Geospatial Assessment of Land Use/Land Cover Dynamics and Future Predictions Using Markov Chain Cellular-Automata Simulations in Rajouri District of Jammu and Kashmir, India

Abstract

Land use/land cover (LULC) change significantly influences a range of environmental and socio-economic issues, including climate change, deforestation, biodiversity loss, soil degradation, ecosystem services, and food security, at local, regional, and global levels. In the northwestern Himalayan region, particularly in Rajouri district of Jammu and Kashmir (J&K), LULC change has profound environmental and socio-economic implications. Understanding the temporal and spatial dimensions of LULC change is crucial for assessing the impact of human activities on the region’s environment. The present study aimed to analyze LULC change in Rajouri district of J&K, India over a 30-year period from 1990 to 2020 and to project future LULC dynamics for the next 30 years up to 2050. Landsat imagery with a supervised classification technique was used for classification and generation of LULC maps. Moreover, CA Markov model was used to predict the future LULC status of the area. The model validation exhibited strong performance, with Kappa statistics exceeding 0.90, indicating a high level of reliability in the projections. The results indicate considerable changes in different land use classes from 1990 to 2020. Over the 30-year period, dense forest showed the maximum reduction of about −20.69 Km², followed by open forest (−15.87 Km²) and grassland (−13.75 Km²). Wasteland showed the maximum increase of about +28.24 Km², followed by built-up (+17.90 Km²) and cropland (+12.50 Km²). The cumulative impact of deforestation from 1990 to 2020 amounts to approximately 43.17 Km², while afforestation efforts only managed to reclaim 6.61 Km² of land. The future prediction using the CA Markov model suggests further changes in LULC patterns, with built-up, cropland, and wasteland projected to increase exponentially by 2050, accompanied by sharp declines in forests. Therefore, policymakers should prioritize sustainable land management and forest conservation strategies to mitigate the potential negative impacts of LULC changes on the environment, ensuring balanced and sustainable development.

1. Introduction

Land serves as an essential natural asset, holding significant economic, social, and biophysical importance [1]. Although often used interchangeably, ‘land use’ and ‘land cover’ carry distinct meanings. Land cover refers to the physical features of the Earth’s surface, encompassing vegetation, water, soil, and other natural elements. Conversely, land use denotes the specific human activities and purposes applied to these physical features like urbanization, infrastructure development, agriculture, etc. [1,2]. Land use land cover (LULC) change refers to the inter-conversion of different land use types and arises from complex interactions between humans and the physical environment [3]. Globally, LULC is continuously changing, particularly accelerated by rapid economic development [4]. This change is a foremost ecological concern of global consequence due to its substantial effects on biogeochemical cycles, climate, and biodiversity [5,6]. While changes in land use often result in economic benefits for humans, they simultaneously jeopardize the stability of the natural environment [7]. The majority of these changes occur without careful planning and manifest as environmental degradation, water scarcity, and swift urbanization [6,8]. Indeed, changes in land use not only modify the spatial extent and dimensions of various LULC classes but also trigger a cascade of events that can lead to the degradation of multiple ecosystems [8,9]. Such changes can heighten susceptibility to landslides and flooding, while also creating conditions conducive to the transmission of infectious diseases. Moreover, these alterations may lead to increased sedimentation, accelerated soil degradation, disturbances in hydrological regimes, and substantial declines in biodiversity at both local and global scales [10]. LULC changes in mountainous regions have far-reaching implications due to their fragile ecosystems, variable climate, and complex terrain, making them highly sensitive to environmental changes. The Himalayan region, in particular, has witnessed significant LULC changes in recent decades, resulting in land becoming less productive or entirely unproductive. This trend poses a considerable challenge to the region’s sustainable development [11]. Hence, a comprehensive understanding of processes and implications of LULC change is fundamental for advancing sustainable land management, mitigating environmental risks, and promoting balanced interactions between human systems and the natural environment [12].

Traditional ground verification and the remote sensing (RS) technique are the methods available to detect LULC changes. However, the ground verification method is time consuming, manpower intensive and it is difficult to collect data from areas which have rugged topography and are inaccessible like most of the areas in the Indian Himalayas [13]. On the other hand, RS-based techniquesare the most efficient means to overcome these shortcomings as they can provide data relating to the spatial as well as temporal pattern of LULC over a large geographical area in an ordinary and reliable manner. Recent advancements in space technology have led to significant improvements in the quality and accessibility of remotely sensed data. Nowadays, RS and geographic information systems (GIS) have become indispensable tools for acquiring accurate and timely spatial data of LULC [13,14]. RS imagery from various earth observation satellites such as the Landsat series of the United States serves as a rich data source for extracting, analyzing, and simulating LULC changes over time [15,16]. Meanwhile, GIS provide a versatile platform for the acquisition, storage, visualization, and analysis of digital data crucial for detecting changes [17]. Consequently, the fusion of RS methods with GIS technology has become a standard practice for delineating and characterizing diverse land use patterns across varying time frames, significantly facilitating the mapping and classification of LULC in a study area [18,19].

Beyond mapping and monitoring, land use change models have increasingly evolved into valuable decision-support instruments, enabling the quantification and analysis of landscape transformations, projection of future trends based on historical trajectories, and validation of simulated scenarios against observed patterns. Such modelling approaches also provide a scientific basis for formulating sustainable development strategies [20].

Among these approaches, the CA-Markov model combining cellular automata (CA) with the Markov chain has been widely applied for robust assessment and prediction of LULC dynamics [21,22]. By integrating remote sensing datasets within a GIS environment, the model generates spatially explicit and geographically consistent outputs suitable for detailed LULC investigations [23]. The Markov chain component estimates transition probabilities based on prior land cover states, while the CA module allocates these transitions spatially according to neighborhood configuration and proximity rules [24]. Owing to this complementary structure, the CA-Markov framework has been successfully implemented in numerous regional studies to simulate future LULC scenarios [24,25,26,27].

Given the persistent anthropogenic pressures on land systems and the strong interdependence between ecosystem processes and human activities, intensified LULC dynamics are anticipated in the coming decades [28,29]. Consequently, evaluating future LULC scenarios is essential for informed planning, sustainable resource management, and the design of appropriate mitigation and adaptation strategies [30].

Various studies in different parts of the world have focused on elucidating the consequences of LULC alterations, such as their profound impacts on resource productivity, climate dynamics [31], soil integrity [32], biodiversity conservation [33], ecosystem services [34,35,36] and even public health concerns [37]. Concurrently, there has been a concerted effort to identify the drivers propelling LULC changes [36,38]. However, before elucidating the consequences of LULC changes, conducting a quantitative analysis of the processes and trends associated with LULC transitions is prerequisite for acquiring a deeper understanding of LULC dynamics. Such analyses serve as a foundational step in aiding policymakers in delineating targeted improvement objectives for specific regions and implementing appropriate strategies. In this regard, LULC mapping and temporal assessments function as indispensable tools for sustainable development, while ensuring alignment with broader sustainability initiatives and fostering harmonized progress across multiple sectors [1,39].

Over time, numerous studies have been conducted in the Himalayan region, particularly in the Kashmir Himalaya, to assess and map LULC changes [8,40,41,42,43,44,45,46,47,48,49]. However, despite these efforts, the Pir Panjal region in the western Himalayas, specifically Rajouri district in the Indian union territory (UT) of Jammu and Kashmir (J&K), has remained relatively unexplored in terms of comprehensive LULC analysis considering topographical aspects. This region faces various ecological challenges including deforestation, forest fires, land degradation, soil erosion, grazing pressure, and agricultural expansion, etc.

Therefore, to address this research gap, this study aims:

(i)

To evaluate and map the patterns of LULC changes over a period of 30 years from 1990 to 2020 in Rajouri district, J&K, India.

(ii)

To predict LULC changes for the next three decades, up to 2050.

(iii)

To analyze deforestation and afforestation patterns based on LULC analysis within the district.

The present study provides updated and reliable LULC datasets that improve understanding of long-term land use dynamics in District Rajouri. These datasets can support sustainable land management initiatives such as forest conservation, afforestation programs, watershed management in hilly terrains, and regulation of urban expansion. By identifying areas of forest decline, built-up expansion, and agricultural transformation, the study offers spatially explicit information that can assist planners and policymakers in prioritizing conservation efforts, mitigating land degradation, and supporting climate change adaptation strategies in the northwestern Himalayan region.

2. Materials and Methods

2.1. Study Area

The present study was conducted in Rajouri, a hilly district located in the Indian UT of J&K (Figure 1). Rajouri is characterized by rugged terrain with moderate to very steep slopes, situated at the foothills of the snow-covered Pir Panjal Mountains in the northwestern Himalayas. Encompassing an area of 2630 Km², it is bordered by the Line of Control to the west, Poonch district to the north, Reasi district to the east, and Jammu district to the south. The district is divided into 13 tehsils and is known for its fertile, mountainous land. Maize, wheat, and rice are the primary crops cultivated in the region, with irrigation primarily sourced from the Tawi River originating from the Pir Panjal Mountains. According to the 2011 census, Rajouri has a population of 642,415 individuals, with a population density of 240 individuals per square kilometer and a sex ratio of 860 females per 1000 males, compared to the UT’s sex ratio of 889 females per 1000 males (http://rajouri.nic.in, accessed on 5 April 2024). The predominant LULC types in the district include agricultural land, built-up, forest land, and grasslands etc. [50]. The vegetation in the area ranges from sub-tropical to alpine (https://forest.jk.gov.in, accessed on 5 April 2024).

2.2. Data Source

Landsat satellite imagery was used as the primary data source for studying the LULC pattern and changes. The satellite images for the year 1990, 2000, 2010 and 2020 were downloaded from United States Geological Survey (USGS) Earth Explorer (https://earthexplorer.usgs.gov, accessed on 20 October 2023) (

Table 1. Details of the satellite images used in the present study.

Satellite	Landsat Scene Identifier	Sensor	Acquisition Date	Cloud Cover (%)	UTM Zone	Datum	Resolution	Path/Row	Source
Landsat 5	LT51490371990123ISP00	Thematic Mapper (TM)	3 May 1990	6.00	43	WGS84	30 m	149/37	USGS Earth Explorer
Landsat 5	LT51490372000135AAA02	Thematic Mapper (TM)	14 May 2000	2.00	43	WGS84	30 m	149/37	USGS Earth Explorer
Landsat 7	LE71490372010122SGS01	Enhanced Thematic Mapper (ETM)	2 May 2010	2.00	43	WGS84	30 m	149/37	USGS Earth Explorer
Landsat 8	LC81490372020142LGN00	Operational Land Imager and Thermal Infrared Sensor (OLI/TIRS)	21 May 2020	1.44	43	WGS84	30 m	149/37	USGS Earth Explorer

).

2.3. Workflow

The overall workflow followed in the present study is summarized in Figure 2. A detailed description of the methodology followed at each stage is provided in the subsequent sections.

2.3.1. Image Pre-Processing

Raw satellite images are prone to systematic and random errors, making them unsuitable for direct feature identification and applications. Thus, in this study, image corrections were conducted using ArcGIS software (version 10.8), following established methods described in the literature [51]. Specifically, standard image processing techniques such as extraction, layer stacking, radiometric correction, atmospheric correction and geometric correction/georeferencing [52] were applied to the four Landsat images utilized in the study. To ensure consistency among multi-temporal datasets acquired from Landsat 5, Landsat 7, and Landsat 8, Digital Numbers (DN) were converted to Top-of-Atmosphere (TOA) reflectance using sensor-specific radiometric rescaling coefficients provided in the metadata (MTL) files following standard USGS guidelines [53,54]. Atmospheric correction was subsequently performed using the Dark Object Subtraction (DOS) method to minimize atmospheric scattering effects [53]. To maintain spectral comparability across sensors, only corresponding reflective bands including Blue, Green, Red, Near-Infrared (NIR), and Shortwave Infrared 1(SWIR1) were selected, while thermal bands were excluded due to differences in spectral range and radiometric resolution. All images were georeferenced to the Universal Transverse Mercator (UTM) projection, employing the World Geodetic System (WGS) 84 datum, zone 43 N, to ensure accurate spatial alignment and compatibility with other geospatial datasets.

2.3.2. Land Use Land Cover Classification Scheme

Following the classification scheme established by the National Remote Sensing Centre (NRSC) Hyderabad, India, nine land use and land cover (LULC) categories were specifically selected for analysis, taking into consideration the characteristics of the study area. The classes encompass ‘Agriculture Plantation’, ‘Built-up’, ‘Cropland’, ‘Dense Forest’, ‘Grassland’, ‘Open Forest’, ‘Snow Cover’, ‘Wasteland’, and ‘Water Bodies’(Figure 3). The detailed description of the LULC classes has been summarized in

Table 2. LULC classification scheme.

S. No.	Class	Description
1	Agriculture Plantation	Found surrounding various agricultural plots. It mainly includes area under agroforestry and some horticulture area.
2	Built-up	Area designated for settlements and commercial purposes, encompassing buildings, houses, transportation networks, and industrial/manufacturing zones.
3	Cropland	Area utilized for cultivating various crops such as maize, wheat, paddy, and other seasonal crops.
4	Dense Forest	Land characterized by a tree canopy density exceeding 40%.
5	Grassland	Land predominantly covered by herbaceous plants.
6	Open Forest	Land with shrubs and a tree canopy density ranging between 10% and 40%.
7	Snow Cover	Area covered by perpetual snow.
8	Wasteland	Area devoid of vegetation, including sediments, exposed rocks, landslide zones, and degraded forest areas.
9	Water Bodies	Includes mostly perennial rivers, small streams, and some alpine lakes within the study area.

.

2.3.3. Classification and Mapping of LULC

After the formulation of the classification scheme, the widely employed supervised classification technique with Maximum Likelihood (ML) algorithm was used to map all the LULC classes following previous studies [8,51,55,56,57]. In supervised classification, image pixels are classified by an image analyst using a specific algorithm and representative training sites for each land cover type are generated, which serve as reference samples. These samples are used to generate spectral signatures that represent the statistical properties of the classes of interest [51]. The ML algorithm, widely recognized for its robustness in land cover studies [58], classifies image pixels based on the probability that they belong to a particular class, assuming a normal distribution of training data [59].

In the present study, training sites were carefully selected to enhance classification accuracy, following a detailed examination of satellite imagery and cross-verification with Google Earth data [55,60,61]. Google Earth integrates high to very high-resolution satellite and aerial imagery from multiple providers, with spatial resolution varying by location and zoom level [62]. The imagery available for the study area was substantially finer than the 30 m Landsat data used for classification. Various band combinations derived from the selected reflective bands (Blue, Green, Red, NIR, and SWIR1) were generated to enhance class separability. Infrared composite images were used to capture vegetation heterogeneity, while NIR-Red combinations were applied to distinguish built-up and barren surfaces. The training samples were compiled into a signature file, forming the basis for supervised classification. All LULC classification and mapping procedures were performed in ArcGIS 10.8.

2.3.4. Field Survey and Accuracy Assessment

A thorough field survey was conducted to validate uncertain areas, utilizing the global positioning system (GPS) across various parts of the district, covering diverse LULC classes. The field survey also involved capturing photographs depicting different land use classes and human-induced activities in the surveyed areas.

The accuracy of classified maps of 1990, 2000, 2010, and 2020 was evaluated in ArcGIS using Google Earth imagery as reference data to validate the supervised classification [60,63]. A stratified random sampling approach was applied for accuracy assessment. Standard evaluation metrics, including the producer’s accuracy, user’s accuracy, overall accuracy, and the kappa coefficient, were employed to quantify classification performance [64]. A minimum of 50 random points for each LULC class were generated on each classified map, and the results were analyzed through an error matrix [63,65]. Producer’s accuracy indicates the proportion of reference pixels of a class that are correctly mapped, while user’s accuracy represents the likelihood that a pixel classified into a given category actually corresponds to that class on the ground. Overall accuracy reflects the proportion of correctly classified pixels in the image. These measures correspond to omission and commission errors in image classification [64].

The Kappa values are categorized into three groups based on the agreement between the data represented on the classified maps and the actual information on the ground. A value of less than 0.40 indicates poor agreement, while values falling between 0.40 and 0.80 indicate moderate agreement. A Kappa value exceeding 0.80 depicts strong agreement [66,67].

2.3.5. LULC Change Detection

To highlight the changes that have occurred between different time periods, change transformation matrices were created [55]. These matrices depict the land cover changes in each decade, derived from classified images spanning from 1990 to 2000, 2000 to 2010, and 2010 to 2020. Additionally, a change matrix was generated from 1990 to 2020 to evaluate the overall changes in LULC classes between 1990 and 2020.

The matrix tables themselves illustrate the areas that have changed, remained unchanged, and the total change in each class. These change matrices provide valuable details about the spatial distribution of changes in LULC classes [68].

The deforestation and afforestation rates were examined by calculating the change rate for forest covers (dense and open forest area) and calculating the patterns of forest covers losses and gains by following Hu et al. [1].

2.3.6. Prediction of Future LULC Changes Using the Integrated CA-Markov Model

The CA-Markov method within the TerrSet software (v19.06) package was utilized to generate a future LULC map of the study area for the year 2050, aiming to predict LULC changes in the area over the next 30 years (i.e., from 2020 to 2050). This approach combines cellular automata (CA) and Markov chain modelling to account for both spatial structure and geographic directions of LULC changes [69]. Markov chain models describe the probabilistic movements between land states and calculate rates of transfer among various land use types [1].

The Markov model can be represented as a discrete stochastic process consisting of a finite set of states, S = {S₀, S₁, S₂, S₃…S_n} where each state corresponds to a specific LULC class. At any given time t, the system is in state S_t, and it transitions to another state S_j at the subsequent time step with a transition probability denoted by P_ij [17,45,69]. The transition probabilities are arranged in a transition probability matrix that quantifies the likelihood of change from one LULC class to another over a specified time interval. Accordingly, the state of the system at time t + 1 is determined solely by its state at time t, reflecting the memoryless (Markovian) property of the process. This relationship is mathematically expressed as [17]:

$$P_{ij}=\left[\begin{array}{ccc}{p}_{11}& p_{12}\dots & p_{1n}\\ p_{21}& p_{22}\dots & p_{2n}\\ ⋮& ⋮& ⋮\\ p_{n1}& p_{n2}\dots & p_{nn}\end{array}\right]$$

$$(0\le P_{ij}<1 \mathrm{a}\mathrm{n}\mathrm{d}\sum _{j=1}^n{P}_{ij}=1,i,j=\mathrm{1,2},\dots ,n)$$

$$S_{t+1}=P_{ij}\times S_t$$

where P_ij is the state transition probability matrix and n is the land use type number; S is land use status, t or t + 1is the time point.

However, Markov chain modelling alone overlooks spatial distribution and directionality in LULC changes [70]. Therefore, the CA-Markov model was employed, as it integrates spatial contiguity and the geographic distribution of land use classes into the Markov chain framework [71,72]. This hybrid approach enhances the realism of simulated LULC patterns by incorporating neighborhood effects and spatial dependency into the modelling process. The cellular automata (CA) component simulates transitions among multiple LULC categories based on spatial proximity, which is a critical determinant of land transformation processes [69].

CA introduces spatial dependency by considering neighborhood interactions, whereby the future state of a cell is influenced not only by its current state but also by the states of surrounding cells [17]. In the CA framework, the state of a cell at time t + 1 is expressed as a function of its current state, and the neighborhood configuration:

$$S_{t+1}=f(S_t,N)$$

where S represents the state of the cell at time t, N denotes the neighborhood configuration, and f is the transition rule function.

Following previous studies [60,61,73] the modelling process was carried out in sequential steps within the TerrSet sofware (v19.06), beginning with the preparation of input maps and followed by transition probability estimation and spatial allocation. The classified LULC maps of 1990 and 2020 were used as inputs. First, the Markov transition estimator was applied to compute the transition probability matrix and transition areas, indicating the likelihood of each land cover class changing into another by 2050. The 2020 LULC map, together with the transition probability data, served as the base for spatial simulation. The CA component incorporated neighborhood contiguity and spatial proximity rules to allocate the predicted transitions across the landscape. Using this framework, the CA-Markov model generated the projected LULC map for 2050, representing potential changes over the 30-year period.

2.3.7. Validation of CA-Markov Model

To evaluate the reliability of the model and validate the prediction of LULC for 2050, the projected map for 2020 was compared with the actual LULC map of the same year, following the approaches of [69,74,75,76]. According to Pontius [77], the accuracy of simulation results can be assessed using Kappa statistics, which measure the degree of agreement in terms of both spatial location (Kappa for location) and categorical quantity (Kappa for quantity).

For this purpose, a simulated LULC map of 2020 was generated using the 1990 and 2010 LULC maps. This simulated map served as the comparison dataset, while the actual LULC map of 2020 was treated as the reference dataset. The simulated map represents the CA-Markov model output, whose performance was evaluated against the observed land cover conditions [74]. The ‘VALIDATE’ module in TerrSet was employed to compute Kappa statistics, quantifying the correlation between the simulated and reference maps [60,61,69]. Kappa values close to 1 indicate strong agreement, values around 0.75 reflect moderate agreement, and values near or below 0.5 suggest weak agreement [76].

Several forms of Kappa statistics were computed, including Kappa for no information (K_no), Kappa for grid-cell level location (K_location), Kappa for stratum-level location (K_{locationStrata}), and the standard Kappa (K_standard) [60,61]. These indices are mathematically expressed as:

$$K_{\mathrm{n}\mathrm{o}}={\displaystyle \frac{{{\mathrm{M}}_{\mathrm{m}}-\mathrm{N}}_{\mathrm{n}}}{{{\mathrm{P}}_{\mathrm{p}}-\mathrm{N}}_{\mathrm{n}}}}$$

$$K_{\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{n}\mathrm{d}\mathrm{a}\mathrm{r}\mathrm{d}}={\displaystyle \frac{{{\mathrm{M}}_{\mathrm{m}}-\mathrm{N}}_{\mathrm{n}}}{{{\mathrm{P}}_{\mathrm{p}}-\mathrm{N}}_{\mathrm{m}}}}$$

$$K_{\mathrm{l}\mathrm{o}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}}={\displaystyle \frac{{{\mathrm{M}}_{\mathrm{m}}-\mathrm{N}}_{\mathrm{n}}}{{{\mathrm{P}}_{\mathrm{m}}-\mathrm{N}}_{\mathrm{m}}}}$$

$$K_{\mathrm{l}\mathrm{o}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{S}\mathrm{t}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{a}}={\displaystyle \frac{{{\mathrm{M}}_{\mathrm{m}}-\mathrm{H}}_{\mathrm{m}}}{{{\mathrm{K}}_{\mathrm{m}}-\mathrm{H}}_{\mathrm{m}}}}$$

Here, the subscripts denote the level of information: n = no information, m = medium information and p = perfect information. The capital letters represent the type of information function at a given location x:N(x) = no information, H(x) = stratum-level information, M(x) = grid-cell level information, K(x) = perfect grid-cell information with imperfect stratum-level information, and P(x) = perfect grid-cell information across the entire landscape. In this context, x refers to a specific grid cell or location on the map.

Together, these indices provide information into the discrepancies between the simulated and reference maps, reflecting both differences in category allocation and the variation in proportional representation of classes across the landscape [76].

3. Results

3.1. Accuracy Assessment of LULC Maps (1990–2020)

The comprehensive accuracy assessment of the LULC maps in the study area for the years 1990, 2000, 2010, and 2020 yielded overall classification accuracies of 86.17%, 87.45%, 88.09%, and 90.85%, respectively. The user’s accuracy (UA) for 1990 ranged from 79.17% to 92.59%, for 2000 (79.25% to 96%), for 2010 (83.02% to 97.96%), and for 2020 (88.68% to 98%). Furthermore, the producer’s accuracy (PA) for 1990 ranged from 79.31% to 93.48%, for 2000 (77.59% to 93.75%), for 2010 (79.31% to 95.83%), and for 2020 (83.93% to 97.92%). The Kappa coefficient values, which indicate the agreement between the classified and reference data, were 0.84, 0.86, 0.87, and 0.90 for 1990, 2000, 2010, and 2020, respectively (

Table 3. Accuracy assessment showing user’s accuracy (UA), producer’s accuracy (PA) and Kappa coefficient for different time periods.

Class	1990		2000		2010		2020
	UA (%)	PA (%)	UA (%)	PA (%)	UA (%)	PA (%)	UA (%)	PA (%)
Agriculture Plantation	79.17	86.36	79.25	91.30	83.02	93.62	88.68	95.92
Built-up	87.04	92.16	88.24	91.84	95.83	86.79	96.00	90.57
Cropland	86.00	86.00	86.27	88.00	81.82	88.24	86.79	90.20
Dense Forest	82.14	79.31	83.93	85.45	83.05	90.74	87.27	90.57
Grassland	89.58	79.63	91.84	77.59	90.20	79.31	92.16	83.93
Open Forest	82.76	81.36	83.93	85.45	86.79	88.46	87.04	88.68
Snow Cover	92.59	90.91	96.00	88.89	97.96	84.21	98.00	90.74
Wasteland	88.68	88.68	90.57	87.27	84.62	88.00	90.57	90.57
Water Bodies	87.76	93.48	88.24	93.75	92.00	95.83	92.16	97.92
Overall accuracy (%)	86.17		87.45		88.09		90.85
Kappa Coefficient	0.84		0.86		0.87		0.90

).

3.2. Distribution of LULC Classes Across Different Time Periods

Table 4. Classified LULC statistics for the year 1990, 2000, 2010 and 2020.

LULC Class	Area in 1990		Area in 2000		Area in 2010		Area in 2020
	Km2	%age	Km2	%age	Km2	%age	Km2	%age
Agriculture Plantation	53.14	2.02	53.64	2.04	54.59	2.07	56.11	2.13
Built-up	54.01	2.05	56.84	2.16	62.06	2.36	71.91	2.73
Cropland	790.13	30.04	793.66	30.18	798.94	30.38	802.63	30.52
Dense Forest	959.03	36.46	954.05	36.27	946.55	35.99	938.34	35.68
Grassland	315.05	11.98	312.59	11.89	307.15	11.68	301.30	11.46
Open Forest	240.07	9.13	234.96	8.93	229.77	8.74	224.20	8.53
Snow Cover	44.59	1.70	43.05	1.64	41.13	1.56	38.77	1.47
Wasteland	148.47	5.65	156.36	5.95	166.47	6.33	176.71	6.72
Water Bodies	25.51	0.97	24.85	0.94	23.34	0.89	20.03	0.76
Total Area(Km2)	2630.00		2630.00		2630.00		2630.00

and Figure 4 show the distribution of LULC classes for four different years 1900, 2000, 2010 and 2020. Out of the total nine LULC classes, it was revealed that in 1990, the district was predominantly covered by dense forest (36.46%), followed by cropland (30.04%), grassland (11.98%), open forest (9.13%), and wasteland (5.65%). Other classes accounted for a smaller portion of the district, with built-up covering approximately 2.05%, agriculture plantation 2.02%, snow cover 1.70%, and water bodies 0.97% of the total area. Shifts in the LULC status occurred for subsequent assessment years (2000, 2010 and 2020) with dense forest, grassland, open forest, snow cover and water bodies showing reduction and agriculture plantation, built-up, cropland and wasteland showing increase. However, dense forest continued to be the dominant LULC class (36.27%, 35.99% and 35.68%, for the year 2000, 2010 and 2020 respectively) followed by cropland (30.18%, 30.38% and 30.52% respectively) grassland (11.89%, 11.68% and 11.46% respectively), open forest (8.93%, 8.74% and 8.53% respectively) and wasteland (5.95%, 6.33% and 6.72% respectively) (Figure 4).

3.3. LULC Change Pattern

3.3.1. Gains and Losses

LULC transformations from 1990 to 2020 revealed substantial shifts across various LULC classes of the district (

Table 5. Relative change of LULC for four different decades.

Class	1990–2000	2000–2010	2010–2020	1990–2020
	Change in Area (Km2)	Change in Area (Km2)	Change in Area (Km2)	Change in Area (Km2)
Agriculture Plantation	+0.50	+0.95	+1.52	+2.97
Built-up	+2.83	+5.22	+9.85	+17.90
Cropland	+3.53	+5.28	+3.69	+12.50
Dense Forest	−4.98	−7.50	−8.21	−20.69
Grassland	−2.46	−5.44	−5.85	−13.75
Open Forest	−5.11	−5.19	−5.57	−15.87
Snow Cover	−1.54	−1.92	−2.36	−5.82
Wasteland	+7.89	+10.11	+10.24	+28.24
Water Bodies	−0.66	−1.51	−3.31	−5.48

), reflecting the dynamic nature of the landscape. In the first decade from 1990 to 2000, notable reductions were observed in the areas of open forest (−5.11 Km²), dense forest (−4.98 Km²), grassland (−2.46 Km²), snow cover (−1.54 Km²), and water bodies (−0.66 Km²). Conversely, there was an expansion in the areas of wasteland (+7.89 Km²), cropland (+3.53 Km²), built-up (+2.83 Km²) and agriculture plantation (+0.50 Km²), during this decade. Similar trends were observed in the second decade from 2000 to 2010. Reductions were apparent in the areas of dense forest (−7.50 Km²), grassland (−5.44 Km²), open forest (−5.19 Km²), snow cover (−1.92 Km²), and water bodies (−1.51 Km²). Concurrently, there was an increase in the areas of wasteland (+10.11 Km²), cropland (+5.28 Km²), built-up (+5.22 Km²), and agriculture plantation (+0.95 Km²). Moving to the third decade from 2010 to 2020, noteworthy changes persisted in LULC patterns and again reductions were evident in the areas of dense forest (−8.21 Km²), grassland (−5.85 Km²), open forest (−5.57 Km²), snow cover (−2.36 Km²), and water bodies (−3.31 Km²). In contrast, significant expansions continued to occur in wasteland (+10.24 Km²), built-up (+9.85 Km²), cropland (+3.69 Km²), and agriculture plantation (+1.52 Km²), during this decade. Overall from 1990–2020, over a period of 30 years, the area occupied by dense forest showed the maximum reduction of about −20.69 Km² followed by open forest (−15.87 Km²), grassland (−13.75 Km²) and snow cover (−5.82 Km²) of the total area occupied by the respective classes in 1990. The area occupied by wasteland showed maximum increase of about +28.24 Km² followed by built-up (+17.90 Km²), cropland (+12.50 Km²), and agriculture plantation (+2.97 Km²), respectively (Figure 5).

3.3.2. Conversion of LULC Classes Through Change Matrix Analysis

The transformation matrices generated for different time periods i.e., between 1990 and2000 (

Table 6. Transition matrix between 1990 and 2000.

Class	Agriculture Plantation	Built-Up	Cropland	Dense Forest	Grassland	Open Forest	Snow Cover	Wasteland	Water Bodies	Grand Total (1990 Area in Km2)
Agriculture Plantation	53.05	0.09	0.00	0.00	0.00	0.00	0.00	0.00	0.00	53.14
Built-up	0.00	54.01	0.00	0.00	0.00	0.00	0.00	0.00	0.00	54.01
Cropland	0.50	1.03	788.60	0.00	0.00	0.00	0.00	0.00	0.00	790.13
Dense Forest	0.00	0.34	0.72	954.02	2.34	0.72	0.00	0.89	0.00	959.03
Grassland	0.09	0.64	3.37	0.02	309.87	0.13	0.00	0.93	0.00	315.05
Open Forest	0.00	0.47	0.93	0.00	0.38	234.06	0.00	4.23	0.00	240.07
Snow Cover	0.00	0.00	0.00	0.00	0.00	0.00	43.05	1.54	0.00	44.59
Wasteland	0.00	0.24	0.00	0.01	0.00	0.05	0.00	148.17	0.00	148.47
Water Bodies	0.00	0.02	0.04	0.00	0.00	0.00	0.00	0.60	24.85	25.51
Grand Total (2000 Area in Km2)	53.64	56.84	793.66	954.05	312.59	234.96	43.05	156.36	24.85	2630.00

), 2000 and 2010 (

Table 7. Transition matrix between 2000 and 2010.

Class	Agriculture Plantation	Built-Up	Cropland	Dense Forest	Grassland	Open Forest	Snow Cover	Wasteland	Water Bodies	Grand Total (2000 Area in Km2)
Agriculture Plantation	53.48	0.16	0.00	0.00	0.00	0.00	0.00	0.00	0.00	53.64
Built-up	0.00	56.84	0.00	0.00	0.00	0.00	0.00	0.00	0.00	56.84
Cropland	1.05	1.64	790.97	0.00	0.00	0.00	0.00	0.00	0.00	793.66
Dense Forest	0.00	0.86	1.70	945.49	2.94	0.62	0.00	2.44	0.00	954.05
Grassland	0.06	1.23	4.22	1.04	303.37	0.67	0.00	2.00	0.00	312.59
Open Forest	0.00	0.79	1.96	0.00	0.84	228.29	0.00	3.08	0.00	234.96
Snow Cover	0.00	0.00	0.00	0.00	0.00	0.00	41.13	1.92	0.00	43.05
Wasteland	0.00	0.51	0.04	0.02	0.00	0.19	0.00	155.60	0.00	156.36
Water Bodies	0.00	0.03	0.05	0.00	0.00	0.00	0.00	1.43	23.34	24.85
Grand Total (2010 Area in Km2)	54.59	62.06	798.94	946.55	307.15	229.77	41.13	166.47	23.34	2630.00

), 2010 and 2020 (

Table 8. Transition matrix between 2010 and 2020.

Class	Agriculture Plantation	Built-Up	Cropland	Dense Forest	Grassland	Open Forest	Snow Cover	Wasteland	Water Bodies	Grand Total (2010 Area in Km2)
Agriculture Plantation	54.32	0.26	0.00	0.00	0.00	0.00	0.00	0.01	0.00	54.59
Built-up	0.00	62.06	0.00	0.00	0.00	0.00	0.00	0.00	0.00	62.06
Cropland	1.21	2.71	795.02	0.00	0.00	0.00	0.00	0.00	0.00	798.94
Dense Forest	0.00	2.85	1.50	937.31	2.03	0.15	0.00	2.71	0.00	946.55
Grassland	0.58	1.90	4.10	1.02	297.70	1.85	0.00	0.00	0.00	307.15
Open Forest	0.00	1.77	1.60	0.00	1.57	222.09	0.00	2.74	0.00	229.77
Snow Cover	0.00	0.00	0.00	0.00	0.00	0.00	38.77	2.36	0.00	41.13
Wasteland	0.00	0.33	0.36	0.01	0.00	0.11	0.00	165.66	0.00	166.47
Water Bodies	0.00	0.03	0.05	0.00	0.00	0.00	0.00	3.23	20.03	23.34
Grand Total (2020 Area in Km2)	56.11	71.91	802.63	938.34	301.30	224.20	38.77	176.71	20.03	2630.00

) and 1990 and 2020 (

Table 9. Transition matrix between 1990 and 2020.

Class	Agriculture Plantation	Built-Up	Cropland	Dense Forest	Grassland	Open Forest	Snow Cover	Wasteland	Water Bodies	Grand Total (1990 Area in Km2)
Agriculture Plantation	52.62	0.51	0.00	0.00	0.00	0.00	0.00	0.01	0.00	53.14
Built-up	0.00	54.01	0.00	0.00	0.00	0.00	0.00	0.00	0.00	54.01
Cropland	2.76	5.38	781.99	0.00	0.00	0.00	0.00	0.00	0.00	790.13
Dense Forest	0.00	4.05	3.92	936.22	7.31	1.49	0.00	6.04	0.00	959.03
Grassland	0.73	3.77	11.69	2.08	291.20	2.65	0.00	2.93	0.00	315.05
Open Forest	0.00	3.03	4.49	0.00	2.79	219.71	0.00	10.05	0.00	240.07
Snow Cover	0.00	0.00	0.00	0.00	0.00	0.00	38.77	5.82	0.00	44.59
Wasteland	0.00	1.08	0.40	0.04	0.00	0.35	0.00	146.60	0.00	148.47
Water Bodies	0.00	0.08	0.14	0.00	0.00	0.00	0.00	5.26	20.03	25.51
Grand Total (2020 Area in Km2)	56.11	71.91	802.63	938.34	301.30	224.20	38.77	176.71	20.03	2630.00

), reveal the varying intensity of change within the study area across different LULC classes. These matrices illustrate the altered and unaltered areas, as well as the total change within each class. The diagonal values within each transition matrix denote the unchanged area of each LULC class. The values in the rows indicate the area lost by a specific LULC class, while the values in the columns signify the area gained by that class. It is evident from these transformation matrices that there has been a continuous interconversion of area among different LULC classes. For instance, between 1990 and 2020, the study area witnessed significant deforestation, resulting in the conversion of forested areas into wasteland, cropland, grassland and built-up areas. Conversely, afforestation efforts led to the conversion of some grassland and wasteland into forested areas. However, despite afforestation efforts, there was an overall decrease in forest cover within the area.

Furthermore, the shrinking of water bodies and reduction in snow cover played a significant role in the expansion of bouldery and sediment areas, consequently contributing to the increase in the wasteland category. This expansion notably enlarged the wasteland area within the district. While grasslands saw some expansion due to deforestation, the conversion of grassland into cropland and built-up areas led to an overall reduction in grassland coverage. The built-up class experienced gains from various sources including forest, grassland, cropland, wasteland, and water bodies etc. The establishment of agriculture plantations resulted from the conversion of cropland and grassland, further highlighting the dynamic nature of LULC changes over time. These changes emphasize the various pressures driving land transformation, such as deforestation and urbanization, while also emphasizing the potential for conservation efforts, like afforestation, to counteract these changes.

3.4. Deforestation and Afforestation Analysis

The district’s forest cover, categorized as ‘Dense Forest’ and ‘Open Forest’ in the present study, forms the dominant landscape feature; however, it has been subject to significant depletion over the past three decades. Analysis reveals a concerning trend of conversion, particularly evident between 1990 and 2000, during which approximately 11.02 Km² of forest land transitioned to alternative land use categories. This period saw a minimal compensatory effort, with only 0.93 Km² of other land use types mainly grassland and wasteland undergoing afforestation initiatives. The subsequent decade, from 2000 to 2010, witnessed a slight rise in deforestation, with about 15.23 Km² of forested area lost to various alternative land uses. While this period also saw some afforestation activities, primarily on 2.54 Km² of grassland and wasteland, it was insufficient to counterbalance the loss. Inevitably, this trend persisted from 2010 to 2020, where approximately 16.92 Km² of forested land succumbed to deforestation, giving way to crop land, grassland, built-up areas, and wasteland. Despite efforts to mitigate this loss, evidenced by afforestation initiatives covering 3.14 Km², the net loss of forested area remained substantial.

The cumulative impact of deforestation from 1990 to 2020 amounts to approximately 43.17 Km², while afforestation efforts only managed to reclaim 6.61 Km² of land. This imbalance highlights the pressing need for concerted conservation measures and sustainable land management practices to safeguard the ecological integrity and biodiversity of the district’s forest cover (Figure 6).

3.5. Future LULC Prediction Through CA Markov Model

According to the analysis, the built-up land area in the entire study region was 71.90 Km² in 2020, and it’s projected to increase to 101.7 Km² by 2050. The results showed that the rapid development of built up area would be accompanied by a sharp decline in forests and grasslands (

Table 10. LULC change prediction in district Rajouri between 2020–2050.

LULC Class	Area in 2020 (Km2)	Area Predicted in 2050 (Km2)	Predicted Change in Area (2020–2050) (Km2)	%Age Change
Agriculture Plantation	56.11	59.59	+3.48	+6.20
Built-up	71.90	101.7	+29.80	+41.45
Cropland	802.63	813	+10.37	+1.29
Dense Forest	938.34	908.78	−29.56	−3.15
Grassland	301.30	285.23	−16.07	−5.33
Open Forest	224.20	205.67	−18.53	−8.26
Snow Cover	38.77	30.66	−8.11	−20.92
Wasteland	176.71	211.51	+34.80	+19.69
Water Bodies	20.03	13.86	−6.17	−30.81

). Cropland is expected to experience a slight increase, rising from 798.64 Km² to 802.63 Km² by 2050.

The predictions indicate substantial losses in both dense and open forest areas, with dense forests decreasing from 938.34 Km² to 908.78 Km² and open forests declining from 224.20 Km² to 205.67 Km² over the specified time frame (Figure 7). Grasslands are also forecasted to undergo significant reductions, decreasing by about 16.7 Km² from 301.30 Km² in 2020 to 285.23 Km² by 2050. Water bodies are expected to decrease further from 20.03 Km² in 2020 to 13.86 Km² in 2050. Similarly, snow cover is projected to shrink to 30.66 Km² from its initial extent of 38.77 Km² in 2020. The areas currently occupied by snow and water bodies are anticipated to be converted to barren land, contributing to an increase in the wasteland area within the district.

3.6. Accuracy of CA-Markov Model

The comparison between the simulated and actual LULC map of 2020 demonstrated a high degree of agreement (Figure 8). The spatial patterns of different classes were consistently reproduced by the CA-Markov model, with only minor deviations observed in a few categories. The scatter plot of actual versus simulated areas (Figure 9) revealed a near 1:1 relationship, further indicating that the CA-Markov model successfully captured the proportional distribution of LULC categories. Most points were closely aligned along the fitted line, highlighting the model’s ability to predict the category-wise area with high reliability.

Statistical validation indices supported these visual observations. Agreement due to chance was 0.1000, agreement due to quantity was 0.1388, and grid-cell level agreement reached 0.7029. Disagreement values were minimal, with disagreement at the grid-cell level (0.0280), quantity (0.0303), and strata (0.0000) (

Table 11. Classification agreement and disagreement illustrating the ability of the model to specify quantity and spatial allocation accurately in predicting LULC for the year 2020.

Information of Allocation	Information of Quantity
	No[n]	Medium[m]	Perfect[p]
Perfect[P(x)]	P(n) = 0.5068	P(m) = 0.9697	P(p) = 1.0000
PerfectStratum[K(x)]	K(n) = 0.5068	K(m) = 0.9697	K(p) = 1.0000
MediumGrid[M(x)]	M(n) = 0.4902	M(m) = 0.9417	M(p) = 0.9246
MediumStratum[H(x)]	H(n) = 0.1000	H(m) = 0.2388	H(p) = 0.2440
No[N(x)]	N(n) = 0.1000	N(m) = 0.2388	N(p) = 0.2440
Agreement Chance = 0.1000
Agreement Quantity = 0.1388
Agreement Strata = 0.0000
Agreement Gridcell = 0.7029
Disagree Gridcell = 0.0280
Disagree Strata = 0.0000
Disagree Quantity = 0.0303

).

The Kappa statistics confirmed the robustness of the model (

Table 12. Kappa statistics representing the accuracy of predicted LULC for the year 2020.

Statistic	Index
Kno	0.9352
Klocation	0.9617
KlocationStrata	0.9617
Kstandard	0.9234

). K_no (0.9352), K_location (0.9617), K_{locationStrata} (0.9617), and K_standard (0.9234) all indicated very strong agreement between the simulated and actual maps.

4. Discussion

LULC transitions play a pivotal role in shaping global environmental trends, as they influence processes such as groundwater recharge, evapotranspiration, and the occurrence of natural hazards [78]. Assessing these changes is crucial for understanding the dynamic interactions between anthropogenic activities and natural systems [79].The present study conducted in district Rajouri of J&K, India revealed substantial changes in the LULC classes over a span of 30 years from 1990 to 2020. This period witnessed both losses and gains among various LULC classes, indicating a dynamic transformation in the landscape. During this period, forest cover and grasslands of the district have been significantly reduced to other land use types indicating a disturbing situation with adverse implications for the local climate and ecosystem services of the region [80,81]. The expansion of cropland, settlements and urbanization has significantly contributed to this decline. Globally, cropland expansion has been seen as a major factor for the deforestation [1]. Built-up area has substantially increased at the expense of grasslands, cropland, forests, and water bodies. Furthermore, the future projection using the CA Markov model indicates that this trend of LULC change is likely to continue, with built-up areas projected to increase further at a fast rate to 101.7 Km² by 2050, accompanied by declines in forests, grasslands, and water bodies.

The concurrent decline in snow cover and shrinkage of water bodies in Rajouri represents another critical concern with far-reaching consequences for regional water security. The reduction of snowpacks, largely driven by rising temperatures in recent decades [45,79], diminishes the natural reservoirs of freshwater essential for agriculture, industry, and domestic use. This decline, coupled with the drying of water bodies, reduces overall water storage capacity and intensifies the expansion of barren lands. As formerly productive areas become parched and infertile, land degradation accelerates, heightening vulnerability to desertification. In higher altitudes, expansion of barren land also intensifies soil erosion and mass movement processes [82], further undermining ecological stability.

Studies conducted in the Kashmir Himalayas and other developing countries [43,44,55,83,84] have shown a substantial rise in built-up areas over the past two decades, with projections indicating further expansion in the coming years. In western Himalayan region, Singh et al. [85] reported a 184% surge in urbanization between 1975 and 2015, along with major increases in barren land (+30%) and sharp declines in forests (−11%), water bodies (−8%), scrubland (−6%), and glaciers/snow (−20%). Their projections for 2055 indicated further urban expansion (+63%) and additional forest loss (−9%), highlighting the escalating anthropogenic pressure on natural ecosystems. Saleem et al. [86] reported similar trajectories in Jammu district, where agricultural land (+6.71%), barren land (+6.45%), and settlements (+4.12%) expanded between 1990 and 2020, while vegetation declined significantly (−16.57%). In the Tawi catchment, Jasrotia et al. [87] employed a CA-Markov and SWAT modelling framework to assess past and future land cover changes and their hydrological implications. They reported that between 2000 and 2020, rangeland (−0.01%) and forest land (−0.88%) decreased, while agricultural (+0.88%) and urban land (+0.21%) expanded. Their future projections for 2050 and 2080 indicate further losses in forest land (−6.77% and −8.51%), rangeland (−0.30% and −0.41%), and perennial snow/ice cover (−1.05% and −1.24%), with simultaneous increases in agricultural land (+2.10% and +3.68%) and urban land (+5.47% and +5.88%). Likewise, Bashir et al. [45], documented similar trends in Baramulla district of J&K, north-western Himalaya, where urban expansion during 2000–2020 was accompanied by declines in snow cover, forest cover, agricultural land, and water bodies. Using CA-Markov modelling, they projected continued urban expansion by 2030, driven primarily by economic compulsions, climate variability, and population growth. Furthermore, Bhat et al. [88] documented significant LULC transitions in Hirpora Wildlife Sanctuary, Western Himalayas, where snow cover (−33.76%) and dense forests (−6.80%) showed major losses, while barren/rocky areas (+11.43%) and built-up land (+0.35%) expanded between 1992 and 2021.

Nevertheless, the swift urbanization in the hilly regions of Rajouri district has a notable impact on vegetation cover, leading to potential environmental degradation. Urban expansion is largely driven by rural–urban migration, employment opportunities, rising living standards, natural growth, and urban redefinition [78].The consequences of diminishing forests and grasslands, and increased bare soil are manifold, encompassing compromised environmental quality, heightened risks to ecological biodiversity and wildlife, as well as escalating pollution levels and deteriorating air quality [69]. Urban development, often correlated with population growth and the proliferation of personal and public transportation, is anticipated to lead to rise in environmental pollutants in the future. As urban areas expand, the destruction and degradation of natural purifiers such as plants and soil may occur, intensifying environmental hazards and posing significant threats to the people [89]. Therefore, policymakers and planners must take into account the projected spatial extent and distribution patterns of LULC categories concerning their adverse effects, particularly focusing on housing characteristics, urban sprawl, infrastructure and utilities and vegetation cover [69]. By considering these factors, they can develop effective strategies to mitigate the negative impacts of urban development and ensure sustainable growth while safeguarding the environment and enhancing the quality of life for residents.

In recent years, several interlinked factors have acted as major drivers of LULC change in the region. Among them, rapid population growth, urban expansion, climate variability, economic pressures, and unplanned development have been identified as the most significant influences [8,44]. The continuous increase in human population has intensified the pressure on natural resources worldwide [83]. In combination with urbanization and economic development, this demographic growth has heightened the demand for food, energy, and water, thereby exerting considerable stress on the environment [90]. Climate change adds another dimension to these challenges, affecting the Earth’s systems in multiple ways [91]. In the Kashmir Himalayas, a clear warming trend has been documented, with annual increases of 0.05 °C and 0.01 °C in mean maximum and minimum temperatures, respectively, accompanied by a decline of around 4.22 mm in precipitation per year between 1980 and 2016 [79,92]. These shifts in climate have contributed to glacier retreat in the Himalayas, leading to a decline in snow cover and significant implications for water resources and hydrological processes [93]. In addition to climatic and demographic pressures, institutional and policy-related factors also shape LULC dynamics. In many developing countries, urbanization trajectories are largely determined by government-led policies and interventions, making governance a decisive force in the spatial structure and land use patterns of communities [94]. In the UT of J&K, the absence of a comprehensive land use policy has enabled land to be used according to immediate needs without proper regulation. Although such practices may generate short-term economic benefits, they pose serious risks for the long-term sustainability of land resources, particularly soil health and productivity [8].

The drivers of land use changes, often human-induced, are inherently complex. Understanding their intricacies involves examining how they influence various policies governing land use management [95,96]. Thus, it is imperative for policymakers to develop comprehensive land use policies that account for the complexities of these drivers and promote sustainable land management practices to ensure the long-term well-being of the region and its inhabitants. Evidence suggests that sound policy frameworks yield the most favorable ecological and socioeconomic outcomes at the regional level [97]. Approaches such as community-based management and conservation have proven effective in ensuring sustainable monitoring and fostering social responsibility [98]. Importantly, the formulation of effective land use policies should be informed by bottom-up approaches, where insights from ground-level studies guide decision-making and resource allocation. Such an approach can foster more sustainable and equitable utilization of available resources [97].

Alongside policy considerations, the demand for land-based resources and direct human interventions such as deforestation for agriculture, rapid urban expansion, road construction, logging for fuelwood and timber, forest fires, and overgrazing (Figure 10) remain the most visible forces shaping LULC change. These drivers, however, are not uniform; they vary across space and time, producing spatially and temporally heterogeneous patterns of land transformation [44,99,100]. Therefore, a spatially explicit and temporally informed understanding of LULC change is essential for evaluating anthropogenic pressures on the natural environment [6]. The observed decline in forest cover and expansion of built-up areas in Rajouri highlight the need for sustainable land management (SLM) interventions. Practices such as agroforestry, reforestation, and habitat restoration may help mitigate further ecological degradation and enhance ecosystem resilience in the northwestern Himalayas [101]. These measures could contribute to maintaining soil stability, conserving water resources, and supporting long-term ecological sustainability.

In the present study, the integration of remote sensing and GIS techniques has facilitated a comprehensive and quantitative assessment of LULC change patterns in the Rajouri district. For future research, the use of higher-resolution satellite imagery, coupled with the incorporation of additional ancillary datasets, such as detailed socio-economic, demographic, and climatic information, could significantly enhance the depth and accuracy of analyses. This would enable a more nuanced understanding of the complex processes driving observed land changes, including intra-urban variations and land cover dynamics in hilly and heterogeneous terrains. Such an approach would not only improve the spatial and temporal resolution of LULC studies but also provide critical understanding for planning sustainable land management strategies and addressing environmental challenges in mountainous regions.

5. Conclusions

The current study analyzed the spatio-temporal trends of LULC change in Rajouri district utilizing satellite imagery. The findings indicate marked transformations in various land use systems since 1990. Over this period, forests, grasslands, water bodies, and snow cover have declined, while built-up areas, croplands, agricultural plantations, and wastelands have expanded. The massive land transformation is largely driven by anthropogenic actions and has been mostly adverse in nature, giving rise to multiple environmental issues in the study area. The loss of forests not only threatens biodiversity but also compromises vital ecosystem services such as carbon sequestration, soil retention, and water regulation. Additionally, the encroachment upon forested areas increases the risk of habitat fragmentation and wildlife displacement, further imperiling the region’s ecological balance. Equally troubling is the diminishing extent of grasslands, which play a crucial role in supporting the socio-economic livelihoods of local communities. With many residents dependent on animal husbandry for their sustenance, the decline in grasslands poses a direct threat to their economic well-being and way of life. Moreover, the degradation of grasslands can intensify soil erosion, reduce water infiltration rates, and compromise the resilience of ecosystems to climate variability. The continued loss of water bodies, retreat of snow cover and the expansion of barren lands could have dire consequences for the region’s ecosystems, biodiversity, and socio-economic fabric. Hence, it is crucial to adopt sustainable forest management practices, initiate reforestation projects, and restore degraded grasslands. Furthermore, community-based conservation efforts, coupled with robust policies and regulations, are indispensable for promoting the coexistence of human activities and natural ecosystems.