Assessing land use/land cover change detection of north-eastern watersheds of Kashmir valley using GIS and remote sensing techniques

Although the potential of satellite-based data as a platform for creating important information for land use/land cover (LULC) has been widely recognized since the mid-seventies, the potential of satellite-based data as a platform for creating important information for LULC has yet to be fully realized (Alphan 2003; Erener et al. 2012; Vorovencii et al. 2013; Badreldin & Goossens 2014; Hütt et al. 2016; Onim et al. 2020). Over the past few decades, land use/land cover mapping and change detection techniques have been developed and deployed all over the world (Mucher et al. 2000; Mengistu & Salami 2007; Sahebjalal & Dashtekian 2013; Agaton et al. 2016; Mishra et al. 2020; Abualtayef et al. 2021). Various land use categories are rapidly replacing land cover around the world (Verma & Raghubanshi 2019). When addressing land use changes on the Earth's surface, the notions land use and land cover are frequently interchanged. Understanding LULC and its consequences is a prerequisite for natural resource planning and management (Garedew et al. 2009; Tallis & Polasky 2009; Melese 2016; Meraj et al. 2021). According to these researchers, land use has a huge impact on the functioning of socioeconomic and environmental systems, with significant tradeoffs for sustainability, food security, biodiversity, and the socioeconomic vulnerability of people and ecosystems.

Land-use patterns are influenced by ecological conditions, altitudes, geological structure, and slope, as well as technological, social, and institutional factors (Yin et al. 2010; Lavorel et al. 2017; Liang et al. 2020). The LULC has evolved dramatically as a result of population growth, industrialization, and urbanization. Even if changes in land cover due to land use do not always imply degradation of the land, land-use change is one of the most important causes of global change. This has an impact on biodiversity, water, and the radiation budget, among other aspects of geoenvironmental and natural ecosystems. Climate, biogeochemical cycles, energy fluxes, and people's livelihoods are all affected by changes in land cover condition and composition (Wang et al. 2012; Samie et al. 2020). As a result of rising population and urbanization in India's Himalayan states, large areas of forest cover are being converted to various land uses, resulting in considerable soil erosion (Mishra et al. 2020) Recent analysis, however, indicates that forest cover is increasing in numerous Himalayan locations (Mondal & Zhang 2018). India is rated ninth among countries with the biggest annual forest area change (2010–2015) according to the Forest Survey Report, 2017. Because rapid erosion caused by poor land-use patterns in the India particularly Kashmir Himalaya contributes to deadly floods on the plains, sustainable land use is vital not only for the 115 million Himalayans who live there, but also for the millions more who dwell in the surrounding Indo-Gangetic lowlands (Rather & Farooq 2018; Yousuf et al. 2018; Meer & Mishra 2020; Saleem et al. 2020).

In this study, a multi-temporal Landsat-5 Thematic Mapper satellite images were used to map the evolving pattern of LULC over a decade in four major watersheds of north-eastern Himalayas of Kashmir valley. The biological conditions, elevations, geological structure, and slope all play a role in determining LULC in the studied watershed. Aside from the aforementioned characteristics, the LULC pattern is likely to be influenced by technological, socioeconomic, and institutional setup (Ranagalage et al. 2019; Srivastava & Chinnasamy 2021). With the increasing scope of anthropogenic change and its effects on the ecosystem, having a land resources inventory of a watershed has become critical. Agriculture and related sectors employ approximately two-thirds of Kashmir Valley's workforce, despite the fact that just 21 percent of the state's land is suitable for cultivation. The research region includes four of Kashmir's highly inhabited agrarian watersheds, with a wide range of geophysical and socioeconomic characteristics (Mir et al. 2021; Sarah et al. 2021; Gull & Shah 2022). Farming in the study watersheds is characterized by a strong traditional agricultural system, horticulture, and lately added floriculture. In past few decades, the area under agroforestry in Kashmir Valley expanded swiftly (Rafeeq et al. 2020). Changes in land use of Kashmir watersheds from unirrigated system (forests, barren land) to irrigated system, which consists of a variety of agricultural and horticultural activities, are extremely prevalent due to improved crop production and economic gain. Many studies of LULC in the Indian Himalayas have been conducted (Ritse et al. 2020; Singh & Pandey 2021; Chakraborty & Saha 2022), but hardly any attempts have been made to study the changes in the available land resources of the study watersheds.

The primary objective of this study is to evaluate the geographical and temporal variation of LULC in four significant watersheds of the Kashmir valley's north-eastern Himalayas from 2003 to 2013. This research may provide a substantial contribution to the understanding of Kashmir's mountain ecosystem's LULC pattern and trend change. It could be a useful tool for researchers and decision-makers who deal with the environment on a global scale.

Lidder watershed has Pir Panjal range in the south and south-east, the north Kashmir range in the north-east, and the Zanskar range in the south-west and is situated between latitudes 33°40′N and 34°20′N and longitudes 75°00′E and 75°30′E. With a mean height of 3,169 metres and a standard deviation of 872 metres, the geographical elevation of the watershed ranges from 1,425 metres to 5,187 metres above mean sea level. It covers 1,097 km² and has unique climatic features due to its geophysical setting, which is surrounded on all sides by steep mountain ranges. The Lidder River, which originates in the Kolhoi glacier, runs through the heart of the watershed. The river runs south until it reaches Pahalgam, when it is joined by a significant tributary that originates in Sheshnag Lake. The Lidder River then flows largely west till it joins the Jhelum near Mirgund in Khanbal.

With an area of roughly 1,287 km² and latitudes of 34°00′ N and 34°30′ N, and longitudes of 74°40′ E and 75°30′ E, the Sindh watershed is one of the primary watersheds of the north-eastern Himalayas of Kashmir Valley. With a mean height of 3,397 metres and a standard deviation of 770 metres, the geographical elevation of the watershed ranges from 1,572 metres to 5,382 metres above mean sea level. The Sindh River, which originates in the Machoi glacier, runs through the heart of the watershed. The river runs south until it reaches Domail, where it is joined by a tributary that originates from the Kolhoi glacier. Before joining the Jhelum near Shadipora, the Sindh River travels largely west, passing through various streams such as the Amarnath, Gund, Wangath, and Surfraw.

The Arin watershed, lying between latitudes 34°20′ N and 34°30′ N and longitudes 74°40′ E and 74°55′ E, is one of the minor watersheds in the north-eastern Himalayas of Kashmir Valley. With a mean height of 3,083 metres and a standard deviation of 698 metres, the geographical elevation of the watershed ranges from 1,570 metres to 5,054 metres above mean sea level. It covers an area of 202 km² and, due to its diverse topography, has peculiar climatic features. The Erin River obtains its water from the drainage of numerous streams, which eventually unite near Isrur Tar and empty into Wular Lake.

The Madhumati watershed is located between latitudes 34°25′ N and 34°35′ N, and longitudes 74°30′ E and 74°55′ E, on the southern edge of the Arin watershed. With a mean height of 3,028 metres and a standard deviation of 609 metres, the geographical elevation of the watershed ranges from 1,589 metres to 4,496 metres above mean sea level. It covers 358 square kilometers and has the same climatic features as the Arin watershed. The Madhumati River rises near Sukhnai Gali pass from a tiny lake called Sar and runs west to Athwatoo, where it turns southwest and eventually enters the Wular lake. Winter and spring bring 75 percent of the region's precipitation, which comes in the form of snow and rain, respectively. Because of complicated geography of these watersheds, the area's minimum and maximum temperatures vary greatly throughout the year.

The LULC classes of the Lidder, Sindh, Arin, and Madhumati watersheds were mapped using multi-temporal Landsat-5 Thematic Mapper (TM) imageries from 2003 and 2013 (Table 1). Forest, agriculture, marine, inland water resources, and LULC mapping and monitoring are the principal applications for the TM sensor. Watershed boundaries were also created using Aster DEMs. All of the photographs were obtained from the United States Geological Survey's Earth Explorer website (http://earthexplorer.usgs.gov/). Landsat 5 TM level 1 Top-of-Atmosphere (TOA) radiance datasets were available, and they were projected in the UTM Zone 43 projected coordinated system.

The USGS Earth Explorer website was used to download multi-temporal Landsat-5 TM imageries from 2003 and 2013. Seven spectral bands in a Geotiff format make up Landsat 5 TM, which has a spatial resolution of 30 m. The study region was extracted using picture sub-setting and layer stacking, which combined all the image bands into a single image. The maximum likelihood classifier (MLC) was used for supervised classification. With the aid of true and false colour composites, a signature file reflecting the various LULC classifications was first produced. The categorized LULC maps also underwent an accuracy assessment. A collection of reference pixels that correspond to the geographic points in the categorized image are needed for accuracy assessment (Radoux et al. 2011).

ASTER DEMs were used to extract the borders of the Lidder, Sindh, Madhumati, and Arin watersheds. The Semi-Automatic Classification Plug-in in QGIS software was used to convert satellite images from TOA to bottom-of-atmospheric reflectance (BOA) (Chapa et al. 2019). After that, the watershed border derived from the ASTER DEM was used to subset all of the images, and all five bands of satellite imagery were stacked. A typical synthetic colour composite of both periods was created for mapping (Camilleri et al. 2017).

A classification method that defines the LULC classes was considered while creating the LULC map using satellite images. The number of land-use classes desired is determined by the needs of a certain project for a specific application (Spruce et al. 2018). Water; scrubland and grasses; river beds; non-perennial snow, horticulture; glaciers; forests; exposed rock masses; built-up; barren land; agriculture were chosen as the 11 key LULC classes for mapping the entire study area. Because tree cover in agroforestry systems had a comparable spectral response to orchards in Kashmir, the area under agroforestry was classified as horticulture, while the area under irrigated crops was classified as agriculture.

For land-cover classification, there are two image classification techniques that are used. A method of land-cover type classification employing sample polygons (ground truth points) from known land-cover types is known as supervised classification (Boryan et al. 2011). Unsupervised classification, on the other hand, is a sort of land cover classification from satellite image data in which the user has no idea how many land-cover classes are present in the field (Rozenstein & Karnieli 2011). The supervised form of classification was used in this investigation, with known ground truth points and images synchronized using QGIS and Google Earth. Following the creation of the classification scheme, all of the LULC classes were mapped using one of the most extensively used image classification techniques, maximum likelihood classification. The Image Classification toolbar's training sample drawing tools are used to construct training examples. The picture layer needs to be a segmented raster layer, and there are four drawing tools available: polygons, circles, rectangles, and select segment. An area that belongs to a recognized class is defined in the map display. A training sample is created using a drawing tool. Once the training sample has been drawn or chosen, a new class in Training Sample Manager is created with a default name, value, and colour that can later be altered. To produce a few more training samples for the other classes in the image, the processes are repeated. The list of training samples is then maintained by evaluating and managing these training samples in the Training Sample Manager dialogue box. Before selecting training samples, a detailed empirical investigation of satellite images, Google Earth images, and the toposheet of the watershed was conducted. A minimum of 100 training samples was required in the majority of the classes. Due to the heavy canopy of woods along the river channel and a paucity of water in the river channels, selecting training samples for water was problematic due to the imagery being acquired in mid-November with most of the rivers in the highlands carrying less water than during the monsoon season.

The term ‘accuracy’ is often used to represent the degree of ‘correctness’ of a map or classification in thematic mapping from remotely sensed data. If a thematic map accurately depicts the land cover of the region it depicts, it can be said to have been created using categorization. The degree to which the resultant image categorization agrees with reality or complies with the ‘truth’ is essentially what is meant by classification accuracy (South et al. 2004). Several factors may lead to an accuracy assessment. It could be done, for example, to give a general assessment of a map's quality, to serve as the foundation for comparing various categorization systems, or to try to comprehend error.

Ground control points from Google Earth and Survey of India toposheets were used for the accuracy evaluation, while field observations and location surveys were employed to verify and corroborate the facts on the ground. With the use of GPS and local guides, a field study was done in several regions of the study watersheds, spanning all LULC classes, to verify suspicious spots on the ground. Few regions were inaccessible due to the rugged geography, rough terrain, and steep hills. These points were used to determine the overall accuracy and the Kappa statistic for each of the categorized images.

A post-classification change detection approach was used to examine the changes. In recent decades, many change-detection methodologies have been developed, including image differencing, post classification change matrix, comparison methodology, and principal component analysis (Lu et al. 2004; Al-doski et al. 2013). A Comparison methodology was built from 2003 to 2013 to assess the overall changes in land-use classifications between 2003 and 2013. In the comparison methodology, the total increase or reduction of each class of land cover is assessed as a percentage of the original area, and change in each class of land cover is determined.

To test the correctness of the 2013 categorized map, one of the most commonly used Kappa accuracy assessment techniques was chosen. Planet scope satellite imagery with a high resolution (3.9 metres) was used to assess accuracy. We created 400 stratified random points for each watershed with a minimum of 20 points in each class. The accuracy of the 2003 photos could not be assessed due to a lack of availability and clear Google Earth images. The overall classification accuracy for the classified image of 2013 was found to be 81.63 percent for Madhumati watershed, 79.25 percent for Arin watershed, 82.34 percent for Sindh watershed, and 77.25 percent for Lidder watershed with a Kappa statistic of 0.8012, 0.7862, 0.8131, and 0.7643 respectively. The results of the LULC change matrix demonstrate significant land-cover changes in the studied region for the study period. Table 2 displays the values for the Kappa statistic and total classification accuracy for the classified images of 2013 of different watersheds. Tables 3,⁴⁵–6 show the spatial distribution pattern of LULC as determined by supervised classification, as well as the comparison methodology.

This study used Landsat-5TM to examine and track the changes in LULC pattern in the key watersheds of the north-eastern Himalayas of Kashmir Valley from 2003 to 2013. The findings of this study show that there have been numerous environmental problems in the study region, such as the unplanned urban and industrial growth. Due to the increase in population density and quick urbanization, encroachment usually affects water bodies, scrublands, and forest lands. The fundamental issue in the region is the transformation of open forests into urban zones. Due to urban growth and hotel construction throughout the past two decades, considerable changes in built up area were recorded around the periphery of the Lidder River and Sindh River, both of which are popular tourist sites. Because forest cover is the most common land use, its quality (soils) and amount (area) are inextricably linked to the nature of landforms. The surface runoff and sediment from the area may rise as a result of increased runoff coefficients due to decrease in forest cover and scrublands. This study discovered that agricultural, horticultural, and built-up activities are on the rise, necessitating greater attention to soil protection and efficient use of overland water resources.

This information would eventually aid in the identification of finite resources as well as environmentally significant places that may be designated as hotspots for conservation or repair. On the one hand, resource evaluation would lead to a better understanding of the effects of various developmental activities on these resources, and on the other hand, the planning process. The current study makes use of the RS and GIS approach, which is currently one of the most popular technologies for spatio-temporal analysis that is not achievable with other traditional mapping techniques. The study's findings and analysis have substantial policy implications for sustainable land-use/cover practices in Kashmir's north-eastern Himalayas.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.