Effects on ecosystem services value due to land use and land cover change (1990–2020) in the transboundary Karnali River Basin, Central Himalayas

Valuation of ecosystem services based on land use and land cover changes (LUCC) offers an incentive to people for sustainable use of the natural resources and can encourage people to adjust the land use sustainably. In this study, we used “Object-based Image Analysis (OBIA),” a remote sensing technique to extract the land use and land cover (LULC) of the transboundary Karnali River Basin (KRB, China and Nepal) from 1990 to 2020, and ecosystem services value (ESV) coefficients derived for the Tibetan Plateau has been used to assess the ESV. The basin has highest percentage of forest (33.44%), followed by bare area (30.29%), shrub/grassland (18.5%), agriculture (13.12%), snow/ice (4.36%), waterbody (0.3%), and built-up (0.03%) as of 2020. Over 30 years, 4.07 km2 of the forest has been converted to agricultural land, 3.31 km2 of agricultural land has been encroached by built-up area, whereas 2.82 km2 of snow/ice area has melted into the waterbody. Furthermore, 80.85 km2 of bare area has been converted to snow/ice, and 2138.83 km2 of snow/ice has been changed to bare area. The ESV of KRB has increased by nearly 2.7 million USD from 1990 to 2020, primarily due to the increase in ESV from the conversion of 133.09 km2 of snow/ice to shrub/grassland. The ESV of forest, waterbody, and snow/ice has decreased, whereas it has increased for other LULC classes in the basin. Spatial distribution of LUCC and assessment of ESV can be a tool to facilitate for better provisioning and regulating the resources for the future. Over the 30 years forest, waterbody, and snow/ice has decreased, whereas shrub/grassland, agriculture, bare area, and built-up has increased in the basin. The ESV of the basin has increased by 2.7 million USD in 30 years. Anthropogenic as well as climate change are the drivers of LUCC-driven ESV changes in the basin. Over the 30 years forest, waterbody, and snow/ice has decreased, whereas shrub/grassland, agriculture, bare area, and built-up has increased in the basin. The ESV of the basin has increased by 2.7 million USD in 30 years. Anthropogenic as well as climate change are the drivers of LUCC-driven ESV changes in the basin.


Introduction
Ecosystem services (ES) are the essential contribution of the ecosystem that plays a major part in human wellbeing, livelihood, and survival [1,2]. These services consist of provisioning (food, raw materials, medicinal, etc.), regulating (water and air purification, climate regulation, waste decomposition, etc.), supporting (habitat provision, primary production, etc.), and cultural services (spiritual, historical, recreational, etc.) [3]. ES flows from nature to human well-being through interaction in the form of human capital, built capital, and social capital [2]. ES is can be based on payment mechanism that contributes additional money to nationwide revenue, which further supports in improving livelihoods [4,5]. The quantification and analysis of the valuation of ecosystem services along with its changes can be an essential tool to promote awareness [6], contributing to management of the natural capital [3], and providing an incentive to conserve and promote ecosystems yielding higher valuable services [7]. Therefore, research interests and studies of Ecosystem Services Value (ESV) have spread rapidly and extensively all around the globe [2,8].
The provision of ecosystem services evaluation is directly linked to various land use and land cover (LULC) types [2,3,9,10]. The value of ecosystem services is estimated per unit area by biome and then multiplied by the total area of each biome [2,3]. The alteration or changes in the LULC directly influences biological diversity [11], leading to changes in the ecosystem services that further affect the biological systems to support the basic human needs [12]. These changes also partly determine the vulnerability of people and their dwelling places to climatic, economic or socio-political perturbations [13,14]. Thus, LULC dynamics changes the ESV [15], either increasing the provision of some services while decreasing others that support human needs, which indicates ecological degradation, or vice versa [16], and changes in the ESV depends on the magnitude and direction of the land use and land cover change LUCC in the given area [17].
Most of the LUCC related studies suggest to have an extended time period for LUCC and assessment of ESV [18,19] to draw a clearer picture of the LUCC trend and its effect on ESV assessment. In context of Nepal, the LUCC-driven ESV related studies of the three major river basins are of limited periods, i.e., 17-25 years. The KRB from 2000 to 2017 [20], the Gandaki river basin from 1990 to 2015 [21], and the Koshi river basin from 1990 to 2010 [22]. Moreover, the studies in the three major river basins of Nepal only used two time periods irrespective of the duration of the analysis. This limits addressing the fluctuation in LUCC in between time periods which further affects ESV assessment. This study aims to fill both the gaps by extending the study periods up to 30 years and carrying out decadal LUCC analysis in order to provide a robust land cover database of the KRB for future references.
The remainder of this paper is organized as follows. Section 2 presents an overview of the study area and proposed methodology for the extraction of LULC from satellite images and calculation of ESV adopted for this study. Section 3 discusses the present scenario of LULC, LUCC, and ESV in the KRB. Section 4 further elaborates on the major findings and discussion regarding the status and drivers for LUCC and the dynamics of ESV as compared to other two basins of Nepal. Finally, Sect. 5 presents our conclusions and discusses future work.

Study area
The Karnali River Basin (KRB) lies in the far-western side of Nepal. The transboundary KRB has a small part in China (Burang), and the Karnali river originates from the south of Mansarovar and Rokas lakes. The entire basin extends from 28° 19′ 39″ to 30° 41′ 19″ North in latitude and 80° 33′ 35″ to 83° 40′ 56″ East in longitude (Fig. 1). Of the total area 46,124 km 2 , 3085 km 2 falls in China (6.69% of total area) and 43,039 km 2 (93.31% of total area) in Nepal. As the elevation of the basin ranges from 124 m to above 4000 m, it covers all the climatic zones present in the country, i.e., alpine, sub-alpine, temperate, subtropical, and tropical zones. The basin extends to all the five physiographic regions of Nepal, namely, Terai, Siwalik, Hill, Middle mountain, and High mountain.

Data collection
Landsat Level 1 (20 geometrically corrected scenes) covering the entire study area (Table 1), taken from the website of United States Geological Survey's (USGS) (https:// earth explo rer. usgs. gov/) were used for extraction of LULC classes. The images of the winter season or post-monsoon (November-February) were used, as in this period the vegetation is green to semi-deciduous [23], and the effect of cloud cover and seasonal snow is minimal [24]. However, images of alternative years or seasons were used in case of excessive cloud or snow cover for the given selected time period.

Image processing and land use and cover classification
The use of different sensors and methodologies make it difficult to compare the land cover datasets of different time period [25]. Thus, the methodology devised for the LULC classification used in the study of [20] has been used to remove any discrepancies originating from the use of different sensors and methodologies. The land cover has been estimated using geographic object-based image analysis (GEOBIA), which was also used to extract the national land cover data for 2010 [23,26]. The GEOBIA, carried out using the eCognition Developer 9.0 software, gives better classification results with higher accuracy than the pixel-based methods because it uses spectral as well spatial information [25,27]. The different band layers of the images acquired for different time period (1990,2000,2010, and 2020) were stacked into one image for their respective year. Radiometric and atmospheric corrections needed were carried out to each five images of their respective year. Landsat level 1T data selected were later on corrected by system radiation and topographic correction, which was done using digital elevation model (DEM). The ENVI 5.3 FLAASH atmospheric correction module was adopted to correct the error due to atmospheric effects. Furthermore, Gram-Schmidt pan sharpening method was used to exploit the capability of both panchromatic and multispectral image products of Landsat imagery. After the necessary corrections, mosaicking of the images for the individual year (each five images mosaicked into one for the respective year) were carried out. Co-registration of the images is an important step that facilitates in generating change detection. The mosaicked image of 2020 was used as the base image to co-register other images of different temporal resolutions. After co-registration, each year's image was subjected to segmentation. The "multi-resolution segmentation" algorithm was used, which groups relatively similar pixels into segment or image object [28][29][30]. The image objects thus generated were subject to classification using various indices as well as rigorous visual interpretation. The band ratios of spectral values were used to extract various indices like normalized difference vegetation index (NDVI), normalized difference snow index (NDSI), normalized difference water index (NDWI), and bare soil index (BSI), which facilitates in classification of the image objects to their respective LC classes ( Table 2). The rulesets were developed exploring these indices and spectral bands to classify the image objects into different land use land cover classes. The generated land cover data were validated using high-resolution Google Earth images, previous national [23], and global [31] datasets. A total of 890 sampling points were generated for verification (Fig.  S1). Further, we generated a total of 904, 901, 905, and 890 stratified random points in the classified image of 1990, 2000, 2010, and 2020 respectively, using ArcGIS. The points were then converted to KML (Keyhole Markup Language) format in order to open in Google Earth. Also, the fishnet tool in GIS software was used to check and correct the classified output raster file. The fishnet created had a window of around 10 km × 10 km and overlaid in Google Earth. The classified outputs were then checked with respect to Google earth for verification. A total number of 494 such fishnet covers the entire KRB (Fig. S2).

Ecosystem services value assessment
To estimate the global ESV, 17 ecosystem services and 16 biomes were used [3], from which the ESV coefficients were derived subsequently by multiplication of coefficient values of ecosystem services by the total area of each biome. The global estimation has been modified and adopted for different studies [2,10]. Using the global estimate of ESV, the ESV for Tibetan Plateau (TP) was also derived based on expert-based survey [9]. These modified ESVs have also been used in Koshi river [22], Gandaki river basin [21], and Karnali River Basin [20]. As there are not specific coefficients derived only for Nepal, all the river basins in Nepal used the coefficients derived for the LULC category of the TP (Table 3). Here, the weighted average value calculated between shrubs and grass has been used. Due to the lack of ESV coefficients for the built-up category, it has not been calculated, as in previous studies [2,9]. Multiplication of the coefficient of ecosystem services value with area of respective LULC gives the ESV of each land cover category. Further, the ESV coefficients for time periods other than 2003 have been inflation-adjusted, as monetary valuation changes with demand and price of goods over time [20,33,34]. The inflation calculator devised by the Bureau of Labor Statistics, United States Department of Labor was used for the Consumer Price Index (CPI) inflation adjustment.

Accuracy assessment
The accuracy of LULC classification for the study was computed as 87.92% by producer's and 85.23% by user's accuracy in 2020. The overall accuracy ranged from ~ 86 to 88%. The classification accuracy of each class is shown in Table 4. The forest and agriculture land classification have the highest accuracy. The use of indices helps to differentiate between vegetation and non-vegetation area and choosing the time period after harvesting season effectively segregates forest from agricultural area. Whereas waterbody and built-up have lower accuracy, possibly due to the scattered and rare nature of the land cover classes [35]. The accuracy assessment carried out for 1990, 2000, and 2010 are provided in the supplementary files as Tables S1, S2, and S3, respectively.
There is a decrease in forest area, waterbody, and snow/    ice categories, whereas an increase in shrub/grassland, agriculture, bare area, and built-up categories ( Table 5). The highest annual increasing rate is for bare areas, whereas the highest annual decreasing rate is seen in snow/ice areas. This is mainly due to the seasonal variation of snow cover, even within the same month. There is a low annual decreasing rate of forest and waterbody,

Decadal LUCC
The decadal LC change shows that the KRB has experienced some changes over the last 30 years. In the span of 30 years, 4.07 km 2 forest has been converted to agricultural land, especially in the lower elevation of the basin, whereas 3.38 km 2 of agricultural land has been encroached by built-up area. Birendranagar, the largest city in the basin, has an area increased by 4.7 km 2 during the study period and the population increased by more than 50% by 2017 [36]. The highest change in the basin is in the form of snow/ice converting to bare area amounting to 2138.8 km 2 in the higher altitude of the basin. Meanwhile, 133.09 km 2 of snow/ice has changed to shrub/grassland. It further shows that 2.82 km 2 of snow/ice has been converted to waterbody ( Table 6). It is basically due to the creation of novel glacial lakes or the spreading out of the prevailing ones in the basin. The meteorological datasets in the KRB further show the seasonal and inter-annual temperature and precipitation trends for the winter season from 1980 to 2020 (Fig. 3). The warming rate is faster in higher elevation than in the lower elevation, with pre-monsoon and winter seasons having the highest warming trend in the basin [37]. Also, the observed meteorological datasets from 1980 to 2020 (Fig. 3) show an increasing trend in temperature whereas, there is a decrease in precipitation (− 0.83 mm/year) in the KRB. The decadal analysis (Tables S4, S5, and S6) estimates that forest, waterbody, and snow/ice are in a continuous decreasing trend. In contrast, shrub/grassland, bare area, and built-up are continuously increasing over all the decades. Agriculture, however, has had an increasing trend for the first two decades (from 1990 to 2010) and decreased in the last decade (from 2010 to 2020). This could be due to the agricultural land abandonment basically converting to shrub/grassland and bare area, especially in the hilly region and higher altitudinal regions [38]. The largest decrease in forest area (by 2.79 km 2 ) is in the first decade, followed by 1.63 km 2 and 1.98 km 2 in the latter two decades. Various policies and government intervention has contributed to forest conversation in Nepal [39][40][41].

ESV of the basin
The ESV of KRB for the years 1990, 2000, 2010, and 2020 was 45.88 × 10 8 , 45.87 × 10 8 , 45.92 × 10 8 , and 45.90 × 10 8 USD, respectively. Among all the LC classes, forest provided the largest ESV in the basin, whereas snow/ice provided the least ESV. The ESV was in increasing trend until 2010 and decreased from 2010 to 2020. However, from 1990 to 2020, there is a net increase in ESV by nearly 2.7 × 10 6 USD, without inflation adjustment. There is a decrease in ESV for the forest, waterbody, and snow/ice, whereas ESV of shrub/grassland, agriculture, and bare   (Table 7). With inflation adjustment, the ESV had increased even for the same LC area in different temporal scales. For example, the forest area has decreased in the basin over the time period, however, when ESV is compared with inflation adjustment, the ESV of forest increases significantly, which does not reflect the ground scenario but can be used for reference purposes only. The increase in ESV of the basin (without inflation adjustment) is primarily due to the conversion of bare area and snow/ice to shrubs/ grassland. The decrease in ESV due to conversion of forest to other classes is canceled out by the transformation of bare area and snow/ice to shrub/grassland, which caused the increment in ESV of the basin (Table 7). Since ESV coefficient of snow/ice and bare area are identical for a given time period (Table 3), the inter-conversion of bare area to snow/ice and vice-versa does not influence the ESV of the basin for a given time period. Hence, the most changes influencing the ESV change of the basin are primarily due to increments in shrub/grassland category.

The status of LUCC in the KRB and its drivers
The effect of changing climate affecting the LUCC is predominantly visible in the higher elevation of the Karnali River Basin (KRB). The change from snow/ice category to waterbody amounts to nearly 3 km 2 during the study period, primarily due to the melting glaciers converting into glacial lakes. The increasing temperature trend in winter season in the basin could be the reason for melting of glaciers into lakes. The highest and lowest temperature of 9 °C and 4 °C was observed in the year 2015 and 1983, respectively. Further, mean temperature shows an increasing trend (~ 0.07 °C/year) in the winter season over the study region (Fig. 3b). In the KRB, the glaciers have shrunk by 30% from 1980 to 2010 [42], and this study shows a 52% decrease in snow/ice area over the last 3 decades. Furthermore, decreasing winter precipitation was observed over the KRB during the study period (Fig. 3a). The deficit of rainfall in winter in most of the western region is the main cause of drought over recent years [43,44].
The anthropogenic factor is another major cause of LUCC, which is dominant in the lower elevation of the basin [45]. The decline in the forest area and increment of agricultural area in the basin is basically due to human activities [39]. In the basin, the people primarily depend on the forest for resources, livestock grazing, fuelwood consumption, and timber extraction [45]. However, the decreasing trend of deforestation, in the latter decades, in the basin is largely due to government intervention through various policies such as community forestry [40,41], leasehold forestry, protected forest, collaborative forestry, and promotion of reduction of emission from deforestation and forest degradation in developing countries (REDD+) as a mechanism to control deforestation and degradation [39]. Furthermore, the decreasing trend of agricultural land in the latter decade could be due to the loss of a greater proportion of population on the middle mountain and hill region [46]. Land abandonment is currently an issue in context of whole Nepal which causes environmental as well as socio-economic problems and should be discouraged implementing policies regarding  [47]. From 2000 to 2017, there was an increment by 23.14% of the built-up area in the basin [20], whereas it has increased by 48% from 1990 to 2020. Although the basin has a smaller built-up area as compared to the other two basins of Nepal, the slower development activities tend to be the reason for the LUCC. The KRB is among the under-developed regions of Nepal [20], with the least human development index of 0.386 [48], where economic activities are limited and faces various developmental and conservation challenges [45,49]. Infrastructural development is inevitable in context of developing country like Nepal. Thus, scientific land use zoning incorporating different levels of administrative units are essential for the sustainable LUCC [47]. The KRB has had the least annual growth rate in builtup (0.0004%) ( Table 8) as compared to two other basins, indicating the slowest development activities. Among the three major river basins of Nepal, Koshi basin had the highest annual change rate in forest (− 0.2078%), shrub/ grassland (0.2568), agriculture (0.2443%), and bare area (− 0.2417) categories, due to climatic as well anthropogenic factors [22]. The Gandaki basin had the highest annual change rate in waterbody (0.009%) due to the climatic factors as well as changing river course, and settlement area due to rapid urbanization [21], but the KRB had the maximum annual change rate in the snow/ ice (− 0.1585%) category. The conversion of snow/ice to shrub/grassland is partly due to the seasonal variability of snow cover in different temporal scales, even during the same season and month, which is one of the major limitations of the remote sensing approach. However, the images from the same season i.e. post monsoon were selected as it has minimum snow and cloud cover [50] to minimize this limitation. Furthermore, with the postmonsoon image, it better helps segregate forest, shrub/ grassland and agriculture based on the indices.

ESV dynamics of KRB and its comparison with other two basins
The ESV of the KRB has increased by 2.7 million USD, without inflation adjustment, within 30 years, whereas Shrestha et al. [20] had reported an increase of 1.59 million USD in 17 years from 2000 to 2017. The decrease of forest area only signifies degradation of the forest ecosystem [39,51]. Additionally, the decadal change analysis for an extended temporal resolution shows that during 1990-2000, the deforestation rate was higher, which waned in latter decade as revealed by the low rate of forest change. Although the ESV coefficients of water and forest are higher, the conversions from these classes to others have not had a larger effect than the gain in shrub/ grassland category. At a higher altitude, 2.82 km 2 of snow/ ice have been converted to water, and as the ESV coefficient of water is about 100 times higher than snow/ice, it seems to have had a positive effect on the mountain ecosystem. However, the risks associated with glacial lake expansion to the downstream communities [24,52] cannot be regarded as the betterment of ESV, particularly in the higher altitudes. Thus, further robust research for the quantification of benefit-risk assessment due to changes in ESV in the mountain ecosystem is inevitable [20]. Furthermore, the ESV coefficient of solid form of water i.e. snow/ ice should be higher compared to bare area as changes in the glacier and snow reserves will affect the overall hydrological cycle which influences the natural environment, including biodiversity, and the ecosystem services that glacier-fed rivers provide to society, particularly provision of water for agriculture, hydropower, and consumption [53]. Thus, future research should focus on this aspect. The decrease in global ecosystem services value aggregated to about $4.3-20.2 trillion USD due to LUCC [2] whereas, in the Tibetan Plateau, ESV increased from 1985 to 2000 by $67.10 × 10 8 [54] but decreased by $9.30 × 10 8 from 2000 to 2010 [55] predominantly due to climate warming and socio-economic development [56]. From the ESV related studies in the three major river basins, the KRB and Gandaki river basin have had positive changes in the ESV, whereas in Koshi river basin, ESV had decreased. Substantial LUCC, compared to the KRB, in Gandaki and Koshi river basin have led to considerable ESV changes in the basins. The ESV of Gandaki river basin had an increase of $1.68 × 10 8 from 1990 to 2015 [21], mainly due to an increase in agriculture, forest, waterbody, and wetlands areas, whereas in the KRB the ESV has increased primarily due to conversion of snow/ ice to shrub/grassland category. The ESV of Koshi river basin decreased by $2.05 × 10 8 from 1990 to 2010 [22] primarily due to urbanization, deforestation, and land reclamation, which was quite similar to the ESV change in the KRB due to deforestation and agricultural land encroachment for built-up area. Also, KRB, which lies in the western part of Nepal, has the least socio-economic development as compared to the other two basins [48]. The study in Gandaki river basin used datasets of twotime periods (1990 and 2015) [21] and Koshi river basin used datasets of 1990 and 2010 [22] for the ESV assessment using benefit transfer approach.
However, the benefit transfer approach used in this study and other basins of Nepal, for the assessment of ESV has some limitations as it is static evaluation method [57,58]. This method negates the spatial and temporal differences of different ecosystem types and quality, which lacks to reflect the dynamic changes in the ecosystem function [58,59]. Additionally, considering land cover change and land use intensification Research Article SN Applied Sciences (2022) 4:137 | https://doi.org/10.1007/s42452-022-05022-y separately could better facilitate the impact of climatic and anthropogenic drivers on ecosystem services as they both exert different and distinguishable effects on ESV [60]. Furthermore, benefit transfer approach does not take management practices into account that influences ecosystem services and functions [60] and thus, can be used simply to rank the importance of land use and land cover based on their contribution to the total ESV but not for assessing robustness and sensitivity of the ESV coefficients [61]. Thus, ESV assessment requires application of meta-analysis results to benefit transfer approach or regression functions of individual service value [62] and models incorporating valuation databases that reflect the dynamic changes of ecosystem functions in different spatio-temporal scale [63] as well as disentangling the ramifications of land cover changes and land use intensification.

Conclusion
This study investigated the decadal changes of seven types of land use and land cover in the transboundary Karnali River Basin during 1990-2020 using geospatial techniques and the effects of LUCC on the ESV were evaluated. The results showed that the forest area, waterbody, and snow/ice areas have decreased, whereas shrub/grassland, agriculture, bare area, and built-up areas have increased in the study area during the past 30 years. These changes lead to a net increase in ESV of the basin by ~ 2.7 × 10 6 USD from 1990 to 2020. Further, there was a decrease in ESV for the forest, waterbody, and snow/ice categories, whereas ESV of shrub/grassland, agriculture, and the bare area has increased. The decadal change analysis showed a continuous increasing or decreasing trend for a particular LC class except for agriculture in the KRB. It had an increasing trend for the first two decades and decreased in the last decade due to agricultural land abandonment in the basin. The natural and anthropogenic factors mainly drive the LULC change; hence, regular monitoring with high-resolution data yields better classification results, and assessing ESV can be a tool to facilitate for better provisioning and regulating the resources in the future. The use of ESV coefficients derived from literatures is a limitation of the study and is recommended to use ecosystem services evaluation models which can simulate the dynamic changes in the ecosystem.