Landscape and Urban Planning The changing land cover and fragmenting forest on the Roof of the World: A case study in Nepal’s Kailash Sacred Landscape

Land cover change is one of the most important drivers of forest ecosystem change. The Hindu Kush Himalayan region (HKH) has experienced severe forest degradation but data and documentation are limited. We undertook this study in the Nepalese part of the Kailash Sacred Landscape (KSL), an important transboundary region known for its biodiversity and the scared values. Forest is an important ecosystem within the landscape and provides various goods and services including habitat for many keystone species. However, precise information on forest change and overall land cover change in the area is limited. We analyzed land cover change and forest fragmentation between 1990 and 2009, and the pre- dicted change for 2030. There was a 9% decrease in forest cover and 12% increase in cropland between 1990 and 2009. A further 4% decline in forest cover and 5% increase in cropland was predicted by 2030, together with a slight increase in grassland and barren area. Fragmentation analysis showed a 10% decrease in large core forest between 1990 and 2009, accompanied by an increase in patch forest. A further 10.6% decline in core forest was predicted by 2030, accompanied by an increase in patch, perforated, small-sized core, and mediumsized core areas. The study suggests that expansions of cropland coupled with high dependency on forests are the major drivers of the observed forest change. Recommendations are made based on the results of the study that will help to maintain and restore forest, and support biodiversity conservation and livelihoods.© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


Introduction
Around 75% of the natural forested areas across the world have either been cleared or dominated by human activity since the last ice age (Ellis & Ramankutty, 2008). The global rate of forest loss is currently reported to be 0.6% per year (Hansen, Stehman, & Potapov, ecological integrity (Kettunen, Terry, Tucker, & Jones, 2007). Forest fragmentation, in which the forest is reduced to patches, can have a marked detrimental impact on biodiversity. Among others, it can result in homogenization (Lôbo, Leão, Melo, Santos, & Tabarelli, 2011), reduction in habitat quality for forest-interior species, and loss of forest health due to changes in microclimate and increased susceptibility to edge predators, parasites, and invasive species (Thuiller, Albert, Araújo, Berry, & Cabeza, 2008). Rare and patchily distributed species that require a larger range of a specific habitat are particularly affected by fragmentation (Fenoglio, Srivastava, Valladares, Luciano, & Salvo, 2012).
Apart from the impact on biodiversity, fragmentation can also negatively impact ecosystem processes and the flow of ecosystem services (Burkhard, Kroll, Muller, & Windhorst, 2009), which in turn affects the livelihoods of forest dependent communities (Chettri, Sharma, Deb, & Sundariyal, 2002). The fragmentation process may lead to landscape, ecosystem and habitat degradation (Leal, Filgueiras, Gomes, Iannuzzi, & Andersen, 2012;Schleuning, Farwing, Peters, Bergsdorf, & Bleher, 2011), and biodiversity loss (Crooks et al., 2011). Land cover change is becoming so prominent at a global scale that it is significantly affecting the Earth's ecosystems and functions (Lawler et al., 2013). By 2100, the impacts of land cover change on biodiversity at a global scale is likely to be more significant than climate change, nitrogen deposition, species introductions, and changing atmospheric concentrations of carbon dioxide (Sala et al., 2000).
The Hindu Kush Himalayan region (HKH) extends over more than four million square kilometers and includes all of Bhutan and Nepal and parts of six other countries: Afghanistan, Bangladesh, China, India, Myanmar, and Pakistan. It is the source of ten large Asian river systems -the Amu Darya, Indus, Ganges, Brahmaputra (Yarlungtsanpo), Irrawaddy, Salween (Nu), Mekong (Lancang), Yangtse (Jinsha), Yellow River (Huanghe), and Tarim (Dayan), -and provides water, ecosystem services, and the basis for livelihoods to a population of around 210.53 million people in the region. The basins of these rivers provide water to 1.3 billion people, a fifth of the world's population (Schild, 2008). Endowed with a rich variety of gene pools and species, and ecosystems of global importance (Chettri, Shakya, Thapa, & Sharma, 2008), the region hosts parts of four Global Biodiversity Hotspots: Himalaya, Indo-Burma, Mountains of South-West China, and Mountains of Central Asia (Mittermeier et al., 2004). Approximately 39% of the HKH is comprised of grassland, 20% forest, 15% shrub land, and 5% agricultural land. The remaining 21% are barren land, rocky outcrops, built-up areas, snow cover, and water bodies (Chettri et al., 2008). With 20% coverage, forest is one of the most important ecosystems in terms of habitat for flagship species (Chettri, Sharma, & Zomer, 2012;Kandel et al., 2015) and as a source of provisioning, regulatory, cultural and supporting services (Badola et al., 2010;Kubiszewski, Costanza, Dorji, Thoennes, & Tshering, 2013;Pant, Rasul, Chettri, Rai, & Sharma, 2012). However, the region has witnessed significant deforestation in the past (Ives & Messerli, 1989) which is still ongoing in many areas (Pandit, Sodhi, Koh, Bhaskar, & Brook, 2007). Although the HKH has witnessed significant progress in conservation, with 39% of land in protected areas (Chettri et al., 2008), the region is still facing challenges with the effectiveness of protected area management (Oli, Chaudhary, & Sharma, 2013), and protected areas are often isolated as conservation islands (Chettri et al., 2008). The conservation agenda is facing additional challenges with climate change (Singh, Singh, & Skutsch, 2010) and high rates of absolute poverty in some parts (Gerlitz, Hunzai, & Hoermann, 2012). Moreover, the region is poorly researched and the information available on biodiversity, land cover change, and climate change is far less than required. The fourth and fifth reports of the Intergovernmental Panel on Climate Change (IPCC) explicitly pointed to the HKH as a data deficit area (IPCC, 2007(IPCC, , 2014 Step 2: Land cover change analysis and prediction Change prediction Driver variables Step 3: Analysis of forest fragmentation LC reclassification Fragmentation analysis Fragmentation maps  , 2007). There is an urgent need to address these challenges through sustainable landscape planning, management, and development, and this will be crucial for the millions of people living in the HKH (Ives & Messerli, 1989;Schild, 2008) as well as the billions living in the downstream of the ten river basins. Transboundary landscape-level planning and management for biodiversity conservation, where initiatives are taken beyond the political boundaries for common conservation and develop goals, is an evolving concept in the HKH (Chettri, Sharma, Shakya, & Bajracharya, 2007;Zomer & Oli, 2011), which emerged primarily out of recognition that protection of protected areas (national parks, sanctuaries, wildlife reserves) is essential but insufficient as a biodiversity conservation strategy (Sharma, Chettri, & Oli, 2010). Almost one-third of the protected areas are transboundary, (Chettri et al., 2008) and in these areas, as elsewhere in the HKH, ecosystems and habitats extend across political boundaries. This means that landscape level planning is necessary and management requires regional cooperation if the ecosystems or habitats are transboundary in nature (Sharma, Chettri, Gurung, & Shakya, 2007). Seven transboundary landscapes have been identified across the HKH based on biodiversity significance, representation of ecoregions, cultural significance, and contiguity of ecosystems for conservation and sustainable development of the region (Chettri, Sharma, & Thapa, 2009), and are being used to develop transboundary landscape level planning and management approaches (Fig. 1).
The Kailash Sacred Landscape (KSL) is one of the seven landscapes. It covers parts of far-western Nepal, the central Indian Himalayas, and Tibet Autonomous Region (TAR) of China. A transboundary collaborative programme was established in 2010 between China, India and Nepal to conserve ecosystems and biodiversity in the landscape, while encouraging sustainable resources management (Chettri et al., 2009). The landscape is named after the most prominent feature, Mount Kailash (6714 m) in the TAR, one of the most sacred mountains in Asia (Brockman, 2011). The landscape has a strong cultural connection with India and Nepal as a gateway to Mount Kailash, and is famous for its ancient heritage and religious significance for Hindu, Buddhist, Bon Po, Jain, Sikh, and other religions, following traditions that go back millennia (Adler et al., 2009). The landscape is also home to 93 mammal species, 497 bird species, and 134 fish species among other fauna making it one of the richest areas in the western Himalayas (Zomer & Oli, 2011). Four of Asia's great rivers have their source in the landscape, the Indus, Brahmaputra, Karnali, and Sutlej, providing essential transboundary ecosystem goods and services, both locally and downstream (Zomer & Oli, 2011).
The Nepalese part of the KSL (KSL-Nepal hereafter) covers 13,289 square kilometers (42% of the total area) and covers four mountain districts (Darchula, Humla, Baitadi, and Bajhang) in the far-west of the country. This is one of the most underdeveloped regions of Nepal and faces numerous conservation and development challenges resulting from the harsh climate, poor accessibility, marginality, and high level of poverty, which manifests in a high dependency on natural resources leading to overexploitation (Kunwar, Mahat, Acharya, & Bussmann, 2013;Roy, Schmidt-Vogt, & Myrholt, 2009). Forest plays a vital role in supporting the rural livelihoods in the area. However, poverty driven human pressure is impacting on the existing forest ecosystem and leading to habitat fragmentation, ecosystem degradation, and a decrease in the capacity to provide ecosystem services (Uddin et al., 2015;Zomer & Oli, 2011). Information on the forest status and trends in the area is limited, and there is no cohesive framework available for regular assessment and monitoring of change across the landscape. Even basic questions such as what are the changes in land cover and what is the rate of forest change remain unanswered. Understanding the land cover change, especially forest change and fragmentation, is urgently needed both to understand past development and as a basis for landscape level planning to guide future conservation and development management interventions. The present study was designed to fill this gap and provide the first detailed information about forest change in the landscape. The main questions for the research were: 1. How is the forested ecosystem of the landscape changing over time? 2. How is forest fragmentation contributing to the trend of forest ecosystem change? 3. What recommendations can be made for conservation interventions in the landscape?

Study area
KSL-Nepal is located between 80 • 24 E to 82 • 49 E and 29 • 30 N to 30 • 44 N and covers an area of 13,289 square kilometers, with an elevation ranging from 518 to 7132 m above sea level (Fig. 1). The study area is situated at the junction of the Western Himalayan, Eastern Himalayan, and Central Asiatic regions and is part of the Himalaya Biodiversity Hotspot (Mittermeier et al., 2004). The area has five major ecoregions: Himalayan subtropical broadleaf forest, Himalayan subtropical pine forest, Western Himalayan broadleaf forest, Western Himalayan subalpine conifer forest, and Western Himalayan alpine shrub and meadows (Olson et al., 2001). It is significantly rich in biodiversity with 22 mammals, 12 birds, and 1 reptile listed on the IUCN Red List; 8 mammals, 7 birds, 22 reptiles, and 8 fish endemic to the region; and 35 species of mammals and 73 species of birds listed by the Convention on International Trade in Endangered Species (CITES). Globally threatened species such as snow leopard (Uncia uncia) and Himalayan musk deer (Moschus chrysogaster) inhabit the area. The total human population in 2001 was approximately 564,000. Agriculture is the main occupation for over 70% of the population (Zomer & Oli, 2011). About 22% of the land is covered by forest which is a major source of fuel, fodder and timber among other and play an important role in supporting the subsistence livelihoods. Deforestation is prominent and the demand for forest land for agriculture and rural infrastructure such as roads, electricity, water supply, hospitals, and schools is quite high. Timber harvesting and export has been identified as a major concern, with 3000 cubic meters of timber per annum supplied officially to China and India and an unknown additional amount traded illegally. Shifting cultivation is still practiced with short rotation period (Zomer & Oli, 2011). The fragile nature of the area, unsustainable resources harvesting, coupled with development activities pose a serious threat to the forested ecosystems in the landscape, with implications for biodiversity and people.

Methods
Fig. 2 shows the overall methodological framework. Briefly, land cover maps were developed for the landscape for 1992 and 2009 using object based image analysis of satellite images. Land use and land cover change was then analyzed from the maps using standard software, and a map of potential land cover in 2030 prepared based on the past change and variables taken as drivers. Finally, all the land cover maps were analyzed to detect forest fragmentation. The details of the procedures and reasons for use are given in the following sections.

Analysis of land use and cover change (LUCC)
Land use and land cover change (LUCC) maps were prepared for the landscape for 1990 and 2009 using medium resolution spatial images from Landsat thematic mapper (TM) (1990/1992) and linear imaging self-scanning sensor (LISS III) (2009) ( Table 1). The images were collected at the same time of year to avoid problems of seasonal variation. A hierarchical land cover classification system (LCCS) was used (Table 2) following Gregorio (2005).
Preprocessing of the satellite images was carried out prior to image classification to bring the images into a standard projection so that they could be overlaid. Ground control points (GCP) were generated from the Landsat TM images and used for geometric correction of the IRS LISS-III images into Universal Transverse Mercator (UTM) Zone 44. The images were then re-sampled using a convolution algorithm to a common resolution of 30 m.
Object-based image analysis (OBIA) was performed using eCognition developer software to derive homogeneous image objects through segmentation. OBIA provides a methodological framework for machine-based interpretation of complex classes using both spectral and spatial information and generates better classification results with a higher degree of accuracy than pixel-based methods (Chettri, Uddin, Chaudhary, & Sharma, 2013;Lang et al., 2011;Uddin et al., 2015). Finally, a multi-resolution segmentation algorithm was applied in which homogeneous areas resulted in larger objects and heterogeneous areas in smaller ones. The algorithm helps to merge pixels with their neighbors based on relative homogeneity criteria (Baatz, Arini, Schäpe, Binnig, & Linssen, 2006). Several segmentations were tested using different parameters until a satisfactory result was obtained.
Information on the spectral values of image layers, vegetation indices (Normalized Difference Vegetation Index, NDVI and Normalized Difference Snow and Ice Index, NDSII), and a land water mask was used in the analysis. The layers were created through band rationing, slope, and texture information. Objects with an area smaller than the defined minimum mapping units were merged with other objects (Bajracharya, Uddin, Chettri, Shrestha, & Siddiqui, 2010). Initially sixteen different land cover types were used in the classification process, which were then merged to give a land cover map with seven broad categories. The categories are shown in Table 2. The classified land cover map and forest layers were exported to raster file format for land use and cover change (LUCC) and forest fragmentation analysis. The accuracy of the land cover map was assessed by selecting 250 stratified pixels at random in a Google Earth image as reference, and ascertaining the accuracy of the classification in a field survey using a global positioning system (GPS) to locate the points. The overall accuracy was 84.8% in 1990 and 88.4% in 2009.

Land use and cover change (LUCC) analysis and modeling
The change in land use and land cover (LUCC) was analyzed to depict gains and losses in graphical form. LUCC modeling was carried out using Land Change Modeler (LCM) in IDRISI Taiga software from Clark Labs (Eastman, 2009), which is a well known software for LUCC analysis (Frey & Paul, 2012;Fu et al., 2005). LCM uses the feed-forward concept of a multi layer perception neural network and follows the artificial neural network modeling technique (Atkinson & Tatnall, 1997). LCM has three basic elements: change analysis, transition potential modeling, and change prediction. The change analysis panel provides a rapid quantitative assessment of change by visualizing gains and losses from different land cover categories; transition potential modeling, or net change, shows the results of considering the original land cover areas adding gains, and subtracting losses; and change prediction examines the contributions to change experienced by a single class of land cover. The land cover maps for 1990 and 2009 were imported into IDRISI Taiga in a raster (IDRISI raster) format. Differences in the maps were reviewed and assessed with LCM to visualize and quantify land cover changes. LCM was then used for empirical modeling of land cover change. The dominant transitions from past land cover change were identified, and related to land cover transitions in order to link with drivers of change. Slope and infrastructure development were considered as major drivers of deforestation in the change analysis. Slope can affect LUCC, as forest in flatter and more fertile areas is more likely to be cleared for agriculture; and infrastructure development is related to urbanization and an increase in built-up area, which may lead to an increase in agricultural land at the expense of forest area. Based on this concept, maps of 'distance-to-roads' and 'distance-to-slope' were produced using the 'Euclidean Distance' tool in ArcGIS. These maps were imported to IDRISI raster format and the drivers were incorporated in the LCM as explanatory driver variables of change for a particular transition. A map of potential land areas that will go through transitions was created to predict changes. The prediction year was specified as 2030, and predictions were examined using the change prediction tab. Predictions were also examined taking into consideration the possible interventions, especially within conservation and development projects. The analyzed land cover map for 2030 was then exported for forest fragmentation analysis.

Assessment of forest fragmentation
The ArcGIS Landscape Fragmentation Tool was used to study forest fragmentation and edge effects as described by Vogt et al. (2007). The method is based on a procedure to map forest fragmentation based on forest and non-forest land cover; the tools quantify forest fragmentation from raster land cover maps. The land cover maps for 1990 and 2009 and predicted land cover map for 2030 were reclassified into forest and non-forest classes using ArcGIS spatial analyst. GIS was used to identify patch formation within the extended forests, including edge and perforated areas, and intact and contiguous forested areas (Fig. 3), and assign them to a patch size class as described by Stokes and Morrison (2003). Using a specified edge width of 100 m, forest fragmented areas were classified into 'core' forest-relatively distant from the forest-non forest boundary; 'patch' forest-forests too small to be considered as core forest; 'perforated' forest-boundaries between core forest and relatively small perforations; and 'edge' forest-boundaries of relatively large perforations and the exterior boundaries of core forest regions (Vogt et al., 2007). 'Core' forest was further divided into 'small core' (<101.17 ha), 'medium core' (101.17-202.34 ha), and 'large core' (>202.34 ha) areas.

Land use and cover change (LUCC) and prediction
Forest was the dominant land cover in 1990 with 24.4% of the total area of KSL-Nepal, followed by snow and glaciers (22.3%) and grassland (20.6%) (see Table 2). The forest area had decreased by about 9% (28,070 ha) in 2009, to 22% of the total area. Shrubland also decreased by 9.3% (from 7223 ha in 1990 to 5760 ha in 2009), while cropland increased by 12% (from 187,802 ha in 1990 to 210,189 ha in 2009). Other major changes were a 22.3% increase in barren area and a 23.1% decrease in snow/glacier area. The land cover distribution and changes are shown in Table 2 and Fig. 4. The conversion of land cover between different classes between 1990 and 2009 is shown in the form of a land cover change matrix in Table 3. There was a major change from forest to cropland (17,475 ha) and forest to grassland (9845 ha), from grassland to cropland, and from snow/glacier to barren land. No area was converted to forest.
The predictions for 2030 showed a further loss of 4% of forest cover (5320 ha converted to grassland and 6465 ha to cropland), a 5% increase in cropland, and a slight increase in grassland (Tables 2 and 4). No significant area was converted into forest. Predictions of land cover change (2009-2030) that included interventions that could halt the present rate of deforestation and restore degraded forests showed a 1% increase in forest area. The spatial distribution of land cover for 2030 is shown in Fig. 4.

Forest fragmentation and prediction
The distribution of forest fragmentation showed significant changes between 1990 and 2009, and further change to 2030. In 1990, large core forest was dominant covering 60% of the forest area; edge forest covered 28%, small core 5%, perforated 3.8%, medium core 1.7%, and patch forest 1.5%. In 2009, large core decreased by 10% and perforated forest by 58%, while medium core, small core, edge, and patch forest increased by 6%, 12%, 1.8%, and 26.7%, respectively.
By 2030, the area of large core was predicted to decrease by a further 10.6%, while medium core was predicted to increase by 39%, small core by 14.3%, perforated by 250% (to 30% more than the 1990 value), edge by 0.6% and patch by 10.5%. The details are shown in Fig. 5 and the values are summarized graphically in Fig. 6.

Land use and cover change, and prediction
Overall, the vegetated area in KSL-Nepal is decreasing and nonvegetated area is increasing. The results are in line with the report by Pandit et al. (2007). Land use and cover change is now recognized as one of the most important drivers of global change (Lambin et al., 2001). The overall changes are likely to have a negative impact on biodiversity, ecosystem services, and the subsistence livelihoods of the people living in the region (Hofer & Messerli, 2006;Myers, 1986). The biodiversity is of global significance, especially endemic and endangered species such as Himalayan musk deer (Moschus leucogaster) and Asiatic black bear (Ursus thibetanus), which would be strongly affected by loss of contiguous forest habitat. Zomer and Oli (2011) has already identified forest degradation as one of the  major reasons for biodiversity loss in the Himalayas. The area of KSL-Nepal includes global biodiversity hotspots, important ecoregions, and the headwaters of major rivers, thus the rate of forest change is a matter of global, as well as national and regional, concern. The ecosystem services provided by KSL-Nepal could be at risk. As more than 90% of the population depends on forests for fuelwood and other products this will have a significant impact on livelihoods. The local communities are highly dependent on forest resources and already face challenges in sourcing fuelwood, fodder, and timber (Zomer & Oli, 2011), which will be further exacerbated by a decline in forest cover, thus increasing poverty in the area. The results suggest that expansion of cropland is one of the major drivers of land cover change, as also reported elsewhere in the Himalayas (Alam, Rashid, Bhat, & Sheikh, 2011;Bawa & Seidler, 2015). In KSL Nepal, more than 70% of population is dependent on agriculture (Zomer & Oli, 2011) and the predictions indicate a further increase in cropland and decrease in forest. However, when interventions that supported forest conservation were included in the model, there was an increase in forest area. This indicates that landscape planning in the area should focus on interventions to conserve forest, and thus maintain the rich biodiversity and flow of ecosystem services and secure local livelihoods as suggested by Bawa and Seidler (2015 Forest ar ea (ha) Forest fragmentaƟon Fig. 6. Forest fragmentation and change in KSL Nepal. a mountainous area like this, where the landscape is fragile with steep slopes and high elevation, and loss of soil cover can mean that recovery of the ecosystem can be very slow or even impossible (Halada, 2010).
Land cover change studies are crucial for developing effective plans for natural resource management . A similar study conducted in the mountains of the Peruvian Andes also identified the need for land cover change assessment for conservation planning (Kintz, Young, & Crews-Meyer, 2006). The present study is the first of its kind in this area and provides a baseline for planned conservation and development planning and interventions. The results will not only help to fill the data gap for Nepal and the HKH region, they will also contribute to better understanding of global change and underlying causes, as the global data on land cover change is poor (Turner, Meyer, & Skole, 1994), and the need to conduct land cover assessment at a local scale, especially in developing countries, has been put emphasised by several studies (Lambin et al., 2001;Loveland et al., 2000;Vadrevu, Justice, Prasad, Prasad, & Gutman, 2015).

Forest fragmentation and prediction
Forests generally become fragmented through intensification of human activities (Peres et al., 2010). The results of this study suggested cropland expansion as one of the major reasons for the observed fragmentation along with high dependence on forest resources coupled with illegal timber extraction as shown by literature review. Reddy, Sreelekshmi, Jha, and Dadhwal (2013) in India and Millington, Velez-Liendo, and Bradley (2003) in Bolivia also proposed socio-economic processes as a reason for forest fragmentation. Forest fragmentation increased in KSL-Nepal between 1990 and 2009, and was predicted to increase further by 2030: with less core, edge, and perforated forest, and more patches. Similar results have been noted in different geographic zones in India (Reddy et al., 2013); and the Alpine grasslands of Abiseo National Park (Kintz et al., 2006) and elsewhere (Lele & Joshi, 2009;Saikia, Hazarika, & Sahariah, 2013). However, comparable data from the Himalayas is limited, and the studies that have been carried out have used different temporal and spatial scales (Nagendra, Paul, Pareeth, & Dutt, 2009). Forest fragmentation can have multiple impacts on ecosystem functions (Nagabhatla, Finlayson, & Sellamuttu, 2012), including habitat loss (Fahrig, 2003), change in species abundance (Pardini, de Arruda Bueno, Gardner, Prado, & Metzger, 2010), and overall loss of biodiversity (Biswas & Khan, 2011;Wenguang et al., 2008). The trends and predictions of forest fragmentation together with the land cover change in KSL-Nepal could have a serious impact on the rich biodiversity of the area as well as on cultural, supporting and regulatory services, including the water catchments.
These impacts could be halted in the KSL Nepal if the Kailash Sacred Landscape Conservation and Development Initiative (KSLCDI), which is supporting planning at the landscape level over all three countries, could design interventions that address the issue of forest loss and degradation. Interventions of the type proposed would reverse the decline and enable forest cover to start increasing. Improved understanding of the dynamics of land cover change and forest fragmentation will help resource managers to understand the ecological processes and introduce adaptive and restoration measures (Bharti, Adhikari, & Rawat, 2012;De & Tiwari, 2008;Nagendra, Pareeth, Sharma, Schweik, & Adhikari, 2008). The results of this study provide timely inputs to the implementation of KSLCDI and other interventions. The results will also help governments, non-governmental organizations, and other stakeholders to make decisions and take pro-active action in those areas where the fragmentation trend is marked, to support biodiversity management and safeguard people's livelihoods.

Conclusions
Over a period of 19 years, KSL-Nepal has experienced a decrease in forest cover, and an increase in cropland and barren land. Forest cover is likely to decrease further, and cropland to increase, by 2030. The results indicate that forest fragmentation coupled with land cover change may lead to critical forest degradation with implications for biodiversity, ecosystem services, and people's livelihoods. Globally threatened species and species with a limited habitat range, and forest dependent communities, are likely to be most strongly affected. The study further suggests that cropland expansion, high dependency on forest, and illegal timber extraction are the major drivers of forest change and fragmentation. This study fills an information gap in a poorly researched area with poor data availability, improves the information base at national and regional scales, and may contribute to understanding of global change. Such information is required for informed decision making and planning for landscape conservation and development interventions. In particular, the results will be useful for planning landscape connectivity and corridors, which are important for conserving biodiversity and maintaining the flow of ecosystem services. The implementation of conservation and development interventions by the KSLCDI will benefit in particular as the initiative aims to conserve the rich biodiversity and develop the region sustainably. Future research should aim to increase understanding of the indirect drivers that are causing forest fragmentation and agricultural expansion, and the impact on biodiversity and livelihoods. Based on the results of this study, the following recommendations are made: • extend the study across the entire landscape to see the pattern and processes in the area within other countries and design complementary interventions; • promote landscape connectivity and an ecological network through reforestation and restoration of non-forested areas, especially areas between patch and perforated forests; • regularly monitor the dynamics and patterns of forest and overall land cover change; • promote sustainable forest management practices with incentives to local people, e.g. agroforestry, or forest-based enterprises to diversify the high dependency on forest resources; • support effective implementation of conservation and development initiatives like the KSLCDI in collaboration with local, national, and regional partners; • promote ecotourism involving local people to diversify livelihoods, building on the sacred value for different religions; and • promote alternative energy sources (such as solar energy or biogas) to reduce pressure on the forests for fuelwood.