Spatio-temporal assessment of land use, climate dynamics, and biophysical parameters in the Ranikhola watershed, Sikkim Himalaya, India Open Access

The advent of climate change has precipitated a pronounced warming trend, characterized by a 1.5 °C escalation in temperature, accompanied by a concomitant 15% augmentation in precipitation patterns, over the course of recent decades (IPCC 2013). These changes have significant implications for the fragile ecosystems of the Himalayas, where altitudinal gradients create distinct ecological zones sensitive to shifts in temperature and precipitation (Adhikari et al. 2007). The phenomenon of rapid land cover change, characterized by the replacement of natural ecosystems with diverse anthropogenic land uses, is a pervasive global trend, extensively supported by empirical research (Geist & Lambin 2001). Ranikhola, located in East Sikkim, serves as a vital water source for agriculture, hydropower generation, and drinking purposes for the local community. The region's unique ecological, economic, and cultural identity is threatened by rapid land use/land cover (LULC) changes, driven by population growth, climate change, and development pressures. These dynamics underscore the need for a sophisticated understanding of the interdependencies between human activities and environmental systems, particularly in mountain ecosystems that are highly vulnerable to external pressures (Kala 2005; Rahaman et al. 2020).

LULC changes in the Sikkim Himalaya between 1990 and 2015 illustrate this phenomenon, with a reported 30% increase in agricultural land and a 25% decrease in forest cover, indicating significant landscape modifications (Kumar et al. 2019). Such transformations not only disrupt local ecosystems but also exacerbate environmental vulnerabilities, including biodiversity loss, hydrological imbalances, and increased susceptibility to natural disasters like landslides. Effective land use planning, informed by accurate LULC assessments, is critical for addressing these challenges and ensuring sustainable development (Herold et al. 2008; Roba et al. 2024; Yadav et al. 2024).

Land surface temperature (LST), the earth's radiative skin temperature influenced by solar radiation, varies with LULC types and sunlight exposure, serving as a crucial indicator of energy balance and a key parameter in microclimate studies (Pu et al. 2006; Khandelwal et al. 2018; John et al. 2020). Higher LST values are observed at lower altitudes, fostering conditions conducive to vegetation growth, while cooler temperatures at higher elevations influence snow cover and water availability. Understanding LST patterns and their relationships with environmental indices, such as the normalized difference snow index (NDSI), normalized difference water index (NDWI), and normalized difference vegetation index (NDVI), provides valuable insights into the impacts of LULC changes on ecosystem stability (Husain et al. 2023).

While several studies have examined LULC changes in the Himalayan region, few have integrated advanced methodologies to model and characterize the spatio-temporal trajectories of these changes. For instance, Nawaz & Shrestha (2016) emphasize the role of developmental pressures and population growth in accelerating land cover transitions in the Himalayan region. Saxena et al. (2024) documented significant LULC transformations in Western Rajasthan, underscoring the environmental consequences of forest depletion. Additionally, there is limited research on the relationships between LST and environmental indices, particularly in the context of high-altitude ecosystems. Most studies focus on short-term observations. The absence of long-term climate projections further hinders the ability to anticipate and mitigate future risks (Tesfaw et al. 2023). This study addresses these gaps by leveraging cutting-edge remote sensing (RS) and geographic information systems (GIS) tools, combined with artificial neural networks (ANNs) and climate trend analysis, to provide a comprehensive understanding of the Ranikhola watershed's ecological dynamics.

This study endeavours to tackle the pressing issues of LULC changes, climate dynamics, and biophysical parameters in the Ranikhola Watershed by pursuing the following objectives: (i) To analyse variations in LULC patterns across different altitudinal zones, elucidating the relationships between land use changes and environmental factors. (ii) To examine the statistical trends of key climatic parameters, including annual earth skin temperature (EST), atmospheric temperature, precipitation, surface pressure, wind speed direction, and relative humidity, to gain a deeper understanding of climate change dynamics and their far-reaching impacts on the hydro-cryosphere of the mountainous terrain. (iii) To predict future climate scenarios using high-resolution data and modelling techniques, inform proactive planning and adaptation strategies, and analyze the relationships between LST and environmental indices (NDSI, NDWI, and NDVI), to assess the impact of LULC changes on LST.

The detection and estimation of LULC changes were effectively accomplished through the application of sophisticated ANN algorithms, which facilitated the precise identification and quantification of spatial and temporal transformations in LULC patterns. High-resolution hyperspectral Landsat data and advanced spatial analysis coupled with ANN algorithms are used to detect and forecast LULC changes up to 2113. The ANN model's impressive performance empowers that it could be a promising decision support tool for predicting LULC analysis in Sikkim's Watersheds for sustainable water management of water resources. To project future climate scenarios and assess the potential ecosystem vulnerabilities, a sophisticated climate trend analysis was performed, utilizing the CSIRO-Mk-3-6-0 model, the RCP 8.5 scenario, and Coupled Model Intercomparison Project Phase 5 (CMIP5) data. Mann–Kendall and Sen's slope tests were used to validate the observed climatic trends. These methodologies enable precise quantification of spatial and temporal transformations, offering robust insights into the impacts of LULC changes on ecosystem resilience.

By integrating advanced spatial analysis and predictive modelling, this study provides actionable insights for sustainable land-use planning and climate adaptation. The findings contribute to bridging existing research gaps and inform evidence-based strategies for mitigating the environmental and socio-economic impacts of climate change in the Sikkim Himalaya. Ultimately, this research promotes resilient and sustainable practices in India's ecologically sensitive regions.

In this research, open sources of geostatistical data were utilized. Geospatial data, detailed in Table 1, was acquired from the National Remote Sensing Centre (NRSC), United States Geological Survey (USGS), National Aeronautics Space Administration (NASA). Based on the study area's topographical characteristics, the Cartosat digital elevation model (DEM) tiles with a 30 m resolution were employed for watershed delineation (Muralikrishnan et al. 2011). Landsat satellite images (Thematic Mapper (TM) and thermal infrared sensor (TIRS)) captured between 1993 and 2023, obtained from USGS with minimal cloud cover (less than 10%), were utilized. Thermal infrared bands of Landsat 5 and 9 were employed for temperature and LULC distribution analysis. The study was taken in the dry post-monsoon period (i.e., November and December) to acquire the best cloud-free satellite images over the decade. The Landsat images were used for LULC change analysis and calculation of LST, NDVI, NDSI, and NDWI. Ground truth observations of 20 GPS sites were made to verify LULC classifications and facilitate the identification of each LULC type within the study area. Annual average climatic datasets were sourced from the NASA Langley Research Centre POWER Project (Zhang et al. 2019), measuring air temperature, relative humidity (recorded at 2 m), wind speed, and direction (recorded at 10 m) by using NASA's MERRA2 satellite for the Gangtok meteorological station (latitude 27.331, longitude 88.614). Additionally, the trend analysis was validated using decadal weather data provided by the Indian Meteorological Department.

The present study employed the supervised maximum likelihood classification technique to investigate LULC changes from 1993 to 2023 (Figure 3). This method utilized the false colour composite (FCC) technique, which integrates blue, green, and red bands, to select features such as water bodies and vegetation, using ArcGIS 10.7.1, following the approach outlined by Seyam et al. (2023). A total of approximately 2,400 pixels were chosen from each category as spectral signatures, representing six attributes, to aid in image classification. These samples were grouped into six LULC classes for streamlined analysis and easy assessment of changes. The MOLUSCE plugin in quantum geographic information system (QGIS) was used to simulate transitions between LULC classes and estimate changes over space and time (Baig et al. 2022). A multilayer machine learning algorithm combining ANN and cellular automata (CA) was utilized. A comprehensive overview of the LULC categories is provided in Table 2.

The calculation of LST depends on the LULC of surface area. The emission and reflectance of energy depend on the material of the earth's surface (Alexander 2020). LST fluctuates in response to variations in climate and human activities. As a crucial parameter, it plays a significant role in determining the maximum and minimum temperatures of a specific area. The LST of Landsat 5 was calculated from the following three formulas:

• (i) Adaptation of the digital number, to the spectral radiance (L):

  
• (ii) Convert radiance into BT in Kelvin

  
• (iii) Convert degrees Kelvin then into degrees Celsius

  

The methodologies used for the calculation of LST for Landsat 9 are given in Supplementary Table S1.

This study employs modern biophysical parameters (NDVI, NDWI, and NDSI) to assess climate change by monitoring vegetation, moisture area estimation, and snow accumulation. Supplementary Table S2 shows specific details.

The Q_med sign indicates data trend reflection, while its value indicates the steepness of the trend. To determine whether the median slope is statistically different than zero, one should obtain the confidence interval of Q_med at a specific probability. In this study, Sen's slope was employed to analyze the climate trend, using R software. The results show that a positive Sen's slope value indicates an upward trend, whereas a negative value indicates a downward trend in the time series.

To validate the model, 20 GPS stations were selected where large transmissions occurred. Station-wise daily on-field climate data was recorded using Android apps and instruments to validate estimated rainfall and humidity. In our analysis of the 2023 classified LULC images, kappa, and overall accuracy exceeded standard levels, with values of 97.3397 and 0.965%, respectively (Supplementary Table S3). The small standard error and improved kappa values demonstrate precise classification, suggesting negligible user bias and a high level of accuracy that surpasses random chance, thereby confirming the reliability of the classification results.

The study detected significant changes in all LULC types within the watershed. As of 2023, vegetation remains the predominant land cover, occupying 63.38% of the area. Meanwhile, built-up areas have expanded significantly, rising from 2.84 to 5.33%, with notable growth concentrated around Gangtok city. Barren land expanded from 7.27 to 10.92%, and agricultural land decreased slightly from 12.98 to 12.45% due to traditional dependence. LULC changes occurred as agricultural land was converted to agroforestry systems, shifting land use from agriculture to forestry.

In this study, a matrix table analysis, presented in the form of a heatmap, was employed to precisely quantify the conversion of LULC types into other LULC categories. Table 4 displays the total land area in square kilometres for each LULC category, with columns showing the changes from 1993 to 2008 and rows showing the changes from 2008 to 2023. The matrix table illustrates the land area conversions between different LULC types, where each cell represents the transformed land area. For example, the cell in the sixth column and fifth row indicates that 2.58 km² of dense vegetation were converted to barren land during the period of study. The highlighted red colour values indicate areas of maximum changes, representing significant transitions, while the light green colour highlights a low rate of transition, indicating comparatively stable LULC patterns.

The transition probability matrix is developed by analyzing the future LULC class transitions and their corresponding change levels for each region (Mas et al. 2014). The CA-ANN model utilizes fuzzy logic, with a fixed range of 0–1, tailored to the terrain's characteristics. The CA-ANN's core components are the interactions between interconnected neurons and the dynamic adjustments to the connection strengths that support these interactions.

The MOLUSCE plugin was used to conduct transition potential analysis using an ANN multilayer perception model with a neighbourhood size of 1 pixel, a maximum of 1,000 iterations, a momentum of 0.05, a minimum error of 0.05 for validation, and a validation kappa of 0.9781 (Figure 6). 1,000 random samples are used to spatially represent the data for ANN training. The neural network predicted LULC changes for the years 2038–2113, based on a fixed simulated learning rate of 0.01. The model leveraged LULC data from 2008 to 2023 to forecast changes in 15-year increments, specifically predicting the state of LULC in 2038 through iterative simulations.

Analyzing changes in LULC in areas with complex terrain is difficult. However, using RS and GIS techniques provides a valuable approach to studying and understanding these changes (Meer & Mishra 2020). Following successful model validation, LULC predictions were generated for three future periods: near-future (2038–2053), mid-future (2068–2083), and far-future (2098–2113), at 15-year intervals. The analysis of LULC changes revealed substantial transformations in land cover classes, characterized by a notable expansion of agricultural land and built-up areas, accompanied by a decline in vegetation and water bodies over the forecasted period. A substantial transformation of approximately 3% was observed, transitioning from open forest to agricultural land. The simulated LULC maps for the predicted periods are illustrated in Figure 8.

Biophysical indices help monitor vegetation, temperature, and water resources under climate change. Climate trends of key elements are shown in Figure 9.

The northeastern part of the watershed, detected by NDSI, is snow-covered in winter and barren otherwise (Figure 10). The year 1993 recorded the highest snowfall. However, a declining trend in NDSI values from 0.78 in 1993 to 0.52 in 2008 and 0.57 in 2023 suggests glacial melting and increased snowfall, elevating the risk of Glacial Lake Outburst Floods from higher elevations in the Ranikhola region.

This finding is supported by the LULC analysis results, which show a significant increase in vegetation cover and a corresponding decrease in barren land, including ice, aligning with the observed temperature changes. Moreover, changes in LST can also indicate LULC change, such as deforestation or urbanization, which can have significant impacts on local climatic ecosystems. Remarkably, 27% of the Ranikhola watershed experiences surface temperatures between 11 and 14 °C, emphasizing the significance of understanding these variations for climate and microclimate studies in the region.

Different types of LULC can significantly influence local climate regulation. LST varies across different LULC features, with notable correlations (Kustas et al. 2003; Weng & Yang 2004). The distribution of vegetation, snow, water bodies, and other features acts as a temperature regulator, determining the surface temperature. In the Ranikhola watershed, areas with abundant vegetation and water bodies tend to be cooler compared to those with sparse vegetation and minimal water bodies, highlighting the crucial role of LULC in moderating local climate conditions. In the longitudinal transect along line AB (NE to SW) for the year 2023, a comprehensive analysis of LST, NDVI, NDWI, and NDSI reveals intricate connections between vegetation, temperature, water discharge, and snow cover in the Ranikhola watershed (Figure 4). The data illustrates fluctuations in each indicator concerning the distance from the water source. High NDVI values correspond to lower temperatures, indicating dense vegetation reduces temperature, while low NDVI values signify higher temperatures, reflecting sparse vegetation. Furthermore, negative NDWI and NDSI values signify low water discharge and minor snow cover due to extreme LST. Figure 11 presents a comparative scenario of four indices at three distinct time points. LST varies between −5.2 and 21.69 °C, with an increasing trend observed between 13,000 and 15,000 m and from 27,000 m onwards. Conversely, LST decreases between 0 and 6,000 m, where green cover, seasonal ice accumulation, and water storage contribute to this decline, confirming the negative correlation between green cover, water presence, and LST.

The study employed Mann–Kendall and Sen's slope trend tests to assess the effects of LST changes. These non-parametric tests, suitable for meteorological data, analyzed various climate factors (temperature, rainfall, surface pressure, wind speed, wind direction, and relative humidity) in the study area from 1981 to 2021. The results showed a significant upward trend in temperature and other weather events (Figure 9), confirmed by the p-values in Table 5, indicating a strong correlation between LST changes and climate variability. The extreme climate events over the years have drawn attention to climate change, leading to the rejection of the null hypothesis (H0) in favour of the alternative hypothesis (H1). Positive Sen's slope values (0.2332, 0.0019, 0.019, 0.2312, and 0.0166) indicate a significant upward trend in annual relative humidity, surface pressure, temperature, precipitation, and EST in the Ranikhola watersheds. In contrast, Wd and Ws show minimal negative Sen's slope values, suggesting a slight decreasing trend.

A decadal analysis reveals a pronounced association between LST and multiple environmental parameters, underscoring a meaningful correlation at this timescale. Figure 5 shows that LST and NDVI exhibit positive correlations at multiple time points within the study area, indicating a direct relationship between temperature and vegetation health, indicating favourable conditions for plant growth, amplified by global warming-induced temperature rise (R² values: 0.27 in 1993, 0.09 in 2008, and 0.30 in 2023). This typically occurs during the winter season when vegetation is dormant. In these conditions, the amount of solar radiation reaching the ground was limited, and the insulating effect of vegetation was reduced. A strong negative correlation was found between LST and NDWI, with R² values of 0.04, 0.17, and 0.32, indicating that the presence of water bodies significantly reduces surface temperatures. This highlights the crucial role of water resources in moderating temperature extremes and mitigating the impacts of high temperatures. Interannual variations in vegetation, reflected in NDVI values, can significantly influence LST values (Abbas & Hamdi 2022). The analysis of LST and NDSI demonstrated a strengthening negative correlation (R²: 0.32, 0.19 in 2008, 2023, underscoring the adverse impact of rising LST on snow cover in the watershed. A significant positive correlation (R²: 0.03) in 1993 indicated glacier melting in Ranikhola. Reduction in snow, especially in the upper reaches of the watershed, heightens the risk of rain-on-snow events, increasing the likelihood of flash floods. These findings shed light on the complex interplay between temperature, vegetation, water bodies, and snow cover, underscoring the critical importance of adopting sustainable resource management practices to address the challenges posed by climate change and ensure a resilient future.

This study investigates the impacts of significant LULC changes in the Ranikhola watershed over the past three decades on climate dynamics and hydro-cryospheric processes. Integrating ANNs and high-resolution LULC analysis revealed complex relationships between LULC and climate. The research quantifies relationships between LULC transitions and biophysical parameters, providing projections of future climate and land use trends for informed planning. LULC changes were dynamic over the research period, with notable declines in forest cover and expansions of built-up areas at lower elevations, which were supported by the findings of Mishra et al. (2020). Rising LST and erratic precipitation patterns correlated strongly with observed LULC shifts. Projected climate scenarios indicate persistent warming and increased rainfall variability, exacerbating ecological vulnerabilities, including soil erosion, biodiversity loss, and degraded vegetation health. Gangtok City, a prominent urban centre, was selected for assessing climate change, exhibiting pronounced climate variability with significant upward trends in monthly temperature (+1.5 °C) and annual rainfall (+50 mm) over the past five decades. Analysis of NASA's MERRA2 data reveals variations in average annual temperature (maximum and minimum) and rainfall distribution throughout the study period. Notably, the temporal dynamics of LST can be quantitatively assessed through the computation of urban heat island (UHI) intensity, a methodological approach supported by existing research (Hamdi et al. 2023). Rising temperatures and irregular precipitation will likely amplify heat stress, flooding, and waterlogging in the watershed. These changes threaten crop viability and increase vector-borne disease risks – impacts expected to worsen with continued climate change (Bolan et al. 2024).

This research reveals significant alterations in the Ranikhola watershed's five altitudinal zones (I–V) over three decades, characterised by drastic transformations in land cover, including a notable decline in dense vegetation and water bodies, with a 5.89 km² decrease in water bodies in the lower altitudinal zone. Notably, zones I–IV exhibited increased planted vegetation, while all zones except III and V witnessed a decline in water bodies. The NDSI analysis revealed a declining trend from 0.78 to 0.52 to 0.57, indicating glacial melting and increased snowfall risk, as NDSI >0 is considered to have some snow present (Mohammadi et al. 2023). Conversely, the NDVI values fluctuated narrowly between 0.53 and 0.51, which is crucial in uncovering the long-term patterns of vegetation change, as demonstrated by Fu & Burgher (2015), while the NDWI showed a significant decline from 0.16 to 0.07. A 30-year temperature analysis revealed a pronounced warming trend, with minimum and maximum LST increasing by +6 and +8 °C, respectively. These changes were accompanied by expanding barren land and settlements, declining waterbodies (−2.7%), and reduced vegetation and snow cover, culminating in a significant increase in LST, with 27% of the area experiencing temperatures between 11– and 14 °C. The loss of plant cover, as indicated by NDVI measurements, appears to have contributed to elevated LST values during the research period. Research by Gelata et al. (2023) demonstrates an inverse linear relationship between LST and NDVI, while Almouctar et al. (2024) observed a positive correlation between NDVI and LST. The observed positive correlation between NDVI and LST indicates that areas with denser vegetation experienced elevated surface temperatures, likely due to the Himalayan canopy structure's tendency to trap heat and restrict air circulation within these regions. These findings emphasize the urgent need for sustainable land management and climate change mitigation strategies in the Ranikhola watershed.

Futuristic local climate scenarios were generated using statistical downscaling from the CMIP5 CSIRO-Mk3-6-0 model (Jeffrey et al. 2014) and the RCP 8.5 Assessment report-5. The analysis highlights that maximum and average temperatures are significantly rising, with a projected increase of 0.5–1.5 °C by 2093 (Figure 11). Additionally, there is a decreasing trend in minimum temperature, providing insights into changing climate patterns in the watershed. Moreover, there is a noticeable increase in annual rainfall variability, with a substantial addition of 40 mm of rainfall from 2013 to 2093. These observations signal a potential climate crisis, raising significant concerns for the region's future.

The study's outcomes and futuristic projections have implications for other Himalayan regions experiencing similar LULC changes. The study analysed, LULC alterations play a pivotal role in regulating temperature fluctuations by influencing the Earth's surface heat balance (Tesfaw et al. 2023). The rapid expansion of built-up areas has led to a significant increase in LST. A decline in NDVI values from 10.35% in 1993 to 8.58% in 2023 underscores the impact of LULC changes. Specifically, built-up areas increased by 6.30 km², while vegetation and waterscapes decreased by 4.53 and 6.93 km², respectively. Variations in barren land, vegetation, and urban areas can also influence historical temperature and rainfall patterns, which was related to the findings of Sharma et al. (2023). Elucidating these LULC-climate interactions is essential for developing effective climate change mitigation and adaptation strategies. The approach employed in this study provides actionable insights, enabling stakeholders and decision-makers to make informed choices regarding natural resource planning and management.

The study of LULC changes in the Ranikhola watershed of Sikkim, spanning from 1993 to 2023, reveals significant environmental impacts driven by declines in forest cover and urban expansion, particularly at lower elevations. This analysis employs ANN for LULC classification at a fine spatial resolution of 30 m, highlighting the urgent need for government intervention as these changes threaten local ecosystems and climatic stability. Moreover, the study demonstrates the effectiveness of RS and GIS in analyzing surface variables like temperature, vegetation growth, and urbanization across different geographical zones. Biophysical indices, such as NDVI, NDWI, and NDSI, underscore the vulnerability of vegetation, water resources, and snow cover to warming trends. Therefore, it is crucial to adopt a comprehensive approach to address environmental vulnerability and climate change (EVCC), as these factors intricately relate to the health of the watershed and its resilience to environmental shifts.

Additionally, factors such as glacial retreat and alterations in snow cover due to climate change further complicate the environmental landscape. The investigation places particular emphasis on biophysical parameters, including vegetation indices, soil moisture, and water resource availability. Findings indicate that rising temperatures are negatively influencing vegetation health and reducing soil moisture levels, ultimately impacting local water dynamics. The increase in evapotranspiration rates, alongside rising humidity and pressure, has been linked to accelerated glacier melting, further altering the ecological framework of the watershed. To combat these challenges, the study advocates for integrated strategies focused on sustainable agroforestry, renewable energy adoption, and eco-tourism promotion. It underscores the importance of comprehensive disaster preparedness while recommending initiatives in reforestation, water conservation, and sustainable urban planning. By prioritizing these approaches, policymakers can enhance community resilience, mitigate environmental risks, and preserve the ecological integrity and biodiversity of the Ranikhola watershed for future generations.

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

The authors declare there is no conflict.