluated using the biophysical models mentioned quality and water flow in streams for drinking water and above, so as to determine the location-specific benefits of irrigation, improved water infiltration and regulation for activities in every possible location. local springs, water flows for downstream fisheries, and • Assess marginal value of sediment reductions: A biodiversity. We discuss the potential benefits for water combination of micro-economic modeling, spatial overlays, regulation and water quality later in this report, but due and qualitative methods was employed to estimate the to data limitations we were not able to quantify nor value value of implementing watershed management practices these in this study. Further, the study focuses on the value that reduce sediment and landslide risk, provide local of reducing sediment reaching KGA and does not attempt benefits to landholders, and store carbon. to quantify these values for the Modi Khola (14.8 MW) • Prioritize intervention scenarios: An optimization and Lower Modi 1 (10 MW) hydropower facilities, located tool (ROOT), which uses estimates of implementation upstream of KGA on the Modi tributary. costs and modeled effectiveness of each activity/location was applied to identify optimal portfolios of interventions Existing watershed-scale models for erosion and at different budget levels. sedimentation were combined with newly-developed modeling and economic approaches to address each stage of Stakeholder input was solicited at each of these stages, to analysis, as follows: provide data on sediment concentrations in the Kali Gandaki watershed, data on physical and economic considerations, to • Establish sediment budget: A set of newly collected define feasible activities, and to vet the analytic approaches. field data on sediment concentrations was used with More details on each of these steps are provided in the an existing InVEST model for erosion and sediment following sections. Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  28 3.2. DEVELOPING A SEDIMENT erosion) are modeled in a spatially-distributed way across the landscape. In each case (excluding glacial erosion), the BUDGET estimates of local erosion are modified by a sediment delivery In brief, sediment budgets represent a framework to organize ratio to account for sediment retention on the landscape. and analyze information on sediment processes and their The total sediment from all four sources is then modified to possible response to human interventions (Reid and Dunne account for sediment deposition in stream channels between 2016). Sediment budgets are crucial tools for watershed-level the source and KGA’s reservoir. Each of these analytical steps sediment management as they identify the dominant source is described below and further elaborated in Appendices 1 processes that contribute to the sediment load at specific through 3. Further, the economic analysis (Section 3.3.5) also locations (Reid and Dunne 2016). Creating a sediment budget accounts for the fact that not all sediment that reaches KGA’s can help, for example, to understand which processes result in reservoir will end up being diverted for power generation. the observed sediment load in a specific part of a watershed. 3.2.1. Sheet and rill erosion Sediment budgets cover various erosion processes that Sheet and rill erosion (abbreviated as sheet erosion hereafter) generate sediment (sources), as well as sinks where sediment occurs when soil particles are detached and transported is deposited. Different sources generate sediment at different downslope by the force of rain impact and shallow overland rates, and with variable timing and characteristics in terms flow. Sheet erosion depends on the balance of forces of grain sizes. Locations of sediment generation are then protecting topsoils from being eroded, such as cohesion connected to sediment sinks in downstream areas via of the soil matrix or vegetation, as well as on the erosive processes of sediment transport, first on hillslopes and then forces exerted by rainfall. Eroded and transported particles in river channels (Downs et al. 2018). are typically fine (fine sand and finer) and contain organic material from the topsoil. Hence, location matters, and sediment budgets are ideally derived in a spatially distributed manner to account for the Sheet and rill erosion will occur naturally on most hillslopes impacts of processes along the transport pathway between but can be greatly magnified by loss of vegetative cover specific sediment sources and downstream sinks (Wasson 2003). and degradation of soils. In the Kali Gandaki watershed, The dominant processes generating sediment will depend sheet erosion is the dominant process on the slopes of the strongly on the geographic setting, hydro-climatic conditions, lower watershed. Here, slopes are steep and particularly and the legacy of human interference (Piégay 2016). susceptible to the erosive forces exerted by strong monsoonal rainfall. Farmers in this area have traditionally For Himalayan watersheds, relevant processes of sediment adopted farming practices that minimize soil erosion, such generation typically include glaciation, mass-movement as the construction of terraces along the slope contours. (such as landslides and rockfalls), sheet and rill erosion from Sediment yield from terraces in Nepal has been observed natural hillslopes and agricultural areas, as well as erosion in to vary by around one order of magnitude as a function river channels (Wasson 2003) either from eroding bedrock or of the adopted management practice (Chalise and alluvial sediment (Fort, Cossart, and Arnaud-Fassetta 2010). Khanal 1997). Hence, abandonment or neglect of these All of these processes are subject to anthropogenic alteration terraces because of outmigration might increase erosion ranging from local scales to global scales. In addition, purely from terraces. man-made processes such as erosion from roads or mining can contribute significantly to a watershed’s sediment budget On higher slopes, remaining natural forests likely protect (Sidle and Ziegler 2012). soils from strong erosion. However, very little vegetation exists in the Mustang area. Very low precipitation limits This section gives a brief overview of relevant processes sediment mobilization on these slopes and soil erosion is for the sediment budget of the study area, including their typically limited to areas disturbed by human activities, e.g., prevalent location, possible alteration by humans, and cattle grazing (Fort, Cossart, and Arnaud-Fassetta 2010). modeling approaches to quantify their contribution to the total sediment budget, for both the baseline and future It should also be noted that part of the sediment that reaches management scenarios. the stream network might be deposited in the river channel downstream, and therefore might not reach the KGA In brief, sediment sources (sheet and rill erosion, glacial reservoir or other downstream point of interest (see Section erosion, landslide-mobilized sediment, and road-induced 3.2.5 for a discussion of channel transport). 29 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal Box 3.1: Methods used for modeling hillslope erosion Methods & Tools 3.1: Modeling hillslope sheet and rill erosion. This study modeled sheet and rill erosion on hillslopes using the InVEST model for sediment retention (SDR; Hamel et al. 2015). This model is based on two separate components, accounting first, for soil erosion on single land parcels and second, for the subsequent sediment transport from a land parcel to the next downhill river channel. The first component is based on the Revised Universal Soil Loss Equation (RUSLE; Wischmeier and Smith 1978), according to which erosion can be calculated from E=R * K * LS * C * P where • E is erosion in t/ha/yr • R is rainfall erosivity (units: MJ.mm/(ha.hr)) • K is soil erodibility (units: ton.ha.hr/(MJ.ha.mm) • LS is a slope length-gradient factor (unitless) • C is a cover management factor (unitless) • P is a support practice factor In the above equation, erosivity and erodibility are derived from gridded global data sets. The C values for different land use types (e.g., pasture vs. forests; derived from Nepal-specific land use maps) are derived from C values tabulated in relevant literature. The P factor can be used to parametrize the effectiveness of soil conservation practices to avoid soil runoff from a parcel (P=1: no effective erosion prevention, P=0: support practices fully stop soil runoff). A combination of C and P factors are used to describe the impact of a specific land use (e.g., growing corn) under different conservation practices (e.g., growing corn on a degraded, downslope-tilled plot vs. growing corn on a terraced plot with hedgerows). RUSLE was developed for agricultural plots in the United States, making location specific calibration as well as a consideration of larger-scale topographic complexity via the SDR factor a necessity for landscape scale-applications. This is especially true in places where topographic and climatic conditions are much more extreme than in the locations where the RUSLE equation was developed, such as Kali Gandaki (Benavidez et al. 2018). Therefore, the model suite was calibrated to match average annual sediment loads derived from observed data (Box 2.1) Eroded sediment will be transported downslope but a portion is deposited along the transport pathway. In InVEST, this retention of sediment on the slopes is modeled using a conceptual factor, commonly referred to as sediment delivery ratio (SDR), which is calculated for each pixel as a function of the area upslope of a pixel and the topography of the flow path between the pixel and the nearest stream (Cavalli et al. 2013). With the SDR applied, the final sediment delivery to the streams is Q S,sheet= E * SDR Note that the SDR factor is calculated on a pixel-by-pixel basis (not a single value applied to the entire area), taking into account the landscape context of all upslope and downslope pixels. 3.2.2. Glacial erosion Glacial erosion is limited to high, glaciated parts of the watershed, in this case the Dhaulagiri/Annapurna range and Glacial erosion occurs as the glacial ice mass moves over along the rim of the Mustang plateau. Some of these glaciers the underlying bedrock and scours it. As it scours into erode bedrock of great hardness, such as the Leucogranites (a bedrock, glacial erosion can mobilize significant amounts type of granitic igneous rock) of the North-Eastern Mustang of fine and abrasive particles that are easily transported plateau, while most glaciers along the southern slopes of the in a river, are hard to remove in a desander and have the main chain are located on sedimentary rocks (sandstone and potential to cause significant damage to hydro-electric schist) which might create less abrasive sediment particles facilities downstream. (Parsons et al. 2016). Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  30 Box 3.2: Methods used for modeling glacial erosion Methods & Tools 3.2: Modeling glacial erosion. The sediment load generated by glacial erosion is a function of the run-off from the glacier, which mobilizes sediment at the glacier toe, the size of the glacier, and the properties of the underlying bed-rock, which typically results in a non-linear relationship between the sediment generation of a glacier and the produced melt water discharge (Hallet et al. 1996). There are no specific measurements available for glaciers in the study area. However, for the Gangotri glacier in western Nepal Haritashya et al. (2006) propose a power law of the form log(CSS) = 1.0862 * log(Q g) + 1.3141 where CSS Suspended sediment concentration in g/m3 Q g Discharge from a glacier in m3/s To determine discharge, we assume that the discharge can be determined from a steady state mass balance Q g = Pg - ETg Where Q g mean discharge in m3/s Pg mean precipitation over a glacier in m3/s ETg mean evapotranspiration from the glacier surface in m3/s This notably assumes that glaciers are not losing any mass, an assumption which might become more inaccurate as climate changes, resulting in faster glacier melt and hence an exponential increase in sediment generation. In this study, Pg is interpolated from available rain gauge observations (see Appendix 2: Modeling Landslides) and ETg is derived from global data sets (WorldClim Version 2). Mass movements: Landslides and 3.2.3. LSOs). Second, the mass of sediment mobilized from an rockfalls LSO. Third, the probability with which an LSO will fail, given locally prevailing rainfall conditions. Fourth, the Landslides are a significant and possibly even the dominant runout path originating from an LSO which might affect source of sediment in the Kali Gandaki watershed (Struck et assets (building and infrastructure) outside of the immediate al. 2015). Landslides occur where the forces retaining soil and landslide area. The model intersects LSOs and runout paths the fractured bedrock are exceeded by the downslope force. with known locations of assets to derive a stochastic measure In Nepal, this commonly occurs because of rain-induced changes in slope water saturation as heavy storms hit slopes for the exposure of assets to landslides. The outputs are: that are already saturated by monsoonal rainfalls (Gabet et al. (1) identification of the spatial extent of possible landslides 2004). In the Kali Gandaki watershed, there is strong evidence (LSOs) and their runout, (2) the probability of failure for that landslides are most prevalent along the southern slopes specific LSOs, (3) the amount of sediment mobilized from of the Annapurna and Dhaulagiri and in the Kali Gandaki a potential slope failure, and (4) an estimate for hazards to Gorges in the Upper Kali Gandaki sub-watershed (Struck et specific assets given implementation of different landslide al. 2015; Figure 2.6). These parts of the watershed receive mitigation activities. high amounts of precipitation, slopes are extremely steep and rapidly uplifting, and might be weakened by tectonic This approach goes far beyond common landslide activities along fault lines (Parsons et al. 2016). Landslides can vulnerability maps, in which a factor of safety is calculated damage assets such as roads, fields, or structures that are built for a single rainfall intensity pixel by pixel. While such on a failing slope, as well as when mobilized sediment travels approaches allow to qualitatively describe which parts of a downslope (referred to as “runout”). landscape are relatively more or less prone to slope failure, they do not allow to derive any of the information listed in The landslide model gives four key outputs. First, connected points (1) – (4) above, which are critical to understanding areas of a hillslope that are prone to slope failure and that the value of landslide mitigation measures as part of a might form a landslide (referred to as landslide objects, or watershed management program. The novel approach 31 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal Results of the stochastic connectivity of landslides and runout tool for an area on the middle Kali Gandaki Figure - 3.2:  River (small cutout for location). Red colors indicate landslide probability and brown colors indicate runout probability used in this study allows for estimation of how watershed Nearly all parameters in this model (which includes sediment management interventions can change the probability of delivery from landslides, runout and landslide location) can slope failure for a range of precipitation events. For a high- be derived from global datasets (DEMs, soil depth) or are level overview, see Box 3.3: Methods used for modeling landslide interpolated from location-specific observations (extreme locations and probabilities. Appendix 2 provides more details on rainfall probabilities based on observed rain gauge data). the functioning of the model, its assumptions, the required Some geotechnical parameters, such as soil cohesion and inputs, and the multiple types of spatially distributed internal angle of friction, are not commonly available at information it provides. Appendix 2, Table 4 lists specific watershed scales, and even Nepal-specific observations data sources and values for model parameters. are rare or absent, requiring to fall back on global data Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  32 (Appendix 2, Table 4) and watershed specific calibration of poorly executed cut slope ditches, or increased saturation landslide-related sediment loads (Appendix 2, Section 3.3). from road drainage pipes). All these factors impact slope stability on scales much smaller than the model resolution In the Himalayas, and specifically in Nepal, a significant employed here, therefore modeling road-induced slides is loss of life and damage to infrastructure such as roads and not part of this study. However, the appendix provides a hydropower plants is related to landslides associated with more detailed discussion of the topic (see Section 3.3 and seismic events (Kargel et al. 2016; Schwanghart, Ryan, and Figure 9 and 10 in Appendix A2). Korup 2018). Co-seismic landslides were not included in this study because no spatially distributed information on ground Figure 3.2 shows the results of the landslide model for an acceleration with different return periods was available. area in the middle Kali Gandaki sub-watershed. Locations However, for future studies that focus more on infrastructure of roads and houses are derived from the Open Street Map risk, the landslide model presented here could be readily dataset. Reddish areas show the parts of the hillslope that adopted to calculate failure probabilities for different ground are prone to failure. The different shades of red indicate accelerations with different return periods and estimate the that different connected parts of a hillslope (LSOs) will resulting hazards. fail with different probabilities. The brown cells indicate modeled runout paths that begin at the downslope end of It should also be noted that roads and other infrastructure each LSO. Runout paths are colored in different shades (e.g., irrigation canals along steep hillslopes) can trigger of brown, according to the failure probability of the LSO landslides via a plethora of mechanisms that change hillslope from which they originate. Note that many houses and hydrology and force balances (e.g., weakening of the cutslope, roads are located either on LSOs or on runout paths from overloading of the hill slope, increased infiltration from upslope LSOs. Box 3.3: Methods used for modeling landslide locations and probabilities Methods & Tools 3.3: Modeling landslide locations and probabilities If a specific part of a hillslope is prone to failure is commonly described by a factor of safety of the form ci+δci+(γs-γw*mi )*zi*αi*tan ϕi FSi= γs*zi*sinαi*cosαi The equation is based on the following input variables: • ci soil cohesion [kPa] • δci added cohesion because of plant roots or slope engineering [kPa]. Root cohesion is set to 12 kPa on forested cells (derived from land use; Table 3.6; Vanacker et al. 2003). • γs unit weight of soil [kN/m3] • γw unit weight of water [kN/m3] • m soil water saturation [ - ] • zi soil depth, assumed to be the depth of a potential failure plane [m] • αi slope angle [deg] • ϕi soil internal angle of friction [deg] This factor of safety is calculated on the scale of single cells of a digital elevation model. However, the volume of sediment mobilized by a landslide will depend on the overall size of a landslide, expressed by the relation ELS = 7.24A1.322 LS Where ELS is the mobilized sediment volume in m3 and ALS is the area of the landslide in m2 (Larsen et al., 2010). It should be noted that while this equation is empirical, there is a very good fit between the more than 4500 global observations in the original area and volume data set (R2=0.95; Larsen et al., 2010). To determine the area of a potential landslide, one first determines the maximum extent of slope failure for very wet, fully saturated conditions. Then all parcels on a slope that are topographically connected based on flow paths are grouped into multi-cell landslide objects (LSOs). The threshold 33 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal Box 3.3 (contd.): Methods used for modeling landslide locations and probabilities precipitation required to make each LSO fail (m*) and the probability p* with which m* is exceeded at the specific location of a landslide are determined. This information is derived from a spatially generalized probabilistic analysis of extreme rainfall based on available gauge data obtained from Nepal Department of Hydrology and Meteorology (DHM; See Appendix 2). The model resolution is identical to the resolution of the underlying DEM (30m for this study). Appendix 2, Table 4 lists case-study specific data sources and typical parameter values and data sources and to Appendix 2 for detailed model formulation. Finally, the sediment delivery rate to the streams from a specific landslide object can be calculated as QS,Landslides = ELSρS p*SDR which considers the hillslope connectivity between the location of the landslide and the next downslope stream and the density of mobilized sediment (ρS in [t/m3], herein assumed to be 1.6). Lastly, the empirical method proposed by Rickenmann (2005) is used to identify which downslope cells might be impacted by the runout of a specific landslide (see Appendix 2). Box 3.4: Different approaches for large-scale landslide hazard assessments Methods & Tools 3.4: Approaches for landslide modeling and susceptibility assessment Evaluating landslide hazards on watershed scales is an area of ongoing research. This is because of the complexity of underlying physical processes, the importance of small-scale heterogeneity, e.g., in soil properties, and also because relevant data is rarely available at large landscape scales. For single slopes or small watersheds, tools such as CHASM or Step- TRAMM can be used for detailed assessments of slope stability. However, such tools cannot be applied for a landscape-scale screening of landslide hazards because of their high computational demands and the above-mentioned data limitations. At the watershed scale, various geospatial assessment approaches are commonly used, typically using raster-based data to represent the study area. The area of interest is represented as a set of cells, and each cell has certain parameters assigned to it. Broadly, there are three approaches that are commonly applied: (1) qualitative susceptibility assessments, (2) quantitative susceptibility assessments, and (3) factor of safety assessments (the approach used in this study). All three methods can be applied to larger areas. Some characteristics of each approach are given in the table below. Ideally, all methods will be validated with geo- and time-referenced observations of landslides, whereas approach 2 (quantitative susceptibility) requires such data up front. 1. Qualitative susceptibility 2. Quantitative 3. Factor of safety susceptibility In a nutshell Expert based ranking of Data-driven approach to link Physically based method different driving factors possibly observed landslides to specific considering drivers of slope related to landslides. local factors. stability. Results Qualitative hazard ranking, i.e., Map of cells where landslides Map of cells where landslides maps identifying higher or lower can occur for certain conditions. can occur for certain conditions, landslide risk. or probability of landslide occurrence (this study). Pros +considers local expert + data driven method for a + allows modeling of knowledge specific geography interventions (if their impact on + include factors that cannot + include factors that cannot physical parameters is known) easily be integrated in physically easily be integrated in physically + results in a quantitative risk based approaches based approaches assessment + does not require training data + results in a quantitative risk + does not require training data assessment + transferable between locations Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  34 1. Qualitative susceptibility 2. Quantitative 3. Factor of safety susceptibility Cons - results in a ranking rather - Requires training data, e.g. - only applicable to small and than a quantitative measure of georeferenced data on timing shallow landslides landslide hazard and magnitude of landslides, - significant model uncertainty - based on local knowledge, that are not commonly - complex changes in slope not easily transferred between available. stability, e.g., because of human settings. - trained for specific locations, disturbance, cannot easily be not easily transferable integrated Herein, we propose a geospatial analysis to identify connected areas of possible landslides. The motivation for this analysis is that the hazard created by a landslide will exponentially increase with its area, hence it is of utmost importance to understand if a number of failure-prone cells will all fail together as a large slide, or rather as single small events. Our connectivity assessment also allows us to determine the runout path and possible additional infrastructure at risk along that runout path. The proposed approach for hillslope connectivity is compatible with all approaches described above; that is, results from any of the above-listed methods could undergo post-processing to evaluate the risk of connected cells failing together. 3.2.4. Road induced erosion for road access to markets and infrastructure. Implementing best practices for road construction, including a strategic Roads contribute to the sediment budget via three main planning process to avoid steepest and least stable slopes processes. First, most roads in the Kali Gandaki watershed could present an opportunity to reduce future erosion, but are unpaved, and rain erodes unpaved road surfaces and this strategic infrastructure planning process is not part of de-vegetated or unstabilized cut slopes associated with this modeling study. them. Second, sediment that is cut during the road building processes is often not hauled away but disposed of on the The model for road surface erosion is based on empirical valley side of the road (fill slope). Without proper stabilization, observations from the Virgin Islands, one of the few this sediment is prone to be remobilized and delivered to the streams, a process commonly observed in the study area. empirical models available (Ramos-Scharrón and Third, roads change the subsurface hydrology of slopes and MacDonald 2007). Applying such a location specific model increase the landslide risk. The model developed here covers to a different geography introduces significant uncertainty. the first two processes for road-induced sediment generation Building a Nepal or Kali Gandaki specific model would (see Box 3.5 and Box 3.6). Modeling the link between roads be relatively straight forward using sediment traps and and landslides in a mechanistic manner would require other low-tech equipment to measure sediment delivery detailed information on road design and geotechnical from road segments with different characteristics (slope, parameters which is not available at a watershed scale for precipitation, surface treatment, traffic rates; see Ramos- this area. Scharrón and MacDonald (2007) for a description of the required equipment). Road erosion is a function of the terrain on which a road is built. Roads with steeper gradients will erode faster and Erosion from the road cut is modeled separately. Uncertainty cutting a road in a steeper hillslope will mobilize more material is lower for this model, as most parameters can be derived than cutting on a gentle slope. Road induced erosion is likely from available topographic data. Some details on the design most prevalent in the lower parts of the watershed, where of roads in the study area are derived from photographic population centers are located on very steep slopes. There is evidence (Appendix 3). likely much less erosion from roads in the Mustang region, where there are less roads, slopes are gentler, and there is Otherwise, all data for the road model are derived from lower population and infrastructure. Road-induced erosion global data sources. Open Street Map (OSM) data provide is likely to increase in the future, as more settlements strive the location of the roads and define different categories of 35 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal roads and tracks. A typical width is assigned to each type. comprehensive for all roads in the area. Mean annual rainfall Notably, this approach introduces uncertainty regarding the is interpolated from DHM gauge data. Road gradients and actual cut volume of roads and the OSM data is by no means hillslope angles are derived from the DEM. Box 3.5: Methods used for modeling erosion from road surfaces Methods & Tools 3.5: Modeling erosion from road surfaces Erosion from road surfaces is modeled for each road segment using an empirical equation ER=(ERS+ECS )*ρS ERS+ECS=1.09*(-0.432+fg (S1.5 P))*L*W Variables are: ER Total erosion from a road segment in t/yr ERS Erosion from the road surface in t/yr ECS Erosion from the cut slope in t/yr ρS Sediment density in t/m3 fg Grading factor 4.73 for freshly graded roads, 1.88 for ungraded road. Average (3.305) was used for this study S Gradient of road segment P Precipitation in cm/yr L Length of road segment m W Width of road segment m It should be noted that this specific model was derived for roads in the tropical US Virgin Islands (Ramos-Scharrón and MacDonald, 2007; Ramos-Scharrón and MacDonald, 2005), the use of this model for Nepal is motivated by the absence of Nepal-specific data on road erosion. The precipitation in this study area is similar or higher than on the Virgin Islands for some parts of the watershed, but much lower in others, while slopes throughout the watershed are likely much higher than in the Virgin Islands. It is hence difficult to assign a specific positive or negative error to the use of such a location- specific empirical model to the Kali Gandaki watershed. The analysis assumes that part of the sediment running off a road at its lowest point is retained on the landscape before reaching the stream. This retention is modeled using the same per-pixel SDR factor as used in the sheet erosion model (Box 3.1) so that the final sediment delivery to streams is Q S,Roads = ER* SDR Box 3.6: Methods used for modeling sediment delivery from road cuts Methods & Tools 3.6: Modeling sediment delivery from road cuts The sediment mobilized from the road cut material is calculated based on the width of a road, the local gradient of the hillslope into which the road is cut and by making some assumptions on the cutslope angle (See Appendix 3). These parameters then allow us to calculate the cross-section of the road cut on a specific slope, so that the total cut material from a specific road segment is * =AC L ρS ECut Where * Mass of cut material in t ECut AC cross sectional area of the road cut in m2 L length of the road segment in m ρS sediment density in t/m3 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  36 Box 3.6 (contd.): Methods for modeling sediment delivery from road cuts Note that this equation yields the instantaneous sediment in tons mobilized when the road is cut (or when it needs to be cleared after a landslide). To calculate a rate of annual sediment delivery, the analysis assumed that sediment from the road cut is mobilized over a 25-year time horizon, so that the annual delivery is E*Cut ECut 25 Sediment delivery from the cut material to the stream is calculated using the pixel-specific SDR factor (Box 3.1) so that Q S,Cut = ECut*SDR In this study, changes induced in the factor of safety due to a road intersecting the LSO are not included, but road-induced landslides are an area of great interest and could be pursued in future work (see Appendix 3 for more discussion on this topic). Box 3.7: Road construction practices and standards in Nepal The Government of Nepal targets call for the country’s road network to nearly quadruple, from about 65,000 km (0.44 km/km2) to 220,000 km (1.5 km/km2) by 2030 (NPC 2015). While there are benefits to be realized from better linking rural areas of the country, Nepal’s steep slopes, fragile geology, and climatic conditions pose challenges for building roads that will be durable and minimize impacts on the environment. The planned increase in road construction, especially of earthen roads in rural areas, is expected to increase sediment generation substantially. This would impact water sources, agriculture, vegetation, hydropower operations, and other ecosystem services. As a secondary effect, roads might open access to areas that so far have seen little disturbance from activities such as logging. Such negative impacts may be especially severe if current practices are not reformed. Rural roads are often built using heavy equipment, but with otherwise low budgets, little technical expertise, or best-practice design principles to reduce environmental impacts. Construction of local roads is often politically driven, and may reflect the interests of elite groups, while other actors are often the ones to experience the negative impacts from landslides, vegetation loss, and sediment generation from new roads. Extreme erosion from roads can be reduced during different steps of construction. First, strategic planning of road networks can help identify road networks that minimize the road length required to connect a maximum number of villages. Second, best practices can be adhered for the engineering design of roads and slopes. Last, proper environmental safeguards are required to ensure proper disposal of cut material. There are a number of existing standards for road construction in Nepal. Nepal Road Standards, most recently revised in 2013-2014, provide design parameters for the design of strategic and local road networks based on administrative classification, technical/functional classification, side slopes of embankment, gradients, traffic characteristics, terrain, sight distance and slopes, and other factors. More than a dozen standards and frameworks have been formulated and implemented in road construction and management, such as: Rural Road Standards 1998, Standard Specifications for Road and Bridge Works 2014, Bridge Standards for Strategic Road Network 2009, Environmental and Social Management Frameworks for road management practices in Nepal. Furthermore, bio-engineering techniques for slope and soil protection (e.g., planting local deep-rooted species on bare roadside embankments to reduce soil erosion and stabilize slopes; Howell 1999) are well-known by government officials and are sometimes practiced, but not systematically implemented, for new roads. These standards also state that all roads should be designed and constructed with proper assessment of environmental and social aspects and their impacts as per the umbrella Environmental Protection Act (EPA 1997, currently being amended). However, poor governance means that these requirements are not consistently observed, especially in more rural areas, where funding and technical capacity for improved road building and planning are not available. 37 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal 3.2.5. Sediment transport in river and fluvial supplied to the network. If the transport capacity is lower erosion than what is supplied, deposition of sediment will occur. The model was calibrated by varying the supply from the different Fluvial processes impact sediment budgets in two main sub-watersheds until model results were in agreement with ways. First, sediment eroded from beds and banks can be the observed sediment data. a relevant source of sediment in mountain rivers (Wasson 2003). Second, transport processes in rivers are responsible For the Kali Gandaki watershed, the analysis shows that most for transporting sediment from the location where it enters of the rivers allow for a great amount of sediment transport the river network to a downstream location. A significant (Figure 3.3, left panel). For fine sand, most rivers have a much part of the sediment entering a stream can be deposited on greater capacity to transport sediment than what is supplied floodplains or on sediment bars (Fryirs et al. 2007). Having from upstream (Figure 3.3, left panel). Especially rivers in the such areas of deposition will change the impact hillslope upper and middle Kali Gandaki and tributaries along the management measures will have on sediment transport at a southern slopes of the Dhaulagiri / Annapurna range have a downstream location. very high transport capacity (several thousand Megatons of sediment per year, Figure 3.3, red colors in left panel). The For this study, only fluvial sediment connectivity is evaluated. only rivers that are possibly transport limited (i.e., receive This is because modeling bank and bed erosion in rivers more sediment then what they can transport) are rivers in would require more detailed information on bed-material the Mustang area. This could be because the hillslopes in composition. Also, bank and bed erosion cannot be mitigated the area are not stabilized by vegetation, so that even the with common watershed management techniques small amounts of precipitation in the Mustang area can lead to significant sediment supply. However, because of the low To estimate which amount of sediment can be delivered precipitation, the discharge in the rivers is very low and only from each part of the watershed to KGA, the study used a limited amount of sediment can be conveyed downstream. the CASCADE model to quantify sediment transport capacity in the river network (Schmitt et al. 2018; Schmitt, Results of this analysis show that around 95% of sand and Bizzi, and Castelletti 2016). The CASCADE model is finer material entering rivers will be delivered to KGA, a based on a statistical application of a common sediment finding which is in line with field observations that rivers in transport formula (Wilcock and Crowe 2003) on a whole this watershed are mostly supply limited (i.e., can transport network-scale.2 all sediment entering the channels; Morris 2014). This also implies that sediment reductions due to watershed While the model allows for considering many grain sizes, the management activities, even in the upper watershed, could be model in the analysis is set up to consider only for a single felt far downstream, reducing sediment delivery to KGA. The grain size of medium sand (0.5 mm). This assumption is used sediment transport in the river network as a function of supply as most processes typically targeted by hillslope management and transport capacity is shown in Figure 3.3 (right panel). (i.e., sheet and rill erosion) will deliver relatively fine sediment. Landslides instead would mobilize both coarse and finer sediment fractions. However, for landslides that 3.3. MODELING BENEFITS OF do not directly run-out into the rivers, fine sediment will be WATERSHED MANAGEMENT washed out preferentially from the landslide debris on the hillslopes. However, mechanisms and magnitudes of coarse This section describes the watershed management activities sediment delivery to streams from landslides would merit that are modeled in this study, along with how estimates of further examination and is left for future work. their costs are derived. Next, the methods used to evaluate the impacts of these activities on sediments generated The CASCADE model yields two main types of information. from hillslope erosion, on landslide-related sediment and The first is the transport capacity of the river network, associated risks are detailed, followed by the methods used to which indicates how much sediment of a certain grain size estimate impacts on carbon storage. The valuation methods could be transported in the rivers. The second information applied to estimate the economic values of each benefit is actual sediment transport. Actual sediment transport at stream are summarized (Appendices 2 through 4 provide any location will depend on how much sediment is actually more detailed treatment). The functioning of the model is shown for the Red River Basin in Vietnam in this video https://www.youtube.com/watch?v=S_EYxK4tRlc 2. Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  38  Figure - 3.3: Transport capacity for medium sand (0.5 mm diameter) (left panel) and actual sediment transport for the same grain size (right panel). Note the difference in scales between the two figures 3.3.1. Activities & costs Electricity Authority, Kathmandu University, Department of Roads, Department of Survey, Ministry of Agriculture Several categories of watershed management activities Development, Water and Energy Commission Secretariat, in this analysis are used. The categories were selected as a Basin Management Center Gandaki, Provincial Forest representative sample of management actions that could be Directorates, the World Bank, Paani Program of USAID, taken to improve landscape condition and control sediment. District Soil Conservation Offices, and World Wildlife Fund, Potential activities to be analyzed were chosen for their 7 intervention types (4 on croplands and degraded lands, feasibility and suitability to the local conditions, based on plus 3 on landslide-prone areas) were selected and are shown a review of relevant literature (Dahal and Bajracharya in Table 3.1. 2013; Paudel et al. 2017; S. B. Shrestha 2016; Atreya et al. 2008; ICIMOD 2007) and guidance documents provided The selected categories of activities are applicable to by the Nepal Department of Forests and Soil Conservation different land areas based on their current land management (DoFSC), and refined through stakeholder consultation and physical characteristics. For example, it is assumed during two workshops held in October 2017 and January that cultivated land above 5% slope could be treated 2019 in Kathmandu. Based on the input received during with one or more of the techniques in the category Hill these workshops from representatives of DoFSC, Nepal terrace improvement.3 Depending on local conditions, this Our modeling does not assume that all of the activities listed under each intervention are implemented; rather we model the average effectiveness of the 3. types of activities (based on literature-based estimates of their impacts on model parameters). We assume that upon implementation, the best activity or combination of activities from the list would be selected by local experts to maximize the potential reduction in sediment. 39 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal could include techniques such as terrace improvements, agricultural land is also considered, assuming that there will hedgerows, and/or agroforestry (planting fruit or other always be some sort of access, even if it does not show up in trees among crops). The impact and effectiveness of the the road data-set. It should also be noted that interventions modeled activities was assumed to reflect an average change might reduce the risk of smaller co-seismic landslides. This that such interventions would cause in the landscape. For effect was not analyzed in this study but should be evaluated example, while there are many different types of terracing in future work. possible (e.g. terracing with hedgerows, bench terraces, cut-and-fill), the actual impacts of activities and their Furthermore, we do not model any interventions that implementation cost will vary depending on site-specific affect road erosion, as the engineering solutions required to conditions upon implementation. Therefore, the modeling manage sediment from roads were outside the scope of what assumes that the specific activities selected to implement are normally considered “green infrastructure” watershed “terrace improvement” in a given location would reflect best interventions. We also assume that glacial erosion will not practices based on the local site conditions. be the target of any land management interventions, as most glacial areas are extremely remote with limited potential for For interventions aimed at preventing or mitigating the risk improving vegetation cover. of landslides, this study uses data reported by Dahal and Dahal (2017). From the methods reported therein, we focus Implementation Costs on two types: tree and bamboo plantation, and installation of Costs for each activity were based on a review of literature subsoil drains. Such low-cost engineering measures are not on implementation of similar activities in the Himalayas suitable, however, to address very large landslides with deep- seated failure planes. Landslide-prone areas are, therefore, (Nepal, India, and Bhutan). Studies were chosen that contain classified into four groups with increasing magnitude, and a a detailed and well-documented explanation of costs. Costs prototype portfolio of measures are developed that can be given in the studies were broken out into establishment applied for the mitigation of types 1 through 3. Specifically, and maintenance costs for materials and labor, and many the following classification is proposed: included the fraction of both initial establishment and maintenance costs borne by the landholders adopting the 1. Shallow landslide (<1.5m) in the topsoil (i.e., landslide practice. Labor costs were calculated using the reported depth < soil depth). The minimum depth of an LSO is labor inputs and a common daily wage rate (US $3 per day;)5 given by the cell size and is around 1.4 m. The 1.5 m and all costs were adjusted to a common year (2018). Specific threshold implies that the failure plane is in the range of activities were grouped into the categories given in Table 3.1 deeper rooting plants and trees. above, and the average cost was calculated (Table 3.2). For 2. Landslide depth > 1.5 m but still in the topsoil. Failure the costs of landslide mitigation, the reported cases where plane in the range of deep rooting trees. similar interventions were implemented with the stated 3. Landslide depth > depth of the topsoil, but less than 3 purpose of reducing or mitigating landslide impacts were m. Failure plane in the bedrock (i.e., can’t be reached by used as the basis for the costs. We assume that those studies roots quickly) but still possibly in the range for soft / grey- that reported a combination of grey and green engineering green engineering. approaches – and correspondingly higher costs – were most 4. Landslide depth > 3 m. Deep seated landslides which applicable to landslide class 3. would require massive engineering for mitigation.4 As with the impact of interventions, the cost of Landslide interventions are assumed to be only feasible on implementation will likely vary based on specific site hillslopes not more than 1 km away from a road, as they conditions. But assuming that, in some areas, costs are might require transport of large equipment and material. All understated and in others they are overstated, the aggregate 4. Such landslides are not considered a suitable target for nature-based mitigation measures, but modeling their location is nonetheless useful for hazard mapping and disaster awareness. These results are an important by-product of the analysis reported here. 5. This wage rate was estimated as a representative value from roughly a dozen studies used in the generation of cost figures for Table 3.2. All studies save (Das and Bauer 2012) are summarized in the WOCAT (World Overview of Conservation Approaches and Technologies) database of sustainable land management practices ( https://qcat.wocat.net/en/wocat/ ). Wage rates varied considerably over both the time (studies reported data from years between 2003 and 2014, a period over which both per capita income adjusted for inflation and consumer prices roughly doubled) and location of the studies. As labor expenses constituted a large share of costs, it should be appreciated that such cost figures are imprecise; see “Limitations” below. Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  40 Table - 3.1: Interventions modeled in this study, examples of practices normally included in such programs, and rules for where on the landscape each activity was modeled Intervention Types of practices included Implementation guiding Sediment modeled principles process affected Hill terrace Slope correction on existing Croplands > 5% slope. We assume Sheet improvement1 terracing, planting nitrogen-fixing that there are some existing terraces erosion hedgerow species along the terrace that are in moderately poor condition. margins in single or multiple rows, agroforestry Soil & water Hedgerows, hedgerow inter-cropping, Croplands <= 5% slope. We Sheet conservation practices crop residues, mulches, cover crops, 1 assume there are some existing soil erosion no tillage, reduced tillage, minimum conservation practices in place but not tillage, windbreaks/shelterbelts, currently very effective. buffer strips/greenbelts, conservation trenching, agroforestry Landslide mitigation Revegetating denuded slopes and/or Areas with high risk of landslide Landslide- (class I) bioengineering for slope stabilization failure at a depth of <1.5m and in the mobilized topsoil only. sediment Landslide mitigation Revegetating denuded slopes, Areas with high risk of landslide Landslide- (class II) bioengineering for slope stabilization, failure at a depth >1.5m, but deeper mobilized slope correction and/or excavation than topsoil and with failure plane in sediment of sub-soil drains the range of deep rooting trees. Landslide mitigation Bioengineering for slope stabilization, Areas with high risk of landslide Landslide- (class III) revegetating denuded slopes, sub-soil failure in the bedrock (i.e. below mobilized drainage and/or retaining walls rooting depth), but with a failure plane sediment <3m deep. Reclamation/ Planting fuel and fodder tree species, Degraded forest lands (defined using Sheet rehabilitation of conservation trenching, eyebrow data from (Hansen et al. 2013). erosion degraded land (forest2) pits, revegetation, hedgerow planting across the slope to regenerate degraded areas Reclamation/ Greenbelts, buffer strips, rotational Grasslands Sheet rehabilitation of grazing, fodder planting, silvopasture erosion degraded land improvement (grasslands2) 1 The DoFSC also invests in civil structures (such as check dams, gully dams in combination with other interventions) to prevent erosion, regulate velocities and trap sediment along a streamlets and tributaries. These structures are often included along with other types of landscape treatments reported in the studies from which we draw model parameters and costs (such as those derived from the observed change in sediment loads post-treatment). Therefore, they are implicitly a part of these activities in that the modeled impacts would reflect their contribution to the change in sediment; however, engineering-only solutions such as check dams in isolation of other vegetative treatments are not considered in the modeling approach of this study. 2 Note that the DoFSC does not distinguish in their reclamation of degraded land between these different types, but we separate them because modeling the impacts of these activities will differ depending on the land cover class. 41 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal Table - 3.2: Cost summary for interventions modeled Intervention modeled Cost (USD/ha) Average Min. Max. Soil & water conservation $ 1,100 $ 140 $ 2,200 Hill terrace improvement $ 2,230 $ 50 $ 8,750 Degraded forest rehabilitation $ 1,690 $ 1,080 $ 2,310 Grazing land rehabilitation $ 880 $ 730 $ 1,030 Landslide mitigation I1 $ 3,850 $ 1,260 $ 8,030 Landslide mitigation I2 $ 3,850 $ 1,260 $ 8,030 Landslide mitigation I3 $ 39,480 $ 19,450 $ 59,520 cost of a portfolio of activities will be estimated reasonably in erosion and/or sediment export from implementation accurately as the sum of an average cost per hectare. Many of similar types of activities were collected (see Appendix of the costs indicated here may seem relatively high, but it 5 for details). The management practice factor (USLE P) should be noted that we report gross costs, while studies of was given a value in the mid-range of literature-reported similar interventions often discount the labor or other costs values to represent the baseline condition (meaning that borne by landholders (thereby reporting only net costs). Our the current practices employed are operating at average analysis shows that this cost-share is on average 84% of the efficiency), and a value at the low end of reported values gross costs – explained in detail in Section 3.3.6 – and when were assigned to the intervention scenario (indicating this is taken into account, the net figures are more in line that improved practices would be operating at maximum with costs reported for similar World Bank projects. efficiency). 3.3.2. Impacts on hillslope erosion The SDR model results are used to evaluate the total sediment loading to streams under the baseline and the To represent the impact of each type of watershed intervention scenarios. The output of the SDR model is the management activity, parameters in the sediment delivery loading to streams, which was further scaled by the fluvial ratio model (SDR) are altered to reflect changes in the transport ratio (95%; as explained in Section 3.2.5) to arrive biophysical condition of the landscape caused by the at the amount of sediment avoided going into KGA. intervention. The interventions “hill terrace improvement”, “soil and 3.3.3. Impacts on landslide-related risks water conservation”, “rehabilitating degraded forest”, and 3.3.3.1. Physical impacts “rehabilitation of grazing lands” were assumed to impact It is challenging to quantify the impact of the selected hillslope erosion. For these practices, model parameters landslide treatment strategies on the parameters of the relating to the vegetation cover (USLE C) and management landslide risk model, and therefore the values presented in practice factor (USLE P) were changed. this analysis are a first, expert-based attempt at the parameter estimation. Model parameters for all treated landslides are In the baseline sheet erosion (SDR) model, we assume that changed according to Table 3.3 below (e.g., for all landslides all croplands have some form of soil management in place, of type 1, we apply the appropriate mitigation measure), and that these are operating at average effectiveness. Field which then results in an increase in the factor of safety and studies from the Himalayan region that reported a change a reduction in failure probability. Specifically, the reduction Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  42 Table - 3.3: Landslide classes, mitigation approach assumed and parameters impacted Landslide class Mitigation approach Impacted parameters 1: Shallow top soil Plantation of grass and coir netting on the • Soil cohesion: Increase soil cohesion by 15 KPa entire landslide surface. (Vanacker et al. 2003) Reforestation. 2: Deep top soil Reforestation. • Soil cohesion: Increase by 10 KPa (Vanacker et Slope correction and/or excavation of al. 2003) sub-soil drains. • Saturation: decrease m by 20%6 3: Shallow bed rock Excavation of deep drains • Saturation: decrease m by 20% in slope failure probability is either because of increased 3.3.3.2. Economic impacts soil cohesion or increased drainage resulting in reduced soil Between 1971 and 2013 – a period that did not include the saturation. See Appendix 2 for more details. 2015 earthquake and the landslides it triggered – landslides in Nepal destroyed nearly 19,000 homes, damaged 132 This change in probability of slope failure implies a schools and eight hospitals, destroyed 20,000 hectares of reduction in the sediment load to rivers as well as the crops, covered nearly 400 kilometers of roads, and killed risk to lives, structures and property associated with almost 5,000 people (UNISDR 2015). those landslides. Therefore, the resulting change in LSO failure probability is evaluated by comparing the pre- and It has not proved feasible to place an economic value on post-intervention model results to obtain the change in reduced landslide risks to all these end points. In this study, mobilized sediment. Avoided sediment produced as a result values associated with three of the most important benefits of landslide mitigation is reduced by 5% to account for are estimated: lives lost, structures destroyed, and roads deposition in the river channels and the resulting change damaged.7 A slightly different approach is taken to the in sediment is valued as a benefit for KGA (see Section valuation of each endpoint, and each of the three is treated 3.3.5). The 5% reduction is to account for the effect that in turn below. In each case we calculate the expected net not all sediment that has been mobilized from a landslide present value in perpetuity, that is, the expected value of the and reached the river channels will be transported through loss/damage discounted over all future periods, and assumed the river network (here we assume that a 5% fraction is to be terminated when a slide occurs, on the assumption deposited in the river channels based on our analysis of there is no further risk at the same site. sediment transport in Section 3.2.5). Lives saved Finally, the modeled change in probability of LSO failure A reduction in the probability of landslides occurring as a result of mitigation measures is applied to the values of translates into an expectation of lives saved. Economists lives and assets at risk, as described in the following section. bring reductions in expected mortality risks into cost-benefit 6. The reduction in soil saturation will depend on many local factors, such as soil type, slope, quality of the drainage works, etc., so we assume the 0.2 value. However, the effectiveness of drainage for landslide prevention and its modeling on watershed scales would merit more detailed studies and is left for future work. (Ortigao and Sayao 2004,p. 178) report changes in the factor of safety as a response to drain installation of horizontal drains. They report an increase in the factor of safety in the range of 10 to 40% as a result of reduced soil moisture in the same range. The assumed value of 20% is, hence, rather conservative. See also (Hutchinson 1977) for a detailed discussion of drain designs for slope stabilization. 7 Damage to crops is not estimated, as the detailed data required were not available. Agricultural damages might be separated into components. First, there is the loss of the current year’s crop. Second, there may be costs of restoring land on which a slide has occurred to bring it back into productive use. Such costs should also include loss of infrastructure, e.g., for irrigation, or destruction of terraces. Third, there may be a permanent loss of some land that could be used for growing crops. The value of the first component would depend on the stage of the growing season at which landslide damage occurred. Since a crop would only be destroyed in the field if it had not yet been harvested, at least the costs of harvesting would need to be subtracted from the gross value of the crop in calculating damages; if the crop were destroyed still earlier, costs of cultivation would also need to be subtracted. Quite extensive damage might result in the permanent loss of entire fields (Thapa and Paudel 2002), in which case land value would be an appropriate measure of the value at risk. Regrettably, data on agricultural land prices in Nepal are sparse and so location-specific as to make transferring them highly speculative. 43 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal analysis by estimating the “value of a statistical life” (VSL). Structures The term is something of a misnomer; it might be better The analysis assumes that a structure hit by a landslide is stated as the value individuals assign to a small change in the damaged beyond repair, and the site on which it was located is mortality risk they face (Cameron 2010). The VSL has been lost to further use. The value of an asset is determined by the estimated by various researchers using a number of methods, yearly return it provides – in the case of a structure, its rental including compensating wage differentials required to accept value – divided by the required rate of return. The rate of riskier jobs, “stated preference” surveys that ask respondents return an owner requires to hold an asset at risk of destruction directly what they would pay to reduce the risk of death (or will sum the compensation required for waiting a year – the what payment they would accept to tolerate a higher risk discount rate – and the probability of the asset’s destruction of death), and studies of housing price differentials between during the year. This calculation is detailed in Box 3.8. more and less risky areas. As noted above, the average structure at risk from As the VSL may be interpreted as the “price of risk,” valuing destruction from a landslide occupies an area of about 45 a reduction in risk is a relatively straightforward exercise once square meters. The Nepal Central Bureau of Statistics’ the more difficult tasks of determining how an intervention Annual Household Survey (CBS 2018) asks respondents to will reduce risk and assigning a VSL have been completed. report either their actual rental payments or estimate the It is then only a matter of multiplying the change in risk by rental value of the home they occupy. In rural areas, the the VSL. If the risk is calculated as a probability of dying in household expenditures on rent average 27,180 NPR (US a landslide in any given year, the value of lives saved from $243). It is assumed that most of the houses at risk from such a risk reduction should be discounted to derive the net landslides in our study area would be in rural areas, so this present value of the risk reduction over the entire period it figure is taken as a representative rent for a structure at is realized.8 risk in the mostly rural Kali Gandaki watershed. It is not possible to distinguish types of structures from the data, We apply the value of a statistical life in Nepal that was except for the footprint of structures at risk, so the rental estimated by a recent World Bank study relating to the costs value of homes is applied to all. To account for homes of air pollution, arriving at a figure of US $34,565 (Sall, of different sizes, average rent is divided by the average private communication; see also (Narain and Sall 2016) for footprint of structures at risk in the data, 45 square meters, a review of VSL estimation procedures). This estimate is to arrive at an estimated rental value of 604 NPR (US $ conservative compared to some other research, (M. Shrestha 5.39) per square meter. It should be noted that this method 2016), and so the resulting values could be even higher under will underestimate the value for structures with multiple alternative assumptions. stories and hence a larger area than what can be inferred from the footprint. While this is a very rough figure, it It is also necessary to have an estimate of the number corresponds to a value of about 225,000 NPR (about US of lives at risk. There are no surveys that document the $ 2,000) for a house at a two percent annual landslide risk. number of people living in areas susceptible to landslides. This is broadly consistent with Nepal’s Post Disaster Recovery This study, however, maps the locations of structures at risk Framework (Government of Nepal 2016) following the 2015 (see, e. g., Figure 3.2). There is also historical data recording earthquake, which offered stipends of 200,000 NPR (US both the numbers of structures destroyed and lives lost from $ 1,786) “per eligible homeowner to assist with housing landslides in Nepal over more than forty years (UNISDR reconstruction or pay for construction of a small core 2015). The ratio of lives lost to structures destroyed in house.” that data is very nearly one-to-four. The average area of a structure in the areas identified at risk from landslides is Roads about 45 meters; hence it is supposed that one life will be Roads in Nepal are often cleared and opened for use again at risk for every 180 (= 4 ∙ 45) square meters of structures after landslides occur. One study found that maintenance at risk. crews remove, on average, between 400 and 700 cubic 8. More complicated formulations might be proposed if the value a person assigned to her life varied with the probability with which she expected to be killed in a future landslide. Such complications are not considered here, as they would likely have little effect in practice. 9. Very large landslides might fully damage a road beyond repair, requiring either its abandonment or reconstruction elsewhere. However, such large landslides likely cannot be mitigated with the watershed management techniques proposed in this report, and their recovery is an engineering problem outside the scope of this study. Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  44 Box 3.8: Methods used for calculating the value of a structure at risk from destruction in a landslide Methods & Tools 3.8: The value of a structure at risk from destruction in a landslide The value of a structure at risk from destruction by a landslide is the earnings that could be realized from owning it this year – its rental value – plus its expected present value next year. Expected present value is calculated by discounting the future value of an asset and multiplying it by the probability that it survives until the next year. Suppose Vs is the value of a structure, R the rent it would command, collected at the end of the year, d the discount rate, and P the risk of its destruction by a landslide. Suppose all these parameters remain constant as long as the structure survives. Then the value of a structure at risk, designated VS, will be R 1-P VS = + VS 1+d 1+d or R VS = P+d The risk of destruction is like an increase in the discount rate, in terms of its effect on the value of the structure. meters of landfall detritus annually from mountain roads The costs of repairing a segment of road that has been (UNEP 2012). It may be more appropriate, then, to model damaged by a landslide depends on a number of factors, the economic value of a reduced risk of landslides as a including the topography of the area affected, the extent reduction in expected costs of repair, rather than as a loss of the damage, and the quality to which the segment is to of asset value, per se.9 The expected cost of road repair in be restored. The cost of new road construction may be a every year is the probability of a landslide occurring times reasonable proxy for the costs of repair, although they may the cost of repairing the section of road damaged. The vary considerably between different types of roads and expected net present value of repair costs can be derived different places. Expenses for several relatively high quality by dividing this expected cost of road repair by the sum roads in Nepal have been estimated at between 600 thousand of the discount rate and the probability of a landslide (see and eight million Nepali Rupees (US $5,357 – 71,430) per Box 3.9). kilometer of road completed (Starkey, Tumbahangfe, and Box 3.9: Methods used for calculating the value of a structure at risk from destruction in a landslide Methods & Tools 3.9: Expected costs of road repairs from landslide Suppose the cost of repairing a segment of road damaged by a landslide is C and the probability of such a landslide occurring is P. The discount rate is δ. Suppose also that, once a landslide has occurred on a segment of road, there is no further risk to that segment. The expected present value of landslide repair costs is, then, the probability that a landslide occurs this year, times the cost of repair, assumed to be paid at the end of the year, plus the discounted expected present value of landslide repair costs next year, weighted by the probability that the landslide has not yet occurred. If K denotes the expected present value of landslide repair costs, then P 1-P K= C+ K, 1+δ 1+δ or PC K= P+δ 45 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal Sharma 2013; Devkota et al. 2014; Suresh Sharma and cover (LULC - e.g., forest, grassland, cultivation) but is also Maskay 1999). affected by land management practices (e.g., the existence of terracing or current tillage practices). We assign values of The figures presented by (Starkey, Tumbahangfe, and Sharma aboveground carbon stocks by LULC class (in Mg/ha) based 2013) was employed here, as they provide specifications that on data provided in Ruesch and Gibbs (2008); for non-forest can be used to extrapolate to roads in other areas. They classes) and the Ministry of Forests and Soil Conservation report costs of 3.9, 4.6, and 5.9 million NPR per kilometer National Forest Reference Level study (MoFSC 2016; for to construct 4.5-meter-wide earthen roads in three different forest classes). Soil carbon stocks are taken from Dahal and locations, each capable of supporting vehicle traffic at Bajracharya (2012). The average of reported values for average speeds of between 70 and 80 kilometers per hour. areas without sustainable soil management practices were Other roads may be more or less expensive depending on assigned to croplands, grasslands, and orchards, while the their specifications, of course, but these estimates provide average reported for forests was assigned to all forest classes. a benchmark against which to calibrate other costs. Taking Baseline carbon stock values are given in Appendix 5. Our the average of the three cost estimates and adjusting them model assumed that carbon storage is a spatially-independent for inflation between 2013 and 2018, the cost becomes 6.35 ecological process, that is, carbon dynamics are not affected by million NPR (US $56,670) per kilometer of road damaged land cover and management practices in neighboring areas. by landslide. To calculate the carbon sequestration from watershed These figures are adjusted for the width of the road and management interventions, we assume changes in above- topographical factors that will determine the volume of and below-ground and soil organic carbon pools based on material that would need to be removed in the event of a the type of land use land cover at the intervention site and landslide, as summarized in section 3.2.4 above and detailed the type of intervention. For soil and water conservation, in Appendix 3. The costs of repair are assumed to vary hill terrace improvement, degraded forest rehabilitation, proportionally with the volume of material that must be and degraded grazing land rehabilitation, we use data from removed to build or restore the road. Cardinael et al. (2018), which give a mean response ratio reflecting the ratio of soil organic carbon (SOC) before and 3.3.4. Impacts on carbon storage after implementation of a variety of agroforestry practices Land management and rehabilitation practices such as (e.g. hedgerows, tree species intercropped with annual crops). those modeled in this study have been shown to increase These activities generally involve the planting of tree species, aboveground carbon storage (by increasing woody vegetation, thereby increasing above- and below-ground carbon pools as through adoption of agroforestry practices; Cardinael et as well as SOC. SOC is by far the largest carbon pool in al. 2018), and improve soil organic carbon (SOC) stocks any of the relevant land classes, so the response ratios from through enhancing soil organic matter and improving soil Cardinael et al. (2018) were multiplied by the total baseline health (Dahal and Bajracharya 2013; Paudel et al. 2017). carbon pool to give the post-intervention carbon storage. Furthermore, carbon stocks can also be preserved by rehabilitation activities that reduce the risk of landslides. In The benefits associated with watershed management actions addition to the harm they cause to people and structures, will in reality take some time to reach full effectiveness, landslides are also associated with carbon emissions because generally on the order of 10 – 20 years, depending on the they expose carbon sequestered in soil to various atmospheric intervention (Vogl et al. 2017). The post-intervention carbon and hydrological processes that may allow that carbon to storage was therefore adjusted to account for the fact that the form greenhouse gases. Therefore, we evaluate the baseline benefit stream is not constant, but rather follows a trajectory carbon storage in the watershed, and estimate the impacts that reaches 100% of the modeled benefit after a certain of watershed management activities on both (1) additional number of years. For simplicity and in the absence of more carbon stored through vegetation and soil management, and refined agronomic and soil data, we assume a linear trajectory (2) avoided loss of carbon through mitigating landslide risk. from zero to full benefits after 20 years. We therefore scale the post-intervention carbon storage values by 0.43. 3.3.4.1. Carbon added via land management We calculate total current carbon stock in soils, above- 3.3.4.2. Avoided carbon losses in landslides and below-ground biomass using the InVEST carbon We assume that over any given time period, reducing the (C) model (Sharp et al. 2014). The amount of C stored in risk of landslides will reduce the present value of emissions each of these pools depends primarily on land use/land due to those landslides, through a combination of lowering Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  46 overall risk and from the effect of discounting as the reduced management scenario is manifested in the probabilities of risk shifts the expected time to failure farther out into the a slide, which are used to create the total expected damages future. This section explains these dynamics in more detail. within a particular scenario: T First, as elsewhere, we assume that: 1) A potential landslide object (LSO) has an average annual DSCEN = Σ t=0 ρSCEN (tL = t) D (tL) probability of sliding in the absence of treatment. That is, the damages are the present value of the damages 2) Landslide mitigation activities reduce this baseline in a particular year, multiplied by the scenario-specific average annual probability by some amount. probability of a landslide in that year, and summed over all 3) The occurrence of a landslide is an “attracting state” -- future years. that is, once it is reached, the average annual probabilities of a slide no longer apply. The present value of the climate benefit associated with a 4) Landslide mitigation benefits take immediate effect from management scenario is the difference in climate damages the time of treatment. with no treatment and those with treatment: Together, these assumptions imply the probability of a PVCLIM = DBASE – DCAT landslide occurring in any specific year is the average annual probability, scaled down by the probability that no slide has Because the above framework is linear in both the initial carbon occurred prior to that year. stock and the social cost of carbon, we estimated a multiplier for how carbon contained in landslide objects within the We also need to make assumptions regarding the fate of the watershed should be translated to monetized benefits from carbon conditional on a slide occurring. In reality, the fate avoided carbon emissions. We use a background average of carbon in a landslide is highly uncertain and dependent failure probability of 0.05, taken from the baseline model on the site-specific conditions. One extreme would be that output, and a decay rate of carbon following a landslide of all carbon becomes oxidized immediately in that year. 0.2, resulting in a final benefits scalar of between 0.0024 and The opposite (also unlikely) possibility is that all carbon 0.011 (depending on the scenario and the modeled change remains effectively sequestered due to immediate regrowth in landslide risk), which was applied to the avoided carbon and protection from dense vegetation. As an intermediate loss from landslide mitigation activities. assumption, we model the carbon as exponentially decaying from the landslides – that is, a certain fraction of the Therefore, assuming a constant social cost of carbon, remaining carbon becomes carbon dioxide each year, once the climate benefits of landslide risk mitigation are then the landslide has occurred. calculated as: More precisely, each year is associated with an exponential PVCLIM = benefits scalar . SCC . C0 decay emissions trajectory E(t | tL), which gives the emissions in year t conditional on the landslide occurring in year tL: This approach assume that all carbon contained in the landslide object, including (previously) live biomass as well E(t | tL) = r [C0 (1 – r) (t–tL–1)] as soil carbon, decays to the atmosphere at a rate of 20% per Where the term in brackets is the carbon remaining at the year once a landslide has occurred, due to the exposure of start of year t, with C0 being the initial carbon stock in previously sequestered soil carbon and the burial of existing the landslide object. (Implicitly, all emissions prior to the vegetation. There are many complex dynamics associated landslide occurring are zero.) with carbon fluxes that were not feasible to model, so these numbers should be treated as notional values that enable a The climate damages D conditional on a landslide occurring rough estimate of the potential magnitude of this benefit in year tL are simply found by multiplying the emissions stream. trajectory by the social cost of carbon (SCC) and discounting: T 3.3.4.3. Economic value of carbon D(tL) = Σ t = tL SCCt E (t | tL) (1 + δ) –t The above methods result in estimates of carbon either stored additionally or as loss avoided from each watershed This value above is not dependent on the watershed intervention scenario. We calculate the value of this carbon management scenario. Rather, the impact of a watershed benefit by using estimates of the social cost of carbon 47 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal from the 2017 report of the High-Level Commission on is 144 MW, which is designed to be achieved at a maximum Carbon Prices (Stiglitz et al. 2017). While carbon pricing is flow rate of 141 m3/s. For roughly half the year, in the months unavoidably imprecise (see, e. g, EPA 2017, which presents during and following the monsoon, this flow is available. estimates of value spanning an order of magnitude) and During the winter and early spring months, however, the flow controversial (see, e. g., (Editors, Journal of Economic declines, and it is not possible to sustain generation at full Perspectives 2015; Tol 2009, 2014), it is useful to have some power. Figure 3.4 depicts this pattern over a three-year period. monetary estimate of climate-related benefits. The High- Level Commission’s results were endorsed in a November From the late spring through the fall, water is plentiful. This 2017 World Bank guidance note on “Shadow Price of means sediment transport is also high. The concentration of Carbon in Economic Analysis” (World Bank 2017). The sand in water increases dramatically with the velocity of flow Commission report recommended using prices ranging in the river (Morris 2014). This is illustrated in Figure 3.5.10 between US$40 and 80 per ton of CO2e in 2020, rising to Comparing Figure 3.4 and Figure 3.5, peak water flow occurs in $50 – 100 per ton in 2030. The Bank’s guidance note also the mid-summer months, when sand concentration also peaks. recommended continuing to extrapolate results from 2030 to 2050 at the 2.25% annual rate of growth projected in Water-borne sand has two related effects. First, it abrades the Commission report for 2020 – 2030. This would result the turbines and other equipment at the plant, reducing in a range of values between US $78 and $156 per ton in operating efficiency and necessitating their repair. Second, 2050. In this analysis, we follow the guidance of the High- it increases the costs of operating the desanding basins11 that Level Commission on Carbon Prices and use a mid-range were designed to intercept and retain much of the sand that estimate of US $60 with $40 taken as a lower bound, and would otherwise pass through the generating equipment. $80 as an upper bound. During the winter and early spring months, flow in the river 3.3.5. Hydropower benefits is below the level required to achieve the plant’s maximum The effects of sediment on hydropower operations at KGA generation potential of 144 MW. At these times, the plant differ by season. Hydropower operations are greatly affected makes use of its limited storage capacity. By filling the by the rate of water flow in the river. The capacity of the plant reservoir at times of day when the demand for power is  Figure - 3.4: Power generation and river flow at KGA 120000 1400 100000 1200 1000 80000 800 60000 600 40000 400 20000 200 0 0 Jan. 2011 July 2011 Jan. 2012 July 2012 Jan. 2013 July 2013 Power generation in MWh per month (left axis) Average river flow in cubic meters per second (right axis) Flow required to support full generation potential (141 cubic meters per second) Source: Adapted from Chhetry and Rana (2015) 10. Note that the vertical scale in Figure 3.5 is logarithmic: each horizontal line marks an order of magnitude increase. 11. While different terms have been used for these structures, such as “silt traps” or “settling basins”, we chose “desanding basins”, as it seemed to be the term most commonly employed. Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  48  Figure - 3.5: Sand concentration entering turbines by month Source: Morris 2014 relatively low, the operator may make use of the storage to The challenges presented in operating a hydroelectric plant time the discharge of water for power generation during the on a river with high sediment loads have been recognized since hours when it is more valuable. Accumulation of sediment the design phase of KGA. The plant was equipped with two in the reservoir reduces storage and thereby restricts the large desanding basins intended to trap and remove particles operator’s ability to generate power when its value is greatest. that could otherwise damage its generating equipment. It is costly to operate these desanding basins, however. The The remainder of this section will give an overview of the second benefit of reducing sediment concentration in the methods used for estimating these different seasonal benefits water diverted for generation is the avoided cost of operating of sediment reduction. the desanding basins. 3.3.5.1. Avoided costs and damages Ideally, avoided damages and avoided costs associated with Reducing the amount of sediment in the water that is reduced sediment concentration would each be estimated diverted from KGA for power generation will result in with straightforward procedures. Damage would be two economic benefits. First, there will be avoided damages. measured by the reduction in operating efficiency resulting Less hard, coarse sediment passing through the generating from abrasion, and abrasion would be related to sediment equipment means less abrasion. This, in turn, means that concentration.12 This approach would require data relating the turbines will generate more electricity per cubic meter sand concentration in the river to abrasion of turbine parts, of water flow through them, they will be less likely to break as well as data relating abrasion of turbine parts to reduction down, and they may require less frequent maintenance. in efficiency. However, these data sets were not available to 12. Reduced abrasion might also reduce maintenance costs if equipment that had suffered less damage needed to be repaired less frequently. Because of the seasonal variation in river flow and, hence, the possibilities for power generation, however, reductions in sediment delivery are unlikely to affect the maintenance schedule. Each of the three generating units at KGA is generally overhauled on a once-every-third-year rotation, with the work planned to occur during the dry season of the year, when water flow is not sufficient to use all three units at full capacity. Because the opportunity cost of having a turbine out of service for overhaul for several weeks when flows are high are substantial, large reductions in sediment would likely be required to motivate a delay in maintenance from one year’s dry season until the next. 49 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal estimate avoided damages. An indirect procedure has, then, but this would mean that less sand would be removed been adopted.13 before it passed through the generating equipment, and the resultant increase in abrasion would reduce operating This indirect procedure exploits the relationship between efficiency more rapidly. avoided damages and avoided costs. A detailed explanation of this procedure is given in Appendix 4, but the intuition The operator would, then, flush the desanding basins underlying it may be explained using some simple economic following a rule under which the marginal avoided damage ideas. Avoided costs and avoided damages are linked, in from an additional cubic meter of sediment passing through that the dam operator strikes a tradeoff between the two. the generating equipment would just balance the marginal It would be prohibitively expensive to prevent any sediment avoided cost of removing that cubic meter of sediment via from reaching the generating equipment. Conversely, if the desanding basins. This outcome is depicted in Figure 3.6. no expense were incurred, little if any sediment would be prevented from causing damage. Both the cost and There are two axes in Figure 3.6. The left axis indicates effectiveness of the desanding basins depend on the amount the marginal cost of removing a cubic meter of sand from of sand that accumulates in them between flushes. They water diverted for generation by more frequent flushing of become less effective as more sediment accumulates in the desanding basins. The red curve, which rises from left them. The operator could remove more sand by flushing the to right, represents the marginal cost of sand removal. The basins more frequently, and hence spare the turbines some right axis in Figure 3.6 indicates the marginal damage from damage. This would mean, however, that customers would not removing a cubic meter of sand, i.e., from allowing that be provided with less power while the basins are flushed. cubic meter to pass through the turbines. The blue curve, Conversely, the operator could generate a steadier supply of which rises from right to left, represents marginal damage. power if the desanding basins were flushed less frequently, The sum of costs and damages will be minimized when the  Figure - 3.6: Conceptual representation of operating the desanding basins to minimize the sum of costs and damages Marginal damage Marginal cost of resulting from removing a cubic not removing a meter of sand (red) cubic meter of sand (blue) Value of reducing a cubic meter of sand from water diverted for generation Sand not Sand removed removed Total sand in water diverted for generation 13. Two different indirect approaches were, in fact, developed. In addition to the procedure reported below, avoided damages were also inferred from a procedure that asked “how much damage must be occurring every year in order to make the observed every-third-year maintenance schedule optimal?” However, data limitations and other considerations made the results of this exercise less reliable than those of the alternative approach reported here. Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  50 marginal value of each is the same, as represented by the The parallel desanding basins at KGA are each 187 meters point where the red and blue curves cross. The volume of long and 40 meters wide (Bishwakarma 2012). They are sand removed in this efficient outcome is measured from left flushed when the depth of sediment accumulated in them to right on the horizontal axis, and the volume of sand not reaches three meters. Thus, a volume of 187 ∙ 40 ∙ 3 ∙ 2 = removed and, consequently, passed through the generating 44,880 m3 of material is removed with each flush.15 equipment is measured from right to left. The volume of sand removed plus the volume of sand not removed must, Finally the fraction of sediment removed is taken as 0.733, of course, be equal to the total volume of sand present in in accordance with estimates of removal efficiency from water diverted for generation, as represented by the distance an International Hydropower Association study of KGA between the two vertical axes. (IHA, n.d.). This estimate is also broadly consistent with the figures on the volume of material removed per flush, The procedure adopted for estimating the avoided costs information provided by NEA on the frequency of flushing, and avoided damage from a reduction in sediment involves and estimates of the volume of sand transported in the river characterizing the conditions under which the marginal cost (Morris 2014).16 of sand removal and the marginal damage from sand that is not removed are equal; that is, characterizing the conditions Details of the calculations applied are given in Appendix under which the red and blue curves in Figure 3.6 cross. 4, and results of these calculations are reported in Such conditions can be represented as a function of Section 4.2.2. three variables: 3.3.5.2. Retention of peaking capacity 1. The cost of flushing the desanding basins,14 as calculated While it is often described as a run-of-the-river plant, KGA from the value of power generation forgone during the was designed to have more than three million cubic meters of time required for flushing; live storage. This is sometimes described as “six hour peaking 2. The volume of sand allowed to accumulate before capacity” (Morris 2014). The three million cubic meters flushing, as reported in dam operating practices; would be sufficient to operate the plant at full generating 3. The fraction of sand captured and removed, and, by capacity (corresponding to 141 m3/s, or a little over half a implication, the fraction that is not removed, as recorded million cubic meters per hour) for about six hours. in studies of dam operations. The storage capacity of the reservoir has declined over time. Flushing is only necessary during the high-flow periods As documented in the previous sections, approximately when sediment concentrations are high. It can, however, 35 million tons of sediment flow down the Kali Gandaki. be timed to occur in off-peak periods when the opportunity This flow was enough to fill the reservoir’s dead storage (the cost of forgone power is relatively low. A value of 6 NPR volume below its hydroelectric intakes) before the plant began (US $ 0.054) per kWh is assigned for the value of generation commercial operation (Morris, 2014). The annual flow of forgone during flushing (see Annex 4 for details). The basins sediment would be enough to fill live storage several times are flushed sequentially, so power can be produced at half of over if it were all retained in the reservoir. Because flows are full capacity using flow through one basin while the other is rapid when sediment concentrations are greatest, however, being flushed. During the nine hours the basins are flushed, most sediment remains suspended, and is transported out then, about ½ ∙ 9 hours ∙ 144 MW = 648 MWh of power of the reservoir, either by releases over the spillway or, as generation is forgone, at an opportunity cost of 648,000 ∙ 6 discussed in 3.3.5.1, by being flushed from the desanding = 3.89 million NPR (US $ 34,730) per flush. Labor or other basins or passed through the generating equipment. costs of flushing are not estimated for lack of data. These are felt to be small compared to the opportunity costs of forgone While only a small fraction of the annual sediment load is generation, however. retained in the reservoir (IHA, n.d.),17 NEA personnel report 14. The cost of flushing the desanding basins is not identical with the cost of removing sand, as the latter reflects changes in the efficiency of removal resulting from difference in the frequency of flushing. 15. As flushing occurs when river flow is high, and all flushed sediments would eventually have made their way downriver, there are few if any downriver costs associated with flushing. 16. While finer and lighter sediments may not be removed as efficiently in the desanding basins, they are of less concern, as they tend to be less abrasive. 17. At 1.5 tons per cubic meter, some 22 million cubic meters of sediment are transported through the Kali Gandaki Dam yearly. If as much as one percent of this load had settled in the reservoir during the dam’s seventeen years of operation, the reservoir would now be completely filled. 51 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal that this has, over time, reduced live storage, an observation the fixed amount of power production the water flow in the confirmed by bathymetric measurements (Morris, 2014). It river will allow is generated. has also motivated management efforts to prevent further accumulation (Morris, 2014). Retention of reservoir capacity Power demand in Nepal shows marked variation over the may provide benefits, then, both in terms of the availability course of a day. Figure 3.7 shows a daily load curve for of storage to meet peak demand and avoided costs of Nepal. Demand peaks in the evening hours between about preventing further losses. 5:00 and 11:00 pm. When it is possible, the dam operator would prefer to store water during the lower demand periods Appendix 4 presents a detailed analysis of the value of of the mid-day and have it available to generate power when reduced sediment load as it relates to retention of reservoir consumers demand more. storage capacity. That analysis may be summarized as follows. Reservoir capacity is only needed during the dry Reservoir capacity allows such intertemporal switching from season, when flow in the Kali Gandaki River is insufficient low- to high-demand periods. During lower-demand periods to support power generation at the designed maximum flow less water can be discharged than flows in and, consequently, rate of 141 cubic meters per second. During and after the less electricity will be generated. By refilling the reservoir, annual monsoon, flow may be several times this rate (see however, the water that is flowing in at the time may be Figure 3.4 above). When flow is below the rate required to discharged later in the day, generating more power when support maximum generation, the total amount of power it is more valuable. This underscores the basic principle of that can be generated over the course of a day depends valuing reservoir capacity. The value of an extra cubic meter on the total volume of flow in the river over the day. The of reservoir capacity is the difference between the value of limited storage capacity of the reservoir does not determine power at peak and off-peak periods.18 It is the difference how much power can be produced. Rather, storage capacity between the value consumers get from a little electricity allows the plant operator to determine when during the day provided when they value it more and the value they forgo  Figure - 3.7: Daily electricity demand (NEA 2018) 18. It might also be possible to derive this value from the monetary or opportunity costs the dam operator would incur to restore a marginal cubic meter of capacity. If the dam is operated optimally, however, this procedure should give the same result; see Appendix 4. Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  52 by having a little less electricity generated when water is kept including appraisals of proposed projects in Ethiopia (World in storage rather than used for off-peak generation. Bank 2008), Nigeria (World Bank 2018a), the countries of the Eastern Nile Basin (World Bank 2009a), China (World Figure 3.7 also suggests a reason for which the estimation Bank 2009b), and several states in India (World Bank 2012, of this difference in values may be challenging. The purple- 2005, 2014). They include implementation or enhancements shaded peak in the figure is labeled “load shedding”. At of terraces; hedgerows, trenches, eyebrow pits and bunds; many times the quantity of power demanded in Nepal planting of grasses, trees, and other vegetation; and reform has exceeded system supply. At such times the rates paid of grazing management. These programs were generally by consumers for the power they purchase may not reflect motivated by an appreciation not just of the benefits that how much they would be willing to pay to purchase more might arise from physically implementing them, but also power, if it were available19. This presents challenges for the of the need for institutional reforms to achieve benefits estimation of the value of power during periods of shortage. landowners had not realized on their own. That is, there These challenges are addressed in more detail in Appendix may be many reasons why landholders are not adopting such 4. In short, a figure of 12 NPR (US $0.108) per kWh is practices voluntarily, but policies are needed to better align adopted for the value of peak power, based on NEA tariffs incentives and overcome barriers to adoption (a broader and some additional considerations as detailed in Appendix discussion of this topic is given in Section 5). Moreover, each 4. The resulting difference between peak and off-peak values of the project appraisals note substantial on-site benefits is 6 NPR (US $ 0.054) per kWh. of better land management. Several predict substantial improvements in agricultural productivity as a result of This figure must be multiplied by the number of days in a program implementation. All of the appraisals project year (assumed to be 180) during which reservoir capacity benefits in excess of costs. In fact, some of the World Bank constrains operations, and a conversion factor giving the Project Appraisal Documents relied only on the agricultural amount of electricity produced per cubic meter of flow benefits of land management practices in their cost-benefit through the turbines (0.284 kWh/m3). Finally, as in all the analyses (e. g., World Bank 2009a, 2008). analyses conducted for this report, a discount rate of 10% is applied to arrive at a net present value. It would not be surprising, then, to find that substantial on-site benefits would be associated with sustainable land 3.3.6. Local (on-site) benefits management practices in Nepal, since such benefits have been identified for other countries at a similar income level Section 1 discusses the various benefits that might arise from that have adopted similar sets of measures. However, we have management interventions implemented on crop, pasture, identified few detailed cost-benefit analyses of sustainable and forest lands. The costs of watershed management are land management measures adopted for Nepal (an exception typically borne by the owners and users of the land on which is Das and Bauer (2012), which finds that hedgerow planting the soil conservation and related practices are implemented. and minimum tillage practices have a positive benefit-cost The benefits of such practices might accrue both to the ratio at a discount rate of 10%). people implementing them and to downstream beneficiaries. The previous sections have largely considered downstream Another set of data does afford some insight into the benefits: sedimentation of reservoirs, damage to power magnitude of benefits likely to be realized by land users generating equipment, and risks to lives and property from when terracing, hedgerows, reforestation, improved grazing landslides. However, landowners may also realize substantial management, and other sustainable land management benefits from erosion control measures they adopt. strategies are adopted, however. The World Overview of Conservation Approaches and Technologies (WOCAT) A number of studies have been undertaken to estimate Sustainable Land Management database contains data economic benefits associated with sustainable land on the costs of establishing and maintaining a number of management programs. The World Bank has appraised different land management practices. It lists information many proposed watershed management projects that drawn from almost 2,000 examples from over 130 countries, consider many of the same activities contemplated for Nepal, including several dozen from Nepal. 19. This may be true at other times, as well, since unlike in a competitive market, what consumers are willing to pay for power may not reflect the full societal cost of providing that power. This issue is discussed in more detail in Appendix 4. 53 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal In Table 3.4, cost data on the types of land management last column in Table 3.4 is labeled as “benefits implied by practices that could be used to control erosion in the Kali land users’ cost share”. As participation in the programs Gandaki watershed have been assembled from eleven is voluntary, the benefits the users perceive they will gain WOCAT studies (nine in Nepal, and another two in must be at least as great as the costs they would bear neighboring India). The column headed “Total NPV of from establishing and/or maintaining the indicated gross cost” gives the net present value (NPV) of establishing practices. Local users’ cost shares are also broken out and maintaining the indicated practice. The WOCAT by establishment and maintenance costs in the data. On data generally break down costs by labor, materials, and average, local users bear about 84 percent of the costs of other expenses, and report them for both the one-off costs these practices. of establishing a practice and the ongoing costs of its maintenance. The latter were discounted at 10% per annum The costs reported in Table 3.4 are, then, multiplied by the (Table 3.4).  factor of 0.84 identified above, on the assumption that cost- bearing share is representative of the broader set of cost These eleven studies are particularly useful for present data available. We then use these impute on-site benefits of purposes because they also contain information on the the interventions modeled in the benefit-cost calculations share of costs borne by local landholders. The second-to- reported below. Table - 3.4: Costs to implement various watershed management activities and benefits to landholders implied based on reported cost sharing Practice Location Total NPV of Benefits implied B/C ratio gross cost by land users’ implied by users’ (US$ per ha) cost share benefits alone Terrace Nepal $ 8,746 $ 7,581 0.87 Ditches, bunds, tree and grass planting Nepal $ 5,598 $ 4,301 0.77 Hedgerows Nepal $ 1,361 $ 1,361 1.00 Contour bunding Nepal $ 52 $ 52 1.00 Contour trench/bund India $ 1,075 $ 323 0.30 Gully plugging with check dams and Nepal $ 725 $ 725 1.00 bamboo planting Controlled gullying by building Nepal $ 69 $ 69 1.00 retaining walls and plantings Riverbank protection by check dams Nepal $ 5,391 $ 4,053 0.75 and grass and bamboo planting Hedgerows, eyebrow pits and trenches, Nepal $ 1,029 $ 1,029 1.00 planting trees and grasses Fodder cultivation on terraces, Nepal $ 2,203 $ 2,203 1.00 abandoned agricultural land Contour trenches, tree and grass India $ 2,308 $ 1,330 0.58 planting Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  54 3.4. WHERE TO INTERVENE? To evaluate each activity’s effectiveness in different locations, a series of hypothetical “full implementation” scenarios are PRIORITIZING WATERSHED created one activity at a time, to represent the landscape as MANAGEMENT ACTIVITIES AND if the activity were implemented everywhere it is possible. LOCATIONS We also generated scenarios to represent combinations of activities (for example, hill terrace improvement, soil The first step in prioritizing where different activities should and water conservation, and forest rehabilitation are be implemented is to understand where in the watershed each simultaneously implemented wherever they are possible). activity can be most effective to achieve a set of objectives, Each scenario is then run through the relevant model (SDR and then use a multi-dimensional optimization approach or landslide) and the total change is estimated for each of the to identify a set of optimal portfolios of interventions that objectives in each of the 821 sub-watersheds. maximize objectives at minimal cost. The objectives used in the optimization are listed in Table 3.5. An optimization approach was deemed to be appropriate for this study, because it allows for development of investment We divide the study area into 821 hydrologically-defined portfolios that meet multiple objectives, and explicitly reports sub-watersheds, with an average size of approximately 900 on trade-offs that exist when prioritizing one kind of benefit ha. Based on stakeholder consultations with DoFSC, this over another. Another approach would be to simply rank the is roughly the size of the individual micro-watersheds that sub-watersheds and activities in terms of the highest benefit DoFSC typically addresses through their current watershed per unit cost, and to select the areas with the highest benefit management programs. Each of these sub-watersheds sequentially until a given budget is exhausted. However, this becomes a “decision unit” – spatial regions representing the method requires that the metric to maximize is selected a smallest area on which an activity (or group of activities) will priori and does not allow for explicit examination of trade- be implemented. While it is technically possible to optimize offs inherent in making that decision. activities at a pixel scale, that level of precision does not align well with the underlying model assumptions, nor is it For this analysis, the Natural Capital Project’s ROOT a feasible unit to implement activities under a community- tool is applied to perform the optimization (Beatty et al. based watershed management program. 2018; Gourevitch et al. 2016). ROOT first summarizes the Table - 3.5: Objectives used to prioritize watershed management activities and locations Objective Unit Beneficiary Valuation approach On-farm benefits of soil Tons of Local landholders Revealed preference based retention sediment/yr on reported cost-share from similar programs Avoided sediment reaching Tons of KGA hydropower plant Avoided damage Kaligandaki reservoir sediment/yr Avoided costs of desanding Peaking capacity maintained Avoided lives lost from USD People at risk from landslides Value of statistical life landslides Avoided damages to USD Structures and associated communities at Rental rate structures risk from landslides Avoided repairs to roads USD Dept of Roads, VDCs and communities Avoided repair costs at risk from landslides Added carbon storage Metric tons National (e.g. REDD+ program), Global Social cost of carbon in 2020 55 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal marginal value of each activity within each sub-watershed Ts is the target value for each objective s. Here we use the into a table. For each of the potential management options, target to constrain the cost, representing different levels of the table contains the value to each objective for each sub- investment in a watershed management program. We ran watershed (calculated as the sum of pixel-level marginal the optimization at a budget constraint ranging from US values within each sub-watershed). $500,000 US $50M. At each of these budget scenarios, a portfolio of interventions was generated with the objective ROOT implements the optimization using binary integer of maximizing the monetized benefits of sediment retention, programming. Formally, the problem is to find the optimal avoided loss of structures, avoided road repairs, and avoided → xij, where the value of each xij is 1 if management option loss of life. After portfolios were generated, on-farm benefits j is chosen for sub-watershed i and 0 if it is not. If all the and the value carbon sequestration were calculated. xij’s are zero for a given sub-watershed, then the choice is to maintain current (baseline) land use. The results show the optimal portfolio of interventions for a given budget, by identifying which sub-watersheds should The optimization problem is be selected for which intervention to maximize benefits and → min C ( xij) minimize costs. We also identify intervention portfolios using → xij different weights on the objectives, to demonstrate how the such that targeting of watershed management activities might change → Vs ( xij) > Ts depending on whether the program prioritizes sediment where reduction, landslide risk mitigation, or reducing on-site → C ( xij) erosion, for example. The model also outputs an agreement is the total cost of the selected management options. map, showing how often each spatial decision unit (SDU) → Vs ( xij) is chosen for a particular activity, regardless of the weight given to different objectives. is the value to objective s of the management choice. The value is given in terms of avoided sediment, avoided damages, etc. (column 2 in Table 3.5 above). Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  56 57 3.5. SUMMARY OF DATA REQUIREMENTS Table - 3.6: Sources and descriptions of data used in this study Data type Description Format/resolution Source General landscape characteristics Elevation ASTER Digital elevation model GeoTIFF, 30m METI/NASA: https://asterweb.jpl.nasa.gov/ resolution gdem.asp Land use and land cover Nepal national land use/land cover, year 2000 GeoTIFF, 30m Nepal Department of Survey resolution Cropping patterns District data on land use and crops grown in the Table of values by Nepal Department of Irrigation district district Degraded forest lands Forest loss/disturbance/change occurring GeoTIFF, 30m Hansen/UMD/Google/USGS/NASA Global between years 2000-2018 resolution Forest Change 2000–2018, version 1.6. https:// earthenginepartners.appspot.com/science-2013- global-forest/download_v1.6.html Hillslope erosion Rainfall erosivity A measure of the intensity of rainfall, estimated GeoTIFF, 1km resolution Rainfall data: WorldClim Version 2 current; from annual precipitation data Erosivity calculation from FAO Soils Bulletin 70, Roose (1996) Soil erodibility Soil property based on texture, indicating how GeoTIFF, 1km resolution ISRIC Soil Grids - https://www.isric.org/ easily the soil detaches to become erosion explore/soilgrids USLE C factor Erosion potential factor based on vegetation Unitless, mapped to land A variety of published literature sources - See type use/land cover Appendix 5 USLE P factor Support practice factor based on management Unitless, mapped to land A variety of published literature sources - See practice type use/land cover Appendix 5 Glacial erosion Precipitation 10-year time series of daily rainfall from Point data Nepal Department of Hydrology and national weather stations Meteorology (DHM) Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal Data type Description Format/resolution Source Location of glaciers Location and extent of glaciers in 2010 Polygon shapefile ICIMOD, dowloaded from: http://data. thethirdpole.net/layers/65 Landslides Elevation ASTER Digital elevation model GeoTIFF, 30m METI/NASA: https://asterweb.jpl.nasa.gov/ resolution gdem.asp Soil Depth Soil depth to bedrock GeoTIFF, 1km resolution ISRIC Soil Grids: https://www.isric.org/explore/ soilgrids Precipitation Daily precipitation data at 35 stations from Point locations of Nepal Department of Hydrology and 1995 - 2014 stations Meteorology (DHM) Hydrologic soil type Required to divide precipitation into runoff and GeoTIFF, 1km resolution ISRIC Soil Grids: https://www.isric.org/explore/ infiltration components soilgrids Infrastructure at risk Locations of structures and roads Line and polygon Open Street Map: https://www.openstreetmap. shapefiles org Land use and land cover Nepal national land use/land cover, year 2000, GeoTIFF, 30m Nepal Department of Survey used to determine root cohesion resolution Road-induced erosion Elevation ASTER Digital elevation model GeoTIFF, 30m METI/NASA: https://asterweb.jpl.nasa.gov/ resolution gdem.asp Road locations Locations of structures and roads Line and polygon Open Street Map: https://www.openstreetmap. shapefiles org Road width and surface Estimated from reported road types Assigned by road type, Open Street Map: https://www.openstreetmap. material width in meters org Fluvial sediment connectivity Elevation ASTER Digital elevation model GeoTIFF, 30m METI/NASA: https://asterweb.jpl.nasa.gov/ resolution gdem.asp Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  58 59 Data type Description Format/resolution Source Discharge Observed discharge at DHM gauging stations, Point locations of gauges Nepal Department of Hydrology and approx. 5 years of daily records Meteorology (DHM) Channel width Manually sampled from Google Earth Numeric value assigned Variety of satellite image sources as provided by to each channel Google Earth Carbon Carbon pools Aboveground, belowground and soil carbon Tonnes/hectare, mapped Most values from Reusch & Gibbs New IPCC pools by land cover & management type to land use/land cover Tier-1 Global Biomass Carbon Map for the Year 2000; Forest values from the National Forest Reference Level of Nepal (2000-2010) Land use and land cover Nepal national land use/land cover, year 2000 GeoTIFF, 30m Nepal Department of Survey resolution Carbon values US$ 40 - 80 per ton CO2e sequestered Value per ton CO2e Stiglitz, et al., 2017. Activities and implementation costs Costs of sustainable land Online databases and peer-reviewed literature Cost per hectare WOCAT SLM database: https://qcat.wocat. management (SLM) on costs of implementing SLM, reported as or net/en/wocat/; Dahal, Hasegawa, Bhandary, & practices converted to 2018 US $ per hectare Yatabe, 2010; Das & Bauer, 2012; Devkota et al., 2014; Sharma 2003 Economic valuation Value of power generated Tariff rates, costs of imports and purchases 2018 US$ per KwH Nepal Electricity Authority Annual Reports, 2009 from independent power producers, Nepalese – 18. rupees (NPR) per kWh by conditions of purchase and time of day, converted to 2018 US$. Estimates of shadow price of unserved demand Nepalese rupees (NPR) J. P. Shrestha & Shrestha, 2016; R. S. Shrestha, during periods of load shedding or US$ per kWh 2011; Timilsina & Toman, 2016; Karki, Mishra, & Shrestha, 2010 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal Data type Description Format/resolution Source Maintenance Scheduling, duration, and costs of overhauls, NPR converted to 2018 Nepal Electricity Authority Annual Reports, 2009 flushes, and other procedures US$ – 18; data obtained from NEA; interviews with NEA personnel; Morris, 2014, and Bishwakarma, 2012, for physical dimensions. Reservoir capacity Initial reservoir capacity and subsequent Millions of cubic meters Morris, 2014; interviews with NEA personnel changes Value of homes at risk Rental value of rural homes, pro-rated per NPR per annum, Nepal Central Bureau of Statistics Annual from landslide damage meter of area converted to 2018 US$ Household Survey 2016/17. Area from data on structures at risk. Cost of road repairs from Costs of road construction, corrected for road NPR per km converted Road construction costs taken from (Starkey, landslide damage width to 2018 US$ Tumbahangfe, & Sharma, 2013); corrections for with from OSM data. Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  60 4. RESULTS © Oliver Foerstner/Shutterstock.com 4.1. BASELINE CONDITIONS fracturing of rocks along fault lines are critical for improving understanding of where and how sediment is generated in 4.1.1. Sediment budget the watershed. The sediment budget for the 5 major sub-watersheds of the Kali Gandaki was determined from sediment measurements A comparison of observed sediment load to our multi- performed by Kathmandu University. These measurements model approach with separate models for hillslope erosion help to determine the contributions of the Mustang Plateau, (SDR), landslides, roads, and glaciers shows that the models the Upper Kaligandaki, the middle and lower Kali Gandaki, generally perform well in terms of total modeled sediment the Modi Khola and Myagdi Khola tributaries (see also Figure loads, although the models tended to over-predict sediment 2.6). A key finding is the great diversity in sediment load and load from some tributaries and under-predict load from the yield, which is not aligned with the spatial distribution of Mustang and the Upper Kaligandaki. rainfall in the sense that the tributary watersheds receiving most of the watershed’s precipitation do not contribute the Because landslides make up the largest part of the sediment most to the its sediment budget. This suggests that improved budget of each sub-watershed, we focus calibration on the data on geologic factors such as lithology, uplift rates, and landslide model, which is in line with the understanding that 61 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal landslides and other mass movements are the most important need for ongoing monitoring at multiple locations to build factors in the sediment budget of this region (Struck et al. a longer record of sediment dynamics, and ideally to also 2015). Specifically, we modify the soil cohesion in each sub- collect more evidence on what processes produce sediment watershed (see Appendix 1 for details). This assumes that in different locations. each sub-watershed is a homogeneous unit with regard to the geomorphic processes impacting landslides. While this is To identify which processes generate most sediment in a simplification, it should be noted that these units are indeed different parts of the watershed, we use the previously distinct with regard to their topography, climate and geology defined 821 sub-watersheds as unit of analysis (Figure (lithology, uplift, fracturing), key factors that influence the 4.2, left panel). Notably, landslides produce most of occurrence of landslides. the sediment in the upper and middle watershed, while hillslope erosion dominates in the south, center-west and Model calibration greatly improved the model’s fit to around the rim of the Mustang Plateau (Figure 4.2, right observations (Figure 4.1). Hillslope erosion (red) makes up panel). The sediment load of very few, high elevation only for a small part of the observed sediment load of the sub-watersheds is dominated by glaciers. Roads are not various sub-watersheds (black squares). According to other dominant in any major sub-watershed. Figure 4.3 shows observations, we assumed that the majority of sediment the sediment generation by process and sub-watershed, in the watershed is generated by landslides (yellow) and so and Table 4.1 gives the modeled sediment yield for model calibration focused on the landslide model as described the most dominant processes – hillslope erosion and above. Results show that landslides make up for a majority landslides – by land use type. These maps are a first step of each sub-watershed’s sediment load, and especially in the in understanding which activities may be implemented to upper Kali Gandaki (draining to Nayapul) and the middle manage sediment, as interventions will be most effective Kali Gandaki (draining to Modi Beni). The high diversity when they are targeted to the dominant sources of in sediment load between sub-watersheds points to the sediment in the relevant areas.  Figure - 4.1: Comparison of modeled and observed load from the multi-model suite, including the calibrated mass- movement/landslide model (yellow) Observed load is the same as Figure 2.7, error bars indicate ± 1 standard deviation in observed loads X106 12 Glaciers SDR 10 Landslides Roads Observations 8 Standard deviation Sediment load [t/yr] 6 4 2 0 som l ani ni at aki u yap e alg top dib nd Jom ngh Na liga Ta Mo Ma Ka Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  62  Figure - 4.2: Sediment load from each sub-watershed to the streams (left) and the processes dominating sediment load in each sub-watershed (right) Table - 4.1: Total mean sediment load and sediment yield (per unit area) for major land uses in the study area. Note that glaciers are not included, because their contribution to sediment loads are calculated separately and do not include hillslope erosion nor landslides. Land use Sediment load Sediment yield, Sediment yield, Sediment yield, (t/yr) total hillslope erosion landslides (t/ha/yr) (t/ha/yr) (t/ha/yr) Cliff 241,600 167.5 65.7 101.8 Cultivation 5,008,400 47.2 30.1 17.1 Forest 5,934,800 38.2 1.7 36.5 Grass 7,118,100 45.4 3.8 41.6 Barren Land 4,806,300 17.9 8.2 9.7 Bush 3,969,200 108.1 12.3 95.8 Pond or Lake 182,600 65.5 34.3 31.2 Sand 314,400 31.9 19.4 12.5 Waterbody 42,200 28.5 17.1 11.4 Built Up 2,100 6.8 2.7 4.1 Nursery 800 3.9 2.3 1.6 Airport 1,000 4.2 1.8 2.4 Scattered Tree 3,900 13.7 3.0 10.6 Orchard 500 0.5 0.1 0.4 63 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  Figure - 4.3: Sediment load from each process in the sediment budget and each sub-watershed to the streams. Processes considered are: Hillslope erosion (a), landslides (b), roads (c) and glaciers (d). Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  64 If all possible watershed management activities that we model class. Less than 10% of all buildings are on landslides or on here were to be implemented in the Kali Gandaki watershed downstream runout pathways (Figure 4.4, yellow line in left (covering about 39% of the total land area, as barren panel). This finding is logical, because it is unlikely that a lands, glaciers, cliffs, built up areas, etc. were excluded from high percentage of buildings will be constructed in places consideration), the total avoided sediment is approximately that are obviously at risk of landslides. Of all buildings at 6.5M tons/year, or 20.5% of the estimated fine sediment risk, most fall in a low risk category (<10 %) and even in this load of 31.7 Mt/yr. A previous study (World Bank 2018b) category, many more structures are at risk because they are estimated a possible 8% reduction in sediment from land located on a potential runout pathway, rather than directly management activities in the lower watershed only; however, on a landslide object (LSO). These results are then the basis that study did not include landslide mitigation measures and for the economic analysis, which considers monetary losses only considered activities in the Middle and High Mountain because of destroyed structures and lost lives. Population physiographic regions. These findings make sense in light of centers that are located within the runout path of landslides the fact that Himalayan geology is known to be unstable and are places where a high risk of sliding corresponds to a high background sediment production very high. Understanding density of values at risk. the scope for watershed management to control sediment problems can help to set realistic expectations for what such The percentages at risk are much higher when it comes to programs can achieve. In reality, a combination of green roads. In total, more than 40% of roads are at risk (Figure and grey engineering solutions will likely be needed to fully 4.4, yellow line in right panel). Again, most of the segments minimize the negative impacts of sedimentation in this area. at risk (around 17% of all segments) are in the lowest (< 5% failure probability). However, compared to houses, a much 4.1.2. Landslide risk greater percentage falls into higher risk classes (10 – 50% pa). Landslide hazards are unique in that they not only produce Similar to houses, there are much more roads at risk because sediment, they also threaten lives and infrastructure. The they are on a runout path, rather than because they are results of our landslide hazard mapping assess the landslide directly located on an LSO. Figure 3.2 shows a comparison hazard of different structures and roads in the watershed of the modeled high-risk areas for both landslides and their (Figure 4.4). The x-axis shows groups of buildings (left panel) runout potential, overlaid with data on homes and roads. To and roads (right panel), grouped by the failure probability of the extent that these data sets could be improved in future their associated landslide and runout hazards, and the y-axis versions of this work, a more complete picture of assets at reports the percentage of the total buildings/in each failure risk could be developed.  Figure - 4.4: Buildings (left) and roads (right) at risk, binned by the failure probability of the landslide/runout they are located on. Lines show cumulative values. 25 50 On LSOs On LSOs On Runout 45 On Runout Cumulative, on LSO Cumulative, on LSO 20 Cumulative, on Runout 40 Cumulative, on Runout Cumulative, all Cumulative, all 35 Percent of road segments [%] Percent of structures [%] 15 30 25 10 20 15 5 10 5 0 0 0 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.05 0.15 0.25 0.35 0.45 0.55 0.65 0.75 0.85 0.95 0 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.05 0.15 0.25 0.35 0.45 0.55 0.65 0.75 0.85 0.95 Failure probability[-] Failure probability[-] 65 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal 4.2. ECONOMIC VALUES OF 4.2.1. The value of an optimal portfolio WATERSHED MANAGEMENT Results for watershed management portfolios ranging in cost from US $500,000 to $50M show that such programs can In this study, we derived a net present value (NPV) of five have a significant, positive impact across many sectors (Table benefit streams that result from implementation of the 4.2). The benefits are driven largely by local benefits and the watershed management activities described in Section 3.3.1: value of avoided lives lost in landslides, with the next highest 1) reduction of sediment to benefit the KGA hydropower beneficiary being downstream hydropower (Figure 4.5). facility, 2) avoided damages to structures and roads due to landslide mitigation, 3) avoided lives lost due to landslide At a US $500,000 budget, each $1 invested yields $4.38 mitigation, 4) changes in carbon stocks, and 5) on-site in benefits, but this ratio drops as budgets are increased. benefits to landholders. We use a discount rate of 10%, a However, even with an investment of US $50M, the program value consistent with practice among development agencies still has a positive benefit: cost ratio, even without considering (Bonzanigo and Kalra 2014) and deemed appropriate the carbon sequestration benefit. There is a large increase in by some commentators for Nepal (Das and Bauer 2012). carbon benefits between US $20M and $50M, this is due to Additional, non-monetized benefits from the program of the fact that costs were allocated in these portfolios without interventions could include improvement in water quality consideration for carbon benefits. The interventions with the and water flow in streams for drinking water and irrigation, greatest non-carbon benefits are for mitigating landslides, improved water infiltration and regulation for local springs, which accrue relatively lower carbon benefits. Once all sites water flows for downstream fisheries, and biodiversity. We for cost-effective landslide mitigation are treated, and budget focus here on the five monetized benefits and examine how is still available, the focus of the intervention portfolios shifts benefits scale as a function of implementation budget and toward reclamation of degraded forest and grazing land, the program’s primary objective(s). which carries with it higher carbon sequestration benefits.  he multiple values of watershed management. The benefits are driven largely by local benefits and the Figure - 4.5: T value of avoided lives lost in landslides, with the next highest beneficiary being downstream hydropower (KGA). Note that X-axis location only represents distinct budget scenarios; it is not proportional to the cost of each portfolio $70 $60 USD, Millions $50 Carbon $40 Lives Roads $30 Structures $20 Hydropower Landholder benefits $10 $ $ 0.5M$ 1.0M$ 2.0M$ 3.0M$ 5.0M$ 7.0M$ 10.0M$ 20.0M$ 50.0M Budget Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  66 In the following sections, detailed results for each benefit generation should be set equal to the marginal damage of stream are given and discussed in the context of a single not reducing a cubic meter of such sediment. This marginal illustrative portfolio: that of a US $1M investment, which cost is calculated from three factors: the cost of flushing has an overall benefit: cost ratio of 3.2. the desanding basins, the volume of sediment disposed with each flush, and the fraction of all sediment in water Figure 4.6 shows the benefit: cost ratio of the modeled diverted for generation that is removed and flushed from the portfolios of interventions, including high and low bounds desanding basins. on the estimated total benefits. These bounds are based on potential values for each benefit stream using a range of Combining this information and using the formula derived parameter estimates in the economic valuation models (see in Annex 4, the net present value of a one cubic meter Sections 4.2.2 through 4.2.5 below for information on how reduction every year in perpetuity is computed to be 1998 these ranges were developed). These ranges illustrate that NPR (US $17.84) at a discount rate of 10%. the positive economic benefit of watershed management interventions is relatively robust to model assumptions but There are a great many factors that affect the calculation of should not be interpreted as confidence intervals. this number. It would take considerable effort and study to quantify the uncertainty of the estimate. As an illustration, 4.2.2. Value of sediment reduction to Kali it has been assumed that off-peak power is priced at $0.054 Gandaki A per kWh. This assumes that markets clear at this price and In this section, monetary estimates of avoided damages that the price reflects the full societal cost of power. If some and avoided costs are first presented following the process demand went unmet during periods of desanding basin outlined in Section 3.3.5, followed by estimates of the value flushing, however, a higher price might be inferred; this of retained reservoir capacity. More details on the methods might be especially true if alternative generators with local used are given in Appendix 4. air quality or global climate implications were used in such periods. If it were supposed that the value of off-peak power Avoided costs and damages were $0.08 per kWh, the value of a cubic foot reduction Section 3.3.5 argued that the marginal cost of reducing would increase to $26.43. Conversely, expansion in system a cubic meter of sediment from water withdrawn for capacity might lead to a reduction in the price of off-peak  enefit/cost ratio of modeled portfolios (blue points), showing high and low boundaries on estimates (lines). Figure - 4.6: B High and low bounds are based on calculations of ranges of potential values for each benefit based on parameter ranges given in the text. These ranges should be considered illustrative and not to be interpreted as confidence intervals 9 8 7 6 Benefit/Cost 5 4 3 2 1 0 0.5 1 2 3 5 7 10 20 50 Budget (millions USD) 67 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal Table - 4.2: Breakdown of values of investment in watershed management and benefit/cost analysis, for budgets ranging from US $500,000 to $50M Budget Values to Values of landslide reduction On-site Carbon TOTAL VALUE (USD) hydropower benefits value based from based on % on social sediment Avoided costs of replacement & repair Avoided Value of cost-share cost of USD Benefit/ reduction to lives lost, avoided carbon Cost KGA mean per lives lost Avoided Avoided Avoided Avoided year (VSL) structures loss of roads at costs at risk (n) structures risk (km) of road value repairs $ 0.5 M $ 76,000 17 $ 42,000 3.3 $ 189,000 4.20 $ 1,451,000 $ 420,000 $ 12,000 $ 2,190,000 4.38 $ 1.0 M $ 121,000 23 $ 69,000 3.6 $ 206,000 5.67 $ 1,959,000 $ 840,000 $ 24,000 $ 3,219,000 3.22 $ 2.0 M $ 256,000 32 $ 95,000 5.0 $ 286,000 7.93 $ 2,740,000 $ 1,680,000 $ 35,000 $ 5,092,000 2.55 $ 3.0 M $ 415,000 40 $ 126,000 5.2 $ 296,000 9.88 $ 3,413,000 $ 2,520,000 $ 75,000 $ 6,845,000 2.28 $ 5.0 M $ 760,000 52 $ 179,000 5.3 $ 302,000 12.95 $ 4,477,000 $ 4,200,000 $ 140,000 $ 10,058,000 2.01 $ 7.0 M $ 1,056,000 61 $ 215,000 5.6 $ 320,000 15.15 $ 5,238,000 $ 5,880,000 $ 196,000 $ 12,904,000 1.84 $10.0 M $ 1,592,000 66 $ 242,000 6.8 $ 385,000 16.54 $ 5,717,000 $ 8,400,000 $ 289,000 $ 16,626,000 1.66 $20.0 M $ 2,865,000 71 $ 261,000 8.2 $ 462,000 17.71 $ 6,123,000 $ 16,800,000 $ 560,000 $ 27,071,000 1.35 $50.0 M $ 4,370,000 78 $ 290,000 9.4 $ 530,000 19.51 $ 6,745,000 $ 42,000,000 $ 3,764,000 $ 57,699,000 1.15 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  68 power. Increased in electricity imports from India, as well as of $0.054 per kWh, it were assumed that the off-peak price increases in domestic capacity resulting from the opening of were $0.08 per kWh, the value of a cubic meter of storage the new Upper Tamakoshi Plant, for example, might make would decline to $184.70. Conversely, if peaking capacity is power more plentiful, and hence cheaper. If off-peak power slow to be augmented, the value of storage might increase. If, prices declined by one quarter, to $0.041 per kWh, the cubic instead of a difference of $0.054 per kWh between peak and foot of sediment load reduction would only be worth $13.38. off-peak prices, a difference of $0.08 per kWh were assumed, the value of storage would increase to $405.30 per cubic US $17.84 is the estimate value of a cubic meter of sediment meter. Such a difference might also reflect environmental in water diverted for power generation. It is estimated, externalities associated with alternative peaking generation. however, that only about 15% of all sediment transported As with other comparisons of valuation outcomes, it should in the river is carried in water that is diverted for generation. be appreciated that there are many, many other sources of Thus, when crediting watershed management interventions variation in both economic and physical calculations that for the avoided costs and damages they achieve, a value of would affect values; thus, these calculations illustrate how 0.15 ∙ $17.84 = $2.68 per cubic meter of loading reduced is estimate values might vary with some such variations. used (leading to a low-end estimate of $2.01 and a high-end estimate of $3.96). Taking the mid-range estimate of US $273.60 per cubic meter, an additional adjustment must be made to this figure The fraction 0.15 is given in an International Hydropower – which provides the value of a cubic meter of sediment Association study of sediment damage and management occupying space in the reservoir – to arrive at the value of a at the Kali Gandaki Plant (IHA, n.d.). A figure of this reduction in sediment loading to the river by one cubic meter magnitude is plausible. In a detailed study of the Kali (which is the output of the sediment model). It is assumed Gandaki, Morris (2014) finds that the concentration of sand that one quarter of one percent of sediment carried in the suspended in water varies as the fourth power of the rate of river settles in the reservoir. flow. This implies that water flowing at 1,000 cubic meters per second would carry 16 times as much sediment as would Before explaining how this fraction is estimated, it may water flowing at 500 cubic meters per second. From Figure be useful first to consider an upper bound on it. The live 3.4 and Figure 3.5 above, it can be seen that the vast majority storage capacity of the reservoir is about 3 million cubic of sediment is delivered during the few months of the year meters. Annual sediment transport is about 22 million cubic when average flow is on the order of 1,000 cubic meters per meters.20 The plant has been in operation since 2002. If, for second. As 141 cubic meters per second – 14.1% of 1,000 example, 3 million m3/(17 years ∙ 22 million m3/yr) = 0.8 cubic meters per second – are diverted for power generation, percent of the sediment transported in the river had settled a figure of 15% does not seem unreasonable. In any event, in the reservoir, it would already be completely filled. The the value of watershed management varies proportionally fraction must be less than 0.8 percent then. with the assumed fraction of sediment borne in water diverted for generation. If the fraction were assumed to be The International Hydropower Association study cited 10, rather than 15%, value would be reduced by a third; if above estimates that less than one-tenth of one percent of 25, rather than 15%, value would be increased by two-thirds. sediment in the river settles in the reservoir. This is consistent with Morris’s (Morris 2014) estimate that approximately Retention of peaking capacity seven percent of live storage in the reservoir was lost in the Based on the calculations and assumptions presented in first decade of plant operation: seven percent of 3 million Section 3.3.5, we arrive at a net present value of US $273.60 cubic meters would be about 210,000 cubic meters over ten per cubic meter of reservoir storage space retained. Again, years, or around 21,000 cubic meters per year, or a little some sense of how this number might vary with assumptions less than one tenth of one percent of the 22 million cubic can be developed by considering alternative scenarios for meters annual sediment load. NEA personnel have reported electricity pricing. It is the difference between peak and off- more severe capacity loss, however, with perhaps as much as peak prices that determines the value of capacity. If, instead a million cubic meters having been filled by sediment. This of a peak price of $0.108 per kWh and an off-peak price higher estimate of capacity loss over the life of the reservoir 20. Annual sediment transport is about 35 million tons, so at a density of 1.5 tons per cubic meter, a little less than 22 million cubic meters would be conveyed. 69 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal yields the figure of one-quarter of one percent, which was assumptions, recall that the VSL is assumed to be US $ then used in this study for deriving the value of loading 34,565. A number of different empirical procedures have reductions in the watershed. been adopted for estimating the VSL, as well as a number of different procedures for transferring VSL estimates The benefits of capacity reduction ascribed to watershed from one country to another based on per capital income management interventions will be proportional to the or other factors (Narain and Sall 2016). Some recent work fraction of sediment transported that is assumed to settle in has inferred a considerably higher VSL for Nepal, based the reservoir. If the fraction were assumed to be one tenth, on workers’ compensation to migrate in pursuit of better rather than one quarter, of 1% the estimate of value would paying, albeit more dangerous work (M. Shrestha 2016). For be 60% lower. Conversely, if the fraction were assumed to be the purposes of illustration, then a VSL of twice $34,565: four tenths, rather than one quarter, of 1%, the estimate of $69,130 (which is still considerably lower than Shrestha’s value would be 60% higher. Again, however, much higher (2016) central estimate) is considered as a high-end estimate. estimates may become implausible, as they would imply that Conversely, while a much lower figure for the VSL itself the reservoir would soon be completely filled. might be unlikely, it has been assumed that the number of lives at risk is proportional to the number of structures To take the example of a US $1M portfolio of watershed at risk. While the constant of proportionality (one life at management interventions, it is estimated that about risk per four structures) is based on reported fatalities and $120,000 in hydroelectric-related benefits would arise for damages from over forty years of records (UNISDR 2015), KGA. Roughly $96,000 of such benefits would come from the correlation between the series is not perfect. One might, avoided damages and costs, and the remaining $24,000 from then ask, how our figures would differ if only half as many retention of reservoir capacity. lives were at risk, and this is then taken as a low-end estimate for calculating the value of avoided lives lost. 4.2.3. Value of reduction in landslide risk We derive three values for reducing landslide risk through In terms of the values of damages to structures, a range land management measures: avoided lives lost, avoided of figures for our estimate of rental costs could also be replacement cost of structures, and avoided road repairs. considered. For example, gross rental payments might reflect Considering the case of a US $1M budget for watershed payments both for the benefit of occupying a structure and management, approximately six lives would be expected to the annual cost of maintaining it from routine wear and tear. be saved per year. At a VSL of $34,565 per life saved, this If maintenance costs were assumed to be half of gross rental would translate into a benefit of US $196,000 per year or, at payments, the net rental value of structures at risk would be a discount rate of 10%, a net present value of US $1.96M. $2.70, rather than $5.39, per square meter. On the other Under this same scenario, the net present value of reduced hand, our approach assumes that the value of structures at risk of destruction of homes and other structures is estimated risk is proportional to their footprint alone. Whereas if there at about US $69,000 , and the net present value of expected are many structures built with multiple stories, then our cost savings on road repair would be about US $113,000. It estimate might be underestimating the rental value per unit should be noted that the values derived for avoided damages footprint. Therefore, we take $7.19 as an upper estimate of to assets are in reality the long-term increases in asset values the per-square-meter rental value, reflecting the case where associated with the modeled reductions in expected losses reported values were 75% of actual values. (not an estimate of actual damages averaged over a finite time period). Considering the costs of road repairs, again there are many different factors that could affect the cost of road repair, and The expected value of mortality risk reduction could vary it would not be possible to characterize the sensitivity of with a great many factors, including differences in the value results to all, or even a substantial number of them, without assigned to a statistical life and the number of people assumed extensive study. Inasmuch as the estimates reported come to be at risk from landslides. More generally, the probability from the midrange of three estimates, however, it might of a landslide occurring, and the change in that probability simply be noted that the lower end of the range would have as a result of interventions to stabilize slopes, divert runoff, yielded an $51,600 per kilometer of road damaged, the of other measures also depends on a number of uncertain upper end, $78,100. factors. A complete uncertainty/sensitivity analysis was infeasible due to time and resource constraints. However, to Summing the three categories of benefits quantified (using our give some sense of how results might vary under alternative mid-range estimates of value), US $1M spent on watershed Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  70 management is estimated to provide an expected net present estimates may not be exactly applicable to any particular value of benefits in excess of US $2.2M from reductions in context, there are reasons they might give either over- or expected losses from landslides alone. While a number of underestimates. The figure overestimates the share of on-site assumptions have gone into the derivation of this estimate, benefits to the extent that costs of program administration it should be noted that 1) the value of expected lives saved add to the establishment and maintenance costs recorded. generates most of the monetized value estimated, and the On the other hand, some of the market imperfections noted figure used for value of a statistical life, US $34,565, is lower in Section 3.3.6 might drive a wedge between the actual than some estimates in the literature; 2) the road damage value of practices to land users and the cost they would bear estimates have been confined to costs of repair only, and do to implement them; a land user might not be able to borrow not reflect either lost benefits during times when a damaged the funds required for an initial investment in terracing, road may be impassable or the possibility that a road cannot for example. Moreover, an actual watershed management be rebuilt, and would need instead to be relocated to what project would likely adopt practices for which land users would likely be a more circuitous route; and 3) due to data were willing to bear a greater fraction of the cost, other limitations, this study does not estimate potential damages to things being equal. Interventions that are not attractive to agricultural production from landslides, which might also be local land users would be less likely to be proposed. significant in some areas. 4.2.6. The costs of degradation 4.2.4. Value of carbon storage Another way to consider the benefits of watershed Carbon sequestration has the largest potential value from management is to look at the contrary case, where terraces rehabilitation of degraded lands, followed by terrace are abandoned as the population shifts from rural to urban improvement and soil and water conservation. Landslide areas and the land is allowed to degrade. Results indicate that mitigation has a much lower carbon benefit, due to the fact the potential increase in sediment load to the Kali Gandaki that benefits are not immediately and fully accrued, rather River under this scenario reaches 6.3M tons/year. This 20% they are scaled by the change in probability that the treated increase over the current rate is concentrated in the lower landslide will occur. portions of the watershed, and could have huge implications for sedimentation at KGA, as well as existing and planned The US $1M portfolio shows a total carbon benefit of only facilities planned upstream (Figure 4.7). For example, the $13,200 (using the social cost of carbon at $60 per CO2e, sediment load reaching the Modi Khola hydropower facility with a range of between $8,800 and $17,600). This value increases 0.42M tons/year (+34%) in this scenario, and is so low because up to budgets of around US $5M, the the load to the Lower Modi 1 facility increases by 0.85M prioritization of activities is driven by the high values of lives tons/year (+44%). The increased sediment load implies an and avoided infrastructure damage that come with landslide additional net present cost to operations and maintenance mitigation. In scenarios where carbon values are prioritized, at KGA of nearly US $13.5M. While we do not have the values can be much higher. sufficient data to calculate the net present cost of impacts on the upstream hydropower facilities, the large percentage 4.2.5. Local (on-site) benefits increase suggests that funding watershed management to at least maintain terraces and soils in good condition could Assuming that 84% of the cost of the watershed management be a smart investment. The impacts on instability of slopes program is shared by landholders, the on-site benefits of and corresponding landslide impacts could be even greater, a US $1M investment would total $840,000. If the low- although quantifying this impact through a mechanistic (30%) and high-end (100%) reported cost-shares are applied landslide model is outside the scope of the current study. instead, we would expect on-site benefits to range from US $300,000 to $1M in this example. It is worth noting that even if this rough calculation of local benefits is not included 4.3. PRIORITIZING WATERSHED in the total, the benefit: cost ratio still remains greater than one for portfolios up to US $5M. MANAGEMENT ACTIVITIES We have shown that watershed management can provide Our estimate of on-site benefits based on the average significant benefits to downstream hydropower and to local reported cost-sharing by landholders is, of course, an communities, and that the benefits are not evenly distributed imprecise estimate of the true benefits of implementing among different sectors. It is imperative, therefore, to these practices. Over and above the fact that cost-sharing understand where in the Kali Gandaki watershed these 71 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  dditional sediment export that could Figure - 4.7: A carbon), and regardless of the cost of implementation. result from abandonment of watershed Investing in watershed management in the darkest areas, management activities and existing soil therefore, will result in the greatest benefits in terms of conservation structures (e.g. terraces) in downstream water quality and hydropower impacts at KGA. cultivated areas in the lower watershed. The total increase in sediment reaches 4.3.2. Intervention Portfolios 6.3M tons/year in this scenario, an Watershed management activities can be prioritized based on increase of 20% different objectives, which will impact where investments should be focused. In the case of the Kali Gandaki watershed, there are multiple entities involved in promoting and implementing various types of best management practices, with different goals: the DoFSC invests in activities to control sediment and promote healthy functioning watersheds broadly, the Ministry of Agriculture and Livestock Development promotes best management practices to support productive and sustainable farming and grazing practices, and the NEA has a program to invest in sediment management in areas surrounding the KGA reservoir. The results described above are based on a set of optimal activity portfolios made to maximize the total monetized values, across budget levels ranging from US $500,000 to $50M (Table 4.2 and Figure 4.5). Using the ROOT tool, activity portfolios can also be developed to maximize multiple objectives. A set of 1000 scenarios was developed using the ROOT optimizer, setting an objective function to minimize sediment exported to the Kali Gandaki river, minimize local erosion, maximize avoided lives at risk from landslides, maximize the value of landslide risk mitigated for structures and roads, maximize carbon value, and to minimize cost. As noted above, local landholders often agree to assume a portion of the costs of implementation, with the expectation of local benefits in terms of maintaining soil health and productivity. However, if a program were to expect local landholders to bear part of the burden of the cost of watershed management, then it is necessary to ensure that local objectives activities should be prioritized, in order to deliver the greatest (such as maintaining or enhancing agricultural productivity) possible benefits for any given budget level. In the following are being given equal weight with downstream objectives sections, we further narrow our focus to a range of budgets (such as reducing sediment for hydropower operations). Figure more likely feasible for implementation in the near future. 4.9 and Figure 4.10 illustrate the potential trade-off between prioritizing activities for local versus downstream benefits. 4.3.1. Evaluating individual activities The portfolio maps in Figure 4.9 show that when downstream The following figures show the impacts of each individual sediment is the primary focus, reducing sediment through watershed management activity on reducing sediment. mitigating mass movement in landslides along the main stem These figures highlight the sub-watersheds where each type and tributary channels are frequently the preferred options. of activity has the highest potential to reduce sediment load However, when local erosion is the main concern, the focus to the Kali Gandaki River, regardless of its performance on shifts more toward terrace improvement, grazing land and other objectives (such as reducing local erosion or storing forest rehabilitation in the middle hills area. Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  72  odeled sediment reduction by sub-watershed, with full implementation of different management practices. Figure - 4.8: M Note the different scales on each panel A) Modeled sediment reduction from soil & water B) Modeled sediment reduction from hill terrace conservation practices. improvement practices. C) Modeled sediment reduction from degraded forest D) Modeled sediment reduction from degraded rangeland rehabilitation practices. rehabilitation practices. 73 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal E) Modeled sediment reduction from landslide mitigation F) Modeled sediment reduction from landslide mitigation practices – Type I and II. practices – Type III.  ntervention portfolios optimized for two competing objectives (left column: downstream sediment for Figure - 4.9: I hydropower and right column: local erosion reduction) and two budget levels (US $5M and $20M), for comparison. Note that different activities and sub-watersheds are chosen for implementation to meet the different objectives A) Downstream sediment-optimized portfolio, US $5M B) Local erosion-optimized portfolio, US $5M budget budget Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  74 C) Downstream sediment-optimized portfolio, US $20M D) Local erosion-optimized portfolio, US $20M budget budget Figure 4.10 demonstrates this trade-off in another way, soil loss. Each point on the curve represents a scenario of by showing the full set of 1000 optimal portfolios at the interventions. For example, the portfolio shown in Figure US $20M budget and the change in downstream sediment 4.9.C is the point of maximum downstream sediment local erosion achieved in each portfolio. Points to the right reduction on this curve, while the point of maximum on the curve prioritize downstream sediment reduction, reduction in local erosion corresponds to the map shown while points to the left on the curve prioritize reducing local in Figure 4.9.D. Trade-off curve showing performance of 1000 optimal scenarios (US $20M budget) in terms of their reduction Figure - 4.10:  in downstream sediment and local erosion control. Points to the bottom and right on the curve prioritize downstream sediment at the expense of reducing local erosion, while points on the upper left of the curve prioritize local erosion control Millions 4.5 4 D Local erosion reduction (tons/year) 3.5 3 2.5 2 1.5 1 0.5 0 C 0.4 0.6 0.8 1 1.2 1.4 Millions Downstream sediment reduction (tons/yr) 75 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal Finally, prioritizing specific activities and locations can be of this scenario in a given location. Higher agreement difficult when faced with so many different options that can (>75%) indicates that the activity is very cost-effective in benefit different objectives. In this case, an agreement map that sub-watershed, regardless of the relative importance is a useful tool to identify places and interventions that are given to specific objectives that might be considered by the repeatedly chosen in portfolios that optimize for different watershed management program. In this scenario, soil and objectives. Figure 4.11 shows an agreement map for the US water conservation, hill terrace improvement, grazing land $5M portfolio. The colors represent two different activities that rehabilitation, and landscape type III mitigation were not were consistently chosen in at least 750 of the 1000 iterations consistently selected across the iterations. Agreement map for US $5M scenario. The colors show the fraction of all scenarios in which a given Figure - 4.11:  activity was selected. No color indicates the sub-watersheds where each activity was selected in less than half of all iterations. Higher agreement (>75%) indicates that the activity is highly cost-effective in that sub-watershed, regardless of the specific objectives sought with the watershed management program Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  76 4.4. CHANGING CONDITIONS, are unclear. One factor that would clearly make the peaking capacity at KGA and other plants more valuable would be CHANGING VALUES increased reliance on renewable energy sources, such as It is often more analytically tractable to suppose that the solar and wind power on a national level. The intermittency conditions that now determine benefits and costs will of such sources puts a premium on the capacity to maintain continue unchanged in the future. It is also sometimes generation when renewables are not available. reasonable to suppose that present conditions are the best predictors of those that will prevail in the future. For these At the same time, increased dispatching of intermittent reasons, most of the analysis presented above has assumed renewable, e.g., wind and solar, might go hand in hand with that future conditions will continue as in the past.21 By the an increased regional linkage of power grids, which might same token, however, conditions assuredly will change in the reduce reliance on storage at any particular location and future, and so we discuss below how certain changes might improve access to the major storage capacity in India and affect the analysis. China. Similarly, Nepal has a major untapped hydropower potential and less than ten percent of the country’s Future changes can be relevant for many parts of our analysis, estimated technically feasible potential has been exploited. impacting processes of sediment generation, the valuation In addition to the two operational facilities on the Modi of hydropower and storage, and the link between watershed Khola tributary, there are three under construction (Lower management and livelihoods. Some of these changes might Modi Khola, Middle Modi, and Lower Modi 2, with a also be competing, in the sense that they lead to an opposite combined capacity of 45.6 MW) and at least three more effect on a given response variable. For example, sediment for which survey licenses have been issued (Department of yield might increase in the future, but technical advances in Electricity Development, GoN). On the one hand, building hydropower technology (e.g., turbine coatings) might reduce more projects upstream of KGA would increase the number the vulnerability, so that the final net-change in sediment- of beneficiaries of watershed management, as more plants related damages to hydropower remains relatively stable. At would be impacted by changing sediment loads. This implies the same time, many of the uncertain and non-stationary that the value of watershed management to the hydropower processes are highly non-linear, which means that a small sector would increase as generation capacity expands, as change in, e.g., glacial melt, results in a major change in many of these plants are run-of-river, which means that sediment generation. incoming sediments would be passed downstream through the turbines of several plants, potentially causing abrasive A thorough quantitative analysis of all these deeply damage to each in turn, or would need to be flushed from uncertain sources of future change was beyond the scope the desanding basin of one plant after another, thereby of this analysis, however, this section provides a short compounding the benefits of reduced sediment loads. On qualitative discussion of future scenarios for selected sectors the other hand, more plants will also increase redundancy, and hydrologic processes. thereby reducing the value of storage and the overall costs of plant shutdowns for maintenance at any given facility, in Hydropower: Reservoir capacity is valuable only to the terms of foregone energy generation during times of peak extent that flow in the river is insufficient at times to meet demand. Of course, if overall peak demand for electricity also demand. If climate change makes river flow even more increases, storage could become more valuable. The actual irregular than the monsoon cycle now implies, capacity benefits of watershed management for the energy sector as a might become more valuable, as would measures to preserve whole are thus difficult to determine, without more detailed it by reducing sediment delivery. The marginal value of study of the operations of all facilities, expansion plans, and power depends on where supply and demand balance. Both projections of energy demand. are likely to change greatly over time. The value of capacity, however, depends on the difference between the value placed Costs and on-site benefits: Nepal has, as have other on power during periods of high and low demand, so the countries in South Asia, experienced urbanization and effects of potentially uneven supply and demand growth substantial migration from rural areas. If this movement 21. The exceptions are: i) the social cost of carbon estimates, as adapted from (Stiglitz et al. 2017), assume real values risking at a rate of 2.25% per year; ii) we suppose carbon storage values increase over time as plants grow; and iii) we distinguish between the establishment costs of interventions which are assumed to be borne immediately, and their maintenance costs, which we assume begin in the year following establishment. 77 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal of population is matched by reduced agricultural activity lead to an increase in hillslope erosion that might be on steep and vulnerable terrain, then the opportunity cost targeted by increasing vegetation cover through watershed of restoring native vegetation will likely decline as it will management practices. displace less croplands. At the same time, outmigration might make the work-force scarce and hence increase Landslide losses: Some of the most substantial benefits costs of implementation, and on-site benefits of increased identified above arise from saving lives that might otherwise agricultural productivity may be less significant if fewer be lost in landslides. Landslides are only a risk in relatively farms and farmers avail themselves of them. Of course, steep terrain. These tend to be the areas from which food production would need to be made up elsewhere, Nepalis are now moving to cities or overseas. On the other either by expansion in the Terai region of Nepal, from hand, however, experience in other countries suggests that exports, or by intensification of production in existing areas. wealthy people may later be drawn to the more spectacular It may well be, however, that the challenge in the future is viewsheds of ravines and hillsides. Socio-economic growth not so much to make agricultural production sustainable in the watershed and remittances from emigrated family on areas of land devoted to growing crops as managing the members might also lead to a growing value of remaining reversion from farmland to native forests and grass without houses. Landslides are also associated with extreme events, excessive erosion. particularly earthquakes and rainstorms. An increase in extreme precipitation could increase the value of Sediment yield from different processes: Future interventions compared to our current analysis which is sediment yield from different processes is highly uncertain based on climate observations of the past two decades. due to possibly compounding effects of future land use changes in the watershed and global climatic changes. Take, Conclusion for example, the aforementioned uncertainty in future land In general, it should be noted that changing conditions use, which will have a major impact on sediment delivery. – both in terms of economic development and climate For example, Rodrigo-Comino et al. (2018) found that impacts – might greatly change the future value of sediment erosion from abandoned terraces in the Mediterranean management both with regard to sediment generation but can be very high, but that the increase of erosion depends also with regard to the value of ecosystem services. For strongly on the crop types and land management practices example, higher standards of living might greatly increase pre-abandonment. Similarly, many of the makeshift roads the value of structures of risk and hence the value of that are now being constructed to small villages might avoiding destruction of structures by landslides. At the same be abandoned in the future and might, if they are not time, more wealth might also decrease the dependence of the decommissioned properly, continue to yield large amounts population on ecosystem services, e.g., on-site fuel, fodder, of sediment for decades. In turn, increasing affluence in and food production to ensure their livelihoods. remaining population centers might enable investing in paved roads and better road construction practices which A thorough analysis of such future scenarios or the uncertainty might reduce erosion from the roads that remain in service. and non-stationarity in natural processes and scenarios of socio- economic development was beyond the scope of this study. In terms of natural processes, increasing rates of glacial However, it should be noted that there are proven techniques melt that are expected in the Himalayas might exponentially for participatory development of relevant scenarios with local increase the delivery of fine and abrasive sediment, which stakeholders, as well as numerical methods for analyzing could not be mitigated with common watershed management coupled human-natural systems under deep uncertainty, which strategies. Conversely, slightly wetter climate in the Mustang might be very beneficial to narrow down future value ranges plateau, where soils are mostly bare and erodible, could for watershed management in the Himalayas. Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  78 5. CONCLUSIONS & RECOMMENDATIONS © Paul Prescott/Shutterstock.com 5.1. CONCLUSIONS Conservative assumptions were applied throughout the economic analysis, and even so the results show that the This study presents a novel attempt to generate a aggregated benefits of such a program can greatly outweigh comprehensive valuation of the multiple benefits that the costs. The benefits to cost ratio is highest at the can result from implementing a watershed management lower investment levels and decreases to 1.2 at $50M US program to control erosion and sedimentation in the investment. There is both a physical limit and a feasibility limit Kali Gandaki watershed. A physically-based modeling to how much can be achieved with watershed management approach, in combination with micro-economic modeling alone – our results indicate a maximum of 20.5% reduction of major benefit streams, was employed using watershed- in fine sediment load using the types of practices evaluated and region-specific data to evaluate these benefits rigorously. in this study. But as part of a comprehensive sediment In this way, our study goes beyond the often-used approach strategy that includes land management improvements, of simply transferring area-based estimates of the value structural sediment mitigation approaches, reclamation of of watershed benefits from one region to another and degraded lands, and best practices for road engineering, represents a proof-of-concept for how such approaches our results show that a data-driven and targeted program of may be applied in other contexts. watershed management can contribute greatly to a broader 79 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal social benefit through real and significant economic gains conservation and development goals as well as the need for to society. sustainable energy and rural development. The results in Table 4.2 highlight the importance of As with any study that relies on physically-based models and considering multiple benefit streams and sources of value extrapolates landscape-scale effects from local data, there to make the case that investments in watershed services are are uncertainties inherent in the analysis. Every attempt has sound. With the exception of the benefits from landslide been made to use the best available data, vetted through a mitigation , no one sector receives enough benefits to justify stakeholder engagement process. Errors in the underlying 100% of the investment cost. The fact that values accrue data on topography, historical climate, streamflow and to many different sectors also means there is flexibility for sediment concentrations, and uncertainties about the costs building coalitions of different actors and funding sources and characteristics of watershed management practices to underwrite such programs. Consider, for example, the as implemented in specific and varying locations on the case of a US $1M investment portfolio, where the value ground means that the results of this study should be taken of avoided lives lost is the largest single contributor to the as demonstrative, rather than definitive. However, this total benefits from the program (estimated at $2M). Even study overall is conservative in its assumptions and thus if those benefits are ignored, the benefit: cost ratio is still provides evidence that watershed management can have enough to justify making the investment (changing from positive economic benefits that greatly exceed the costs of 3.2 to 1.3). Further, this portfolio assumes a net benefit to its implementation. landholders (from improved soil fertility, water capture, and agricultural productivity, for example) of US $840,000. It is worth noting that the benefits accruing to landholders Considering this scenario from a break-even perspective, the are a large fraction of the total benefits shown in the results local benefits could be 30% lower (at only $580,000) and of this analysis. The assertion that better land-management the overall benefit: cost ratio would still reach 1. On the practices might provide such benefits to the landholders other hand, if the values of landslide risk reduction are fully implementing them may beg the question of why they are realized and funded through disaster risk reduction efforts, not already adopting them. There are several reasons they then local landholders could receive the benefit of watershed may not be. The first may just be that the on-site benefits management practices without any cost-share required of adoption do not fully cover the private costs. Economists on their part. Strategic partnerships between sectors are have modeled farmers’ soil conservation choices as a therefore necessary to pool resources and achieve these problem in the management of a depletable resource (a widespread benefits. seminal paper is (McConnell 1983). They describe farmers as balancing the benefits of enhanced soil fertility against the However, in some cases, targeting investments to benefit costs of measures to maintain or restore fertility. If a farmer one sector will reduce the benefits accrued to other sectors. has struck this balance between on-site benefits and costs, For example, sediment generation affecting hydropower the benefits of doing a little more to prevent erosion and infrastructure may be a huge problem in an area with loss should roughly approximate the costs. Other choices relatively few people, which would argue for an engineering farmers make may be more discrete: whether to establish approach to sediment management, such as building terraces or hedgerows at all, for example. Inasmuch as many retaining walls or sediment-trapping structures rather areas have features such as terraces and hedgerows in place than investing in vegetation-based interventions that bring already, however (Bhattarai 2018), it is reasonable to suppose fuelwood and fodder benefits. Conversely, investing in on- that the costs of expanding their use in other areas would be farm management practices in another area may deliver offset by substantial benefits. huge development gains, but may not address the most critical sources of sediment for the hydropower sector. In addition, several market failures may explain why farmers Mapping and quantifying the sources of sediment and do not adopt on their own practices that might confer net benefit pathways will help policymakers to design equitable benefits. Farmers in developing countries often face credit programs that distribute the costs of sediment management constraints that prevent them from making profitable capital across different actors who receive benefits, and that address investments (Das and Bauer 2012; Blackman 2001). There 22. The total benefits from reducing landslide risks (value of avoided lives lost, avoided loss of structures and avoided road repairs) is greater than the cost of implementation only up to a budget of about US $5M. Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  80 may also be a lack of information. Many projects have been robust payments for ecosystem services (PES) schemes and instituted to demonstrate to farmers the benefits of sustainable can leverage investment from multiple actors. Application land management without necessarily susbsidizing their of the sediment modeling and prioritization tools can adoption. The fact that such projects often are adopted inform the design of watershed management programs to by many local land users without subsidies suggests that reduce sediment and improve water quality, by targeting the land users did not have information regarding their interventions to the best places to achieve particular effectiveness before the program was initiated (see Section outcomes and balancing trade-offs, thereby making such 3.3.6 for several examples). programs more cost effective and transparent. Finally, many successful interventions to encourage more The transportation and disaster risk management sustainable land management practices have focused as sectors can apply the landscape-scale hazard mapping much or more on the institutions for management as developed in this study to estimate the exposure of assets on technologies or practices instituted per se. Seminal such as roads, at a finer spatial resolution than is currently contributions such as (Hardin 1968; Ostrom 1990) available from landscape-scale screening analyses. While document how lack of an effective governance structure can cutting of hill slopes, slope stabilization, landslide risks, water result in the overexploitation of resources with consequent impacts, and other parameters are currently considered degradation of the asset base that provides them. Instances in a typical impact assessment study, the relatively simple of such degradation have often been noted, along with sediment model employed here could be applied to assess the descriptions of the societal attitudes that underlie them. potential for downstream impacts outside the project area. Ahmad (2001), for example, writes of local people who Further, the prioritization tools can be used to identify areas were concerned that the lands they managed would be of particular risk that may require higher standard of impact appropriated by the government if local actions improved assessment and/or consideration of cumulative (rather than their condition. Baig et al. (2013) write of grazing lands project-specific) impacts on ecosystem services. producing at less than a third of their potential because nomadic users frustrated local attempts to institute rotational The hydropower sector can use the valuation and grazing systems. Another study found that the productivity prioritization methodologies to design PES schemes that of grazing lands increased by a factor of ten when local more effectively control sediment from watersheds. Where people were organized to better manage grazing access policy mechanisms exist that require revenue-sharing from (WOCAT 2012). While these studies were conducted in hydropower plants, the sediment budget and prioritization other countries, similar concerns have been cited in grazing tools can be used to identify priority areas for investment that land management in Nepal (Guedel, n.d.). promote rural development (satisfying the motivation for why such policies often exist), while simultaneously reducing There are, then, many benefits that landowners might enjoy sediment-induced impacts on operations and maintenance as a result of interventions to manage watershed to prevent of facilities. erosion, as well as reasons to suppose that landowners may not always act on incentives to supply these benefits The tools also have relevance for environmental and to themselves. Aligning the incentives for landholders with social safeguards, by providing a data-driven and broader societal goals for improving the value of ecosystem systematic way to incorporate ecosystem services impacts into services from watersheds is therefore a policy challenge, environment management plans, ensuring that infrastructure and one that can be informed by the types of information projects are more resilient in the context of other forces and provided by this study: e.g., where watershed management pressures on the landscape. Beyond identifying impacts of practices provide greatest overall economic benefits and how proposed projects, the prioritization tools can also help to these benefits accrue to different sectors. Such a systematic identify mitigation opportunities to offset project impacts to approach allows for further engagement with different ecosystem services. sectors to align interests and leverage resources. Overall, the methods and data resulting from this study The agriculture, forestry, and water sectors can demonstrate why effective and efficient targeting is key to use this valuation methodology to make a case for why achieving the greatest benefits at the lowest costs. Across watershed management programs are good investments. all of these sectors, the use of watershed-scale tools to Understanding and quantifying the benefits that accrue to evaluate and integrate the multiple benefits of watershed different sectors enables the design of more efficient and management into sectoral and cross-sectoral policy and 81 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal planning can be used a strategic tool to build resilience as The sampling campaign did not include bed-load climate change impacts are increasingly felt. Further, the measurements or detailed assessments of grainsize stakeholder-driven process employed here allows for more composition of the bed material. Bed load measurements durable and sustainable solutions, and the science-based, would be very resource intensive and should not be landscape-level assessment uncovers the underlying drivers prioritized. However, assessments of bed material of problems instead of focusing only the individual, localized composition throughout the watershed could yield very results of such problems. important information for modeling bed-load transport (Schmitt et al. 2018; Ferguson et al. 2015) with relatively small effort. 5.2. CAPACITY, DATA, AND • Sediment yield from glaciers: Glaciers are TECHNICAL NEEDS often considered of great importance for sediment management of Himalayan hydropower plants. That This study, like many assessments of hydrologic ecosystem is because glaciers are often perceived as an important services, draws heavily on numerical watershed-scale contributor in terms of total sediment, and especially models. This is because it is in general not possible to in terms of fine and abrasive particles that can pass observe processes generating sediment or storing carbon through the desanders and directly damage turbines. on whole watershed scales. Numerical models also allow Glacial sediment yield might also greatly increase in a us to evaluate the effectiveness of different management warmer climate. Despite the relatively high rates of scenarios applied at scale to achieve objectives of glaciation in the Myagdi and Modi Rivers, the relative watershed management. contribution of these tributaries to the overall sediment budget is relatively small, however our estimates of how Data are crucial to calibrate these numerical models, as much of that fine sediment originates from glaciers is demonstrated in the previous sections of this report. To purely model based. Some measurements of sediment bring the results of studies based on numerical models to the yield directly downstream of glaciers would be of great field, local agencies will require the knowledge to critically value to determine the contribution of glaciers and the evaluate model input data, run models, and compare model mineral composition of the mobilized material (similar outputs to their experience and use the results for prioritizing to Haritashya et al. 2006). their work in watershed management. • Sediment yield from roads Sediment yield from roads is not monitored in the This section presents some brief evaluation of the greatest Kaligandaki area and indeed there are very few studies data gaps in the study region, and some measures to address on that in Nepal. Establishing a network of sediment data gaps and capacity improvements. traps for measuring sediment load from roads could be of great importance for better management and planning Priorities for data collection of road sediment generation. • Suspended sediment data: River suspended sediment • Sediment composition and provenance: While data collected by Kathmandu University were of high initial results of KU indicate that glaciers do not supply quality and the sampling frequency matched the needs of sediment of greater hardness than other erosion processes, this study. Importantly, most of the relevant tributaries, more distributed samples of mineral composition, geo- except for the Aadhi river, were monitored. Sediment chemistry, and isotopes could be a very cost effective monitoring took place from the river banks and using measure to confirm the origins of different minerals, and multiple grab samples. Depth integrated sampling along to provide an independent line of evidence for sediment an entire cross-section would likely yield more accurate provenance in the watershed (Garzanti et al. 2016). results, but would also require expensive equipment such • Exposure data: Data of exposed infrastructure and as a cable crane or a suspension bridge (wading is not buildings is based on Open Street Map data, i.e., data feasible in most of the rivers). For the purpose of ongoing mapped by interested citizens. Our visual quality watershed assessment and management, it is advised to control indicated that these data represent the location sample with a simple but replicable method at many of most structures and roads well. However, there is no locations and for at minimum 5 years. Given the usefulness guarantee that these data are comprehensive. Incomplete of sediment data for watershed management, it should exposure data can lead to underestimating the risk of be explored what options are available for automated natural hazards as well as to underestimating the value sediment sampling (e.g., using turbidity meters). of measures reducing hazards. Given the rapid and Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  82 unplanned development in the watershed, satellite data sediment impacts on upstream hydropower plants, both and machine learning approaches could be used for those currently operational (Modi Khola and Lower infrastructure and building identification. Modi 1) and planned (Middle Modi, Lower Modi • Landslides: Data on landslides in a georeferenced Khola, and Lower Modi 2, among others). format are absent. Such data would be of great value ii. improving the data basis for on-site benefits versus local to better calibrate and validate the landslide model. and total program costs, through improved monitoring Landslide data could be obtained either by trained of site-level impacts and data sharing among programs; citizen scientists (e.g. people who already work on iii. extending the landslide hazard analysis by considering mapping for open street maps) or from satellite data co-seismic landslides and improving data on assets and machine learning approaches. Similarly, better at risk; understanding road and structure damages by landslides iv. exploring partnerships with WOCAT, FAO and and the related repair costs could help to better value other experts in the field of sediment and watershed mitigation measures. management, to bring in other perspectives and link to • Hydromet data: Hydrometeorological data were other knowledge bases on watershed management; mostly of good quality and there is a dense network of v. improving the understanding and technical precipitation stations in the study area. Discharge data approaches for considering episodic phenomena (e.g. from Jomsom gauging stations turned out to be invalid for landslides, floods); most of the year because of changing river cross sections. vi. explore other modeling approaches for watershed If such inconsistencies are observed, mechanisms should hydrology and sedimentation (e.g. SWAT), to compare be in place to correct cross section data as soon as possible, the results of different approaches and thereby develop as data gaps can greatly reduce the data quality. ways to reduce computational complexity (e.g. using • Data on activity implementation: Models that simple coefficients based on calibrated models to quickly predict impacts of watershed management activities rely evaluate particular interventions impacts both locally on data to parameterize the models so that they reflect and downstream); how effective the practices are at restoring vegetation and vii. extending the analysis to model and compare the reducing sediment. More site-specific studies from the impacts of civil works alone, vegetative interventions study region would help to reduce uncertainty in model alone, and then a more integrated approach to improve estimates. Further, specifics on average costs of activities understanding of the relative values and synergies of and the physical & economic conditions that affect costs grey and green infrastructure approaches; would improve estimates of program efficiency. Finally, viii. building a global knowledge coalition (facilitated, local information on the willingness to pay on the part for example, by The World Bank), to collect and of landholders to adopt practices, and the co-benefits make accessible a global database of watershed that they expect to receive from them, would improve intervention costs; greatly our estimates of the total value of implementing ix. develop a capacity-building program including, for such practices. example, virtual trainings related to sediment modeling and management to build up interest and capacity, showcase local progress, and improve access to global 5.3. FUTURE WORK good practice and expertise. This can include existing Further work that could be undertaken to expand on this communities of practices and can slowly expand to analysis could include: other elements of watershed planning, monitoring, management, and instruments (e.g. PES) that can all i. extending the analysis of benefits to the hydropower be better designed on the bedrock of good watershed sector by incorporating detailed economic analysis of management data and analytics. 83 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal 6. REFERENCES (ADB) Asian Development Bank. 2012. “Nepal: Kali Gandaki ‘A’ Hydroelectric Project. Performance Evaluation Report.” https://www.adb.org/sites/default/files/evaluation-document/35901/files/in13-13.pdf. (ICIMOD) International Centre for Integrated Mountain Development. 2007. “Good Practices in Watershed Management: Lessons Learned in the Mid Hills of Nepal.” Kathmandu, Nepal. (MoFSC) Ministry of Forests and Soil Conservation. 2016. “National Forest Reference Level of Nepal (2000 – 2010).” Kathmandu, Nepal. https://redd.unfccc.int/files/nepal_frl_jan_8__2017.pdf. (NEA) Nepal Electricity Authority. 2013. “Kali Gandaki ‘A’ Hydropower Plant Rehabilitation Project: Environmental Assessment.” Bhaktapur, Nepal. Ahmad, Shahid. 2001. “Watershed Management in Pakistan: Achievements and Issues.” In National Seminar on Water Resources, 1–17. Annandale, George W., Gregory L. Morris, and Pravin Karki. 2016. Extending the Life of Reservoirs: Sustainable Sediment Management for Dams and Run-of-River Hydropower. Washington, DC: World Bank. https://doi.org/10.1596/978-1-4648-0838-8. Aryal, Suman, Tek Narayan Maraseni, and Geoff Cockfield. 2014. “Sustainability of Transhumance Grazing Systems under Socio-Economic Threats in Langtang, Nepal.” Journal of Mountain Science 11 (4): 1023–34. https://doi.org/10.1007/ s11629-013-2684-7. Atreya, Kishor, Subodh Sharma, Roshan M. Bajracharya, and Neeranjan P. Rajbhandari. 2008. “Developing a Sustainable Agro-System for Central Nepal Using Reduced Tillage and Straw Mulching.” Journal of Environmental Management. https://doi.org/10.1016/j.jenvman.2007.03.017. Baig, Mirza B., Shabbir A. Shahid, and Gary S. Straquadine. 2013. “Making Rainfed Agriculture Sustainable through Environmental Friendly Technologies in Pakistan: A Review.” International Soil and Water Conservation Research. https://doi.org/10.1016/S2095-6339(15)30038-1. Beatty, C.R., L. Raes, A.L. Vogl, P. Hawthorne, M. Moraes, J.L. Saborio, and K. Meza Prado. 2018. “Landscapes at Your Service: Applications of the Restoration Opportunities Optimization Tool (ROOT).” Gland, Switzerland. https://doi.org/10.2305/IUCN.CH.2018.17.en. Benavidez, Rubianca, Bethanna Jackson, Deborah Maxwell, and Kevin Norton. 2018. “A Review of the (Revised) Universal Soil Loss Equation ((R)USLE): With a View to Increasing Its Global Applicability and Improving Soil Loss Estimates.” Hydrology and Earth System Sciences 22 (11): 6059–86. https://doi.org/https://doi.org/10.5194/hess-22-6059-2018. Bhattarai, Ramanand. 2018. “Traditional Irrigated Rice Terraces [Nepal].” 2018. Bishwakarma, Meg B. 2012. “Performance Improvement of Headworks: A Case of Kalignadaki A Hydropweor Project through Physical Hydraulic Modeling.” Blackman, Allen. 2001. “Why Don’t Lenders Finance High-Return Technological Change in Developing-Country Agriculture?” American Journal of Agricultural Economics. https://doi.org/10.1111/0002-9092.00227. Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  84 Bonzanigo, Laura, and Nidhi Kalra. 2014. “Making Informed Investment Decisions in an Uncertain World: A Short Demonstration,” no. February: 20. Brauman, Kate A, Gretchen C Daily, T Ka’eo Duarte, and Harold A Mooney. 2007. “The Nature and Value of Ecosystem Services: An Overview Highlighting Hydrologic Services.” Annual Review of Environment and Resources 32 (1): 67–98. https://doi.org/doi:10.1146/annurev.energy.32.031306.102758. Cameron, Trudy Ann. 2010. “Euthanizing the Value of a Statistical Life.” Review of Environmental Economics and Policy. https:// doi.org/10.1093/reep/req010. Cardinael, Rémi, Viviane Umulisa, Anass Toudert, Alain Olivier, Louis Bockel, and Martial Bernoux. 2018. “Revisiting IPCC Tier 1 Coefficients for Soil Organic and Biomass Carbon Storage in Agroforestry Systems.” Environmental Research Letters 13 (12): 124020. https://doi.org/10.1088/1748-9326/aaeb5f. Cavalli, Marco, Sebastiano Trevisani, Francesco Comiti, and Lorenzo Marchi. 2013. “Geomorphometric Assessment of Spatial Sediment Connectivity in Small Alpine Catchments.” Geomorphology, Sediment sources, source-to-sink fluxes and sedimentary budgets, 188: 31–41. https://doi.org/10.1016/j.geomorph.2012.05.007. CBS. 2018. “Annual Household Survey 2016/17.” Vol. 17. Chalise, S R, and N R Khanal. 1997. “Erosion Processes and Their Implications in Sustainable Management of Watersheds in Nepal.” FRIEND’97: Regional Hydrology: Concepts and Models for Sustainable Water Resource Management, no. 246: 325. Chhetry, Balendra, and Kumar Rana. 2015. “Effect of Sand Erosion on Turbine Components: A Case Study of Kali Gandaki ‘A’ Hydroelectric Project (144 MW), Nepal.” Hydro Nepal: Journal of Water, Energy and Environment 17: 24–33. https://doi.org/10.3126/hn.v17i0.13270. Collison, A J C, M G Anderson, and D M Lloyd. 1995. “Impact of Vegetation on Slope Stability in a Humid Tropical Environment: A Modeling Approach.” Proceedings of the ICE-Water Maritime and Energy 112 (2): 168–75. http://www.icevirtuallibrary.com/content/article/10.1680/iwtme.1995.27662. Dahal, Bhim Kumar, and Ranjan Kumar Dahal. 2017. “Landslide Hazard Map: Tool for Optimization of Low-Cost Mitigation.” Geoenvironmental Disasters 4 (1): 8. https://doi.org/10.1186/s40677-017-0071-3. Dahal, Nagmindra, and Roshan M Bajracharya. 2013. “Effects of Sustainable Soil Management Practices on Distribution of Soil Organic Carbon in Upland Agricultural Soils of Mid-Hills of Nepal.” Nepal Journal of Science and Technology 13 (1): 133–41. https://doi.org/10.3126/njst.v13i1.7452. Dahal, R.K. RK, S. Hasegawa, N. P. Bhandary, and R. Yatabe. 2010. “Low Cost Road for the Development of Nepal and Its Engineering Geological Consequences.” In Geologically Active: Proceedings of the 11th IAEG Congress. Das, Romy, and Siegfried Bauer. 2012. “Bio-Economic Analysis of Soil Conservation Technologies in the Mid-Hill Region of Nepal.” Soil and Tillage Research. https://doi.org/10.1016/j.still.2012.01.016. Dennedy-Frank, P James. 2018. “Effects of Land-Cover Change on Streamflow: Analysis of Watershed Simulations from around the World.” Stanford University. https://searchworks.stanford.edu/view/12649139. Devkota, S, K Sudmeier-Rieux, I Penna, S Erble, M Jaboyedoff, A Andhikari, and R Khanal. 2014. “Community-Based Bio-Engineering for Eco-Safe Roadsides in Nepal.” Downs, Peter W, Scott R Dusterhoff, Glen T Leverich, Philip J Soar, and Michael B Napolitano. 2018. “Fluvial System Dynamics Derived from Distributed Sediment Budgets: Perspectives from an Uncertainty-Bounded Application.” Earth Surface Processes and Landforms 43 (6): 1335–54. https://doi.org/10.1002/esp.4319. 85 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal Editors, Journal of Economic Perspectives (JEP). 2015. “Editorial Note: Correction to Richard S. Tol’s ‘The Economic Effects of Climate Change.’” Journal of Economic Perspectives 29 (1): 217–20. (US EPA), United States Environmental Protection Agency. 2017. “The Social Cost of Carbon Estimating the Benefits of Reducing Greenhouse Gas Emissions.” 2017. Felix, D, I Albayrak, A Abgottspon, and R M Boes. 2016. “Hydro-Abrasive Erosion of Hydraulic Turbines Caused by Sediment - a Century of Research and Development.” IOP Conference Series: Earth and Environmental Science 49 (12): 122001. https://doi.org/10.1088/1755-1315/49/12/122001. Ferguson, R I, M Church, C D Rennie, and J G Venditti. 2015. “Reconstructing a Sediment Pulse: Modeling the Effect of Placer Mining on Fraser River, Canada.” Journal of Geophysical Research: Earth Surface 120 (7): 1436–54. https://doi.org/10.1002/2015JF003491. Filoso, Solange, Maíra Ometto Bezerra, Katherine C B Weiss, and Margaret A Palmer. 2017. “Impacts of Forest Restoration on Water Yield: A Systematic Review.” PLOS ONE 12 (8): e0183210. https://doi.org/10.1371/journal.pone.0183210. Førsund, FInn. 2015. Hydropower Economics. Springer. Fort, M, E Cossart, and G Arnaud-Fassetta. 2010. “Hillslope-Channel Coupling in the Nepal Himalayas and Threat to Man- Made Structures: The Middle Kali Gandaki Valley.” Geomorphology, Recent advances in landslide investigation, 124 (3): 178–99. https://doi.org/10.1016/j.geomorph.2010.09.010. Fryirs, Kirstie A, Gary J Brierley, Nicholas J Preston, and Mio Kasai. 2007. “Buffers, Barriers and Blankets: The (Dis) Connectivity of Catchment-Scale Sediment Cascades.” CATENA 70 (1): 49–67. https://doi.org/10.1016/j. catena.2006.07.007. Gabet, Emmanuel J, Douglas W Burbank, Jaakko K Putkonen, Beth A Pratt-Sitaula, and Tank Ojha. 2004. “Rainfall Thresholds for Landsliding in the Himalayas of Nepal.” Geomorphology 63 (3): 131–43. https://doi.org/10.1016/j. geomorph.2004.03.011. Garzanti, Eduardo, Jiang-Gang Wang, Giovanni Vezzoli, and Mara Limonta. 2016. “Tracing Provenance and Sediment Fluxes in the Irrawaddy River Basin (Myanmar).” Chemical Geology 440: 73–90. Gourevitch, Jesse D, Peter L Hawthorne, Bonnie L Keeler, Craig R Beatty, Michelle Greve, and Michael A Verdone. 2016. “Optimizing Investments in National-Scale Forest Landscape Restoration in Uganda to Maximize Multiple Benefits.” Environmental Research Letters 11 (11): 114027. https://doi.org/10.1088/1748-9326/11/11/114027. Government of Nepal. 2013. “Nepal - National Population and Housing Census 2011.” Kathmandu. http://old.cbs.gov.np/ nada/index.php/catalog/54. ———. 2016. “NEPAL EARTHQUAKE 2015: Post Disaster Recovery Framework 2016 - 2020,” 3–60. Guedel, Nicole. n.d. “Rehabilitation of Communal Grazing Land (Nepal).” Hallet, B, L Hunter, and J Bogen. 1996. “Rates of Erosion and Sediment Evacuation by Glaciers: A Review of Field Data and Their Implications.” Global and Planetary Change, Impact of Glaciations on Basin Evolution: Data and Models from the Norwegian Margin and Adjacent Areas, 12 (1): 213–35. https://doi.org/10.1016/0921-8181(95)00021-6. Hamel, Perrine, Rebecca Chaplin-Kramer, Sarah Sim, and Carina Mueller. 2015. “A New Approach to Modeling the Sediment Retention Service (InVEST 3.0): Case Study of the Cape Fear Catchment, North Carolina, USA.” Science of The Total Environment 524–525: 166–77. https://doi.org/10.1016/j.scitotenv.2015.04.027. Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  86 Hansen, M C, P V Potapov, R Moore, M Hancher, S A Turubanova, A Tyukavina, D Thau, et al. 2013. “High-Resolution Global Maps of 21st-Century Forest Cover Change.” Science 342 (6160): 850 LP – 853. http://science.sciencemag.org/ content/342/6160/850.abstract. Hardin, Garrett. 1968. “The Tragedy of Commons.” Science. https://doi.org/10.1126/science.162.3859.1243. Haritashya, Umesh K, Pratap Singh, Naresh Kumar, and R P Gupta. 2006. “Suspended Sediment from the Gangotri Glacier: Quantification, Variability and Associations with Discharge and Air Temperature.” Journal of Hydrology 321 (1): 116–30. https://doi.org/10.1016/j.jhydrol.2005.07.037. Hutchinson, N N. 1977. “Assessment of the Effectiveness of Corrective Measures in Relation to Geological Conditions and Types of Slope Movement.” Bulletin of the International Association of Engineering Geology - Bulletin de l’Association Internationale de Géologie de l’Ingénieur 16 (1): 131–55. https://doi.org/10.1007/BF02591469. IHA. n.d. “Nepal - Kali Gandaki Case Study.” ———. 2017. “2017 Hydropower Status Report.” https://www.hydropower.org/2017-hydropower-status-report. Jaquet, Stéphanie, Thomas Kohler, and Gudrun Schwilch. 2019. “Labour Migration in the Middle Hills of Nepal: Consequences on Land Management Strategies.” Sustainability. https://doi.org/10.3390/su11051349. Kafle, K R, and R Bhandari. 2019. “Sediment Monitoring Along The Kaligandaki River.” Kargel, J S, G J Leonard, D H Shugar, U K Haritashya, A Bevington, E J Fielding, K Fujita, et al. 2016. “Geomorphic and Geologic Controls of Geohazards Induced by Nepal’s 2015 Gorkha Earthquake.” Science 351 (6269): aac8353. https://doi.org/10.1126/science.aac8353. Karki, Nava Raj, Arbind Kumar Mishra, and Jayandra Shrestha. 2010. “Industrial Customer Outage Cost Analysis: A Case Study of Nepal.” International Journal of Systems Assurance Engineering and Management. https://doi.org/10.1007/s13198- 010-0011-z. Larsen, Isaac J, David R Montgomery, and Oliver Korup. 2010. “Landslide Erosion Controlled by Hillslope Material.” Nature Geoscience 3 (4): 247–51. https://doi.org/10.1038/ngeo776. Mandle, Lisa, Benjamin P Bryant, Mary Ruckelshaus, Davide Geneletti, Joseph M Kiesecker, and Alexander Pfaff. 2016. “Entry Points for Considering Ecosystem Services within Infrastructure Planning: How to Integrate Conservation with Development in Order to Aid Them Both.” Conservation Letters 9 (3): 221–27. https://doi.org/10.1111/conl.12201. McConnell, K. E. 1983. “An Economic Model of Soil Conservation.” American Journal of Agricultural Economics 65 (1): 81–89. Morris, Gregory. 2014. “Sustainable Sediment Management: Kali Gandaki 144 MW Hydropower Dam, Nepal.” Narain, Urvashi, and Christopher Sall. 2016. “Methodology for Valuing the Health Impacts of Air Pollution: Discussion of Challenges and Proposed Solutions.” World Bank, 1–69. https://doi.org/10.1596/K8849. NEA. 2018. “Annual Report.” Ortigao, J. A. R., and Alberto Sayao. 2004. Handbook of Slope Stabilisation. 1st ed. Springer-Verlag Berlin Heidelberg. https:// doi.org/10.1007/978-3-662-07680-4. Ostrom, Elinor. 1990. “GOVERNING THE.” In Governing the Commons - The Evolutino of Institutinos for Collective Action. 87 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal Pandit, Bishnu Hari, Krishna K Shrestha, and Suman S Bhattarai. 2014. “Sustainable Local Livelihoods through Enhancing Agroforestry Systems in Nepal.” Journal for Forest and Livelihoods. Parsons, Andrew J, Richard D Law, Michael P Searle, Richard J Phillips, and Geoffrey E Lloyd. 2016. “Geology of the Dhaulagiri-Annapurna-Manaslu Himalaya, Western Region, Nepal. 1:200,000.” Journal of Maps 12 (1): 100–110. https://doi.org/10.1080/17445647.2014.984784. Paudel, Deepa, Krishna Raj Tiwari, Roshan Man Bajracharya, Nani Raut, and Bishal K. Sitaula. 2017. “Agroforestry System: An Opportunity for Carbon Sequestration and Climate Change Adaptation in the Mid-Hills of Nepal.” Octa Journal of Environmental Research 5 (1): 022–031. Piégay, Hervé. 2016. “System Approaches in Fluvial Geomorphology.” In Tools in Fluvial Geomorphology, 75–102. https://doi.org/10.1002/9781118648551.ch5. Postel, Sandra L, and Barton H Thompson. 2005. “Watershed Protection: Capturing the Benefits of Nature’s Water Supply Services.” Natural Resources Forum 29 (2): 98–108. https://doi.org/10.1111/j.1477-8947.2005.00119.x. Ramos-Scharrón, Carlos E, and Lee H MacDonald. 2007. “Development and Application of a GIS-Based Sediment Budget Model.” Journal of Environmental Management 84 (2): 157–72. https://doi.org/10.1016/j.jenvman.2006.05.019. Ramos-Scharrón, Carlos E, and Lee H MacDonald. 2005. “Measurement and Prediction of Sediment Production from Unpaved Roads, St John, US Virgin Islands.” Earth Surface Processes and Landforms 30 (10): 1283–1304. https://doi. org/10.1002/esp.1201. Reid, Leslie M, and Thomas Dunne. 2016. “Sediment Budgets as an Organizing Framework in Fluvial Geomorphology.” In Tools in Fluvial Geomorphology, 357–80. https://doi.org/10.1002/9781118648551.ch16. Rickenmann, Dieter. 2005. “Runout Prediction Methods BT - Debris-Flow Hazards and Related Phenomena.” In Debris- Flow Hazards and Related Phenomena, 305–324. Rodrigo-Comino, Jesús, Carlos Martínez-Hernández, Thomas Iserloh, and Artemi Cerdà. 2018. “Contrasted Impact of Land Abandonment on Soil Erosion in Mediterranean Agriculture Fields.” Pedosphere 28 (4): 617–31. https://doi.org/https://doi.org/10.1016/S1002-0160(17)60441-7. Roose, E. 1996. “Land Husbandry - Components and Strategy.” FAO Soils Bulletin 70. Rome, Italy. Ruesch, Aaron, and Holly K. Gibbs. 2008. “New IPCC Tier-1 Global Biomass Carbon Map For the Year 2000.” Oak Ridge, Tennessee. Schmitt, R J P, S Bizzi, A F Castelletti, and G M Kondolf. 2018. “Stochastic Modeling of Sediment Connectivity for Reconstructing Sand Fluxes and Origins in the Unmonitored Se Kong, Se San, and Sre Pok Tributaries of the Mekong River.” Journal of Geophysical Research: Earth Surface, February. https://doi.org/10.1002/2016JF004105. Schmitt, Rafael J P, Simone Bizzi, and Andrea Castelletti. 2016. “Tracking Multiple Sediment Cascades at the River Network Scale Identifies Controls and Emerging Patterns of Sediment Connectivity.” Water Resources Research, 3941–3965. https:// doi.org/10.1002/2015WR018097. Schwanghart, Wolfgang, Marie Ryan, and Oliver Korup. 2018. “Topographic and Seismic Constraints on the Vulnerability of Himalayan Hydropower.” Geophysical Research Letters 45 (17): 8985–92. https://doi.org/10.1029/2018GL079173. Sharma, Sudirendar. 2003. “Rethinking Watershed Development in India: Strategy for the Twenty-First Century.” In Preparing for the Next Generation of Watershed Management Programmes and Projects, edited by Moujahed Achouri, Larry Tennyson, Kumari Upadhyay, and Roger White, 69–76. Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  88 Sharma, Suresh, and Madhuban Lal Maskay. 1999. Community Participation and Environmental Protection in the Construction of Mountain Roads: Promotion of the “Green Road” Approach in Nepal. Transport and Communications Bulletin For Asia And The Pacific. Sharp, R, H T Tallis, T Ricketts, A D Guerry, S A Wood, R Chaplin-Kramer, E Nelson, et al. 2014. “InVEST User’s Guide.” Stanford, CA: The Natural Capital Project, Stanford University. http://naturalcapitalproject.org/InVEST.html. Shrestha, H.R. 2009. “Harmonizing Rural Road Development with Mountain Environment: Green Roads in Nepal.” http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.550.2378&rep=rep1&type=pdf. Shrestha, Jayandra Prasad, and Namrata Tusuju Shrestha. 2016. “Expansion Planning of Electricity Generating System Using the VALORAGUA and WASP-IV Models in Nepal.” Hydro Nepal: Journal of Water, Energy and Environment. https://doi.org/10.3126/hn.v19i0.15352. Shrestha, Maheshwor. 2016. “Get Rich or Die Tryin’: Perceived Earnings, Perceived Mortality Rate and the Value of a Statistical Life of Potential Work-Migrants from Nepal.” Job Market Paper, no. January. https://doi.org/10.1111/j.1365- 2036.2010.04534.x. Shrestha, Ratna Sansar. 2011. “Electricity Crisis (Load Shedding) in Nepal, Its Manifestations and Ramifications.” Hydro Nepal: Journal of Water, Energy and Environment. https://doi.org/10.3126/hn.v6i0.4187. Shrestha, Ser Bahadur. 2016. “Soil Erosion and Payment for Sediment Retention in Kulekhani Watershed.” Agricultural and Forestry University (Nepal). Sidle, Roy C, and Alan D Ziegler. 2012. “The Dilemma of Mountain Roads.” Nature Geoscience 5 (7): 437–38. https://doi.org/10.1038/ngeo1512. Starkey, Paul, Ansu Tumbahangfe, and Shuva Sharma. 2013. “Building Roads and Improving Livelihoods in Nepal.” Stiglitz, Joseph, Nicholas Stern, Maosheng Duan, Ottmar Edenhofer, Gaël Giraud, Geoffrey Heal, Emilio Lèbre la Rovere, et al. 2017. “Report of the High-Level Commission on Carbon Prices.” Ipcc. https://doi.org/10.4103/2231-4040.101015. Struck, Martin, Christoff Andermann, Niels Hovius, Oliver Korup, Jens M Turowski, Raj Bista, Hari P Pandit, and Ranjan K Dahal. 2015. “Monsoonal Hillslope Processes Determine Grain Size-Specific Suspended Sediment Fluxes in a Trans- Himalayan River.” Geophysical Research Letters 42 (7): 2015GL063360. https://doi.org/10.1002/2015GL063360. Thapa, G. B., and G. S. Paudel. 2002. “Farmland Degradation in the Mountains of Nepal: A Study of Watersheds ‘with’ and ‘without’ External Intervention.” Land Degradation and Development. https://doi.org/10.1002/ldr.536. Timilsina, Govinda R., and Mike Toman. 2016. “Potential Gains from Expanding Regional Electricity Trade in South Asia.” Energy Economics. https://doi.org/10.1016/j.eneco.2016.08.023. Tol, Richard S J. 2009. “The Economic Effects of Climate Change.” Journal of Economic Perspectives 23 (2): 29–51. https://doi.org/10.1257/jep.23.2.29. ———. 2014. “Correction and Update: The Economic Effects of Climate Change.” Journal of Economic Perspectives 28 (2): 221–26. https://doi.org/10.1257/jep.28.2.221. Turowski, Jens M, Dieter Rickenmann, and Simon J Dadson. 2010. “The Partitioning of the Total Sediment Load of a River into Suspended Load and Bedload: A Review of Empirical Data.” Sedimentology 57 (4): 1126–46. https://doi. org/10.1111/j.1365-3091.2009.01140.x. UNISDR. 2015. “Desinventar Sendai.” 2015. 89 Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal United Nations Environment Programme (UNEP). 2012. “Economic Analysis of Local Government Investment in Rural Roads in Nepal 2011.” Vanacker, V, M Vanderschaeghe, G Govers, E Willems, J Poesen, J Deckers, and B De Bievre. 2003. “Linking Hydrological, Infinite Slope Stability and Land-Use Change Models through GIS for Assessing the Impact of Deforestation on Slope Stability in High Andean Watersheds.” Geomorphology 52 (3–4): 299–315. https://doi.org/10.1016/S0169- 555X(02)00263-5. Vogl, A.L., B.P. Bryant, J.E. Hunink, S. Wolny, C. Apse, and P. Droogers. 2017. “Valuing Investments in Sustainable Land Management in the Upper Tana River Basin, Kenya.” Journal of Environmental Management 195. https://doi.org/10.1016/j.jenvman.2016.10.013. Wasson, Robert J. 2003. “A Sediment Budget for the Ganga–Brahmaputra Catchment.” Current Science, 1041–1047. Wilcock, Peter R, and J Crowe. 2003. “Surface-Based Transport Model for Mixed-Size Sediment.” Journal of Hydraulic Engineering 129 (2): 120–28. https://doi.org/10.1061/(ASCE)0733-9429(2003)129:2(120). Wischmeier, W H, and D Smith. 1978. “Predicting Rainfall Erosion Losses: A Guide to Conservation Planning.” Washington DC: U.S. Department of Agriculture, Agriculture Handbook No. 537. WOCAT. 2012. “Rehabilitation and Management of Rangelands Rehabilitation and Management of Rangelands [Pakistan].” 2012. https://qcat.wocat.net/en/unccd/view/unccd_56/. (World Bank) International Bank for Reconstruction and Development. 2017. “Shadow Price of Carbon in Economic Analysis Guidance Note.” Washington, DC. http://documents.worldbank.org/curated/en/621721519940107694/ pdf/2017-Shadow-Price-of-Carbon-Guidance-Note.pdf. World Bank. 2005. “Himachal Pradesh Mid-Himalayan Watershed Development Project.” ———. 2008. “Sustainable Land Management Project (Ethiopia).” ———. 2009a. “Eastern Nile Watershed Management Project.” http://documents.worldbank.org/curated/ en/224371468009576741/pdf/450170PAD0P111101Official0Use0Only1.pdf. ———. 2009b. “Xining Flood and Watershed Management Project.” ———. 2012. “Karnataka Watershed Development Project -II.” ———. 2013. “Kali Gandaki A Hydropower Plant Rehabilitation Project: Project Appraisal Document.” http://documents. worldbank.org/curated/en/874211468290710744/Nepal-Kali-Gandaki-a-Hydropower-Plant-Rehabilitation- Project. ———. 2014. “Uttarakhand Decentralized Watershed Development Ii Project.” ———. 2018a. “Nigeria Erosion and Watershed Management Project (Newmap).” http://documents.worldbank.org/ curated/en/513991530329697812/pdf/Nigeria-Erosion-PAD2621-06052018.pdf. ———. 2018b. “Prioritizing Investments for Sediment Retention: Case of Kaligandaki Hydropower Plant.” Valuing Green Infrastructure: Case Study of Kali Gandaki Watershed, Nepal  90 1818 H Street, NW Washington, D.C. 20433 USA Telephone: 202-473-1000 Internet: www.worldbank.org/environment