The present study utilized multi-temporal satellite imagery, meteorological datasets, and ancillary geospatial data to analyze the spatio-temporal patterns of vegetation cover, rainfall variability, and land surface temperature (LST) across the Lohit District. The data sources, their characteristics, and applications are summarized below.

Temporal patterns were assessed for warming or stabilization phases. a) Conversion of Digital Numbers (DN) to Top-of-Atmosphere (TOA) Radiance To convert the digital number (DN) values from Landsat 5 thermal infrared data (Band 6) to TOA spectral radiance (Lλ), the following equation is applied:

Where:

This conversion adjusts for the dynamic range of the sensor, transforming the pixel values from DN to radiance values that are physically meaningful and suitable for further temperature calculation.

b) Conversion of TOA Radiance to Brightness Temperature (BT)

The TOA radiance obtained in the previous step is then converted to brightness temperature (BT) in Kelvin using the Planck function, with the following equation:

Where:

These constants are specific to Landsat 5 and are typically provided in the metadata files (MTL) for each scene. The formula uses logarithmic transformation based on the inverse relationship between radiance and temperature in thermal infrared wavelengths.

c) Conversion of Brightness Temperature to Land Surface Temperature (LST)

Finally, to estimate the Land Surface Temperature (LST), the brightness temperature is corrected for atmospheric effects and emissivity (ϵ\epsilon) using the following equation:

The emissivity value (ϵ\epsilon) accounts for the effects of surface material type (e.g., vegetation, water, urban areas) on the thermal radiation observed by the sensor. Emissivity can be determined based on land cover classification or by using standard values (e.g., 0.95 for vegetation, 0.98 for water bodies).

The calculation of Land Surface Temperature (LST) for Landsat 8 data follows a multi-step procedure that involves radiometric correction, the calculation of brightness temperature (BT), vegetation indices, and the determination of land surface emissivity (ε), followed by the final estimation of LST. The following sections describe the steps involved in the LST calculation, utilizing the thermal infrared bands and associated formulas:

a) Conversion from Digital Number (DN) to Top-of-Atmosphere (TOA) Radiance (L)

The first step in the radiometric processing of Landsat 8 data is to convert the Digital Numbers (Qcal) from the thermal infrared band (Band 10) to Top-of-Atmosphere (TOA) radiance. The relationship is given by:

This step corrects for the sensor's response to incoming radiation, producing the TOA radiance values necessary for further processing.

b) Conversion from TOA Radiance to Brightness Temperature (BT)

Once the TOA radiance is obtained, it is converted to brightness temperature (BT) in Kelvin using the following equation, which is based on Planck’s law:

Where:

This step provides an estimate of the Earth's surface temperature in Kelvin, which is subsequently converted to degrees Celsius by subtracting 273.15.

c) Normalized Difference Vegetation Index (NDVI)

The NDVI is used to quantify vegetation cover, as it has a significant influence on the land surface emissivity (ε). NDVI is calculated using the reflectance values from the red (Band 4) and near-infrared (Band 5) bands of Landsat 8:

NDVI values range from − 1 to + 1, with higher values indicating denser vegetation. NDVI plays a key role in estimating land surface emissivity, as it influences the amount of vegetation cover.

d) Vegetation Proportion (Pv)

The vegetation proportion (Pv) is derived from the NDVI values and reflects the fraction of the land surface covered by vegetation. It is calculated using the formula:

Where:

NDVI min and NDVI max are the minimum and maximum values of NDVI observed in the study area, respectively.

The square root transformation normalizes the Pv values, ensuring they fall within the range of 0 to 1, where 0 represents bare soil and 1 represents dense vegetation.

e) Land Surface Emissivity (ε)

Land surface emissivity (ϵ\epsilon) is a key factor influencing the conversion of brightness temperature to LST. It is estimated from the vegetation proportion (Pv) using the empirical relationship:

This equation reflects the relationship between the amount of vegetation (as indicated by Pv) and the surface’s ability to emit thermal radiation. The emissivity of bare soil and dense vegetation differ, which impacts the final LST calculation.

f) Calculation of Land Surface Temperature (LST)

The term accounts for the relationship between the brightness temperature and the emissivity. The natural logarithmic transformation of ϵ adjusts the temperature for different land surface types both temporal trends and spatial patterns.

To quantify the relationships between vegetation dynamics and climatic variables, Spearman’s rank correlation analysis was applied to examine associations among Normalized Difference Vegetation Index (NDVI), rainfall, and Land Surface Temperature (LST) for the period 1988–2023. Spearman’s correlation coefficient (ρ) is a non-parametric statistical measure that evaluates the strength and direction of monotonic relationships between variables and is particularly suitable for long-term environmental datasets that may exhibit non-normal distributions, outliers, or heteroscedasticity (Conover, 1999; Wilks, 2011).

Annual mean NDVI values derived from satellite imagery were statistically compared with corresponding annual rainfall totals and mean LST values to assess climate–vegetation linkages. This approach minimizes the influence of short-term variability and emphasizes long-term coupling between vegetation condition and hydroclimatic factors, as commonly adopted in large-scale vegetation–climate interaction studies (Myneni et al., 1997; Zhou et al., 2016).

Spearman’s rank correlation was preferred over parametric alternatives (e.g., Pearson’s correlation) because NDVI, rainfall, and LST time series often violate assumptions of linearity and normality, particularly in heterogeneous mountainous environments (Legendre & Legendre, 2012). Statistical significance was evaluated at a confidence level of p < 0.05, consistent with standard practice in environmental and climatological research.

The resulting correlation coefficients were used to characterize the direction and relative strength of associations between vegetation greenness, moisture availability, and surface thermal conditions, providing a quantitative basis for interpreting climate–vegetation interactions discussed in subsequent sections.

Spatio-temporal analysis of Normalized Difference Vegetation Index (NDVI) and Landsat-derived classified imagery from 1988 to 2023 reveals pronounced changes in vegetation density, composition, and spatial distribution across Lohit District. Over the past three and a half decades, the district has undergone a systematic transformation of forest structure, reflecting the combined effects of sustained anthropogenic pressure and hydroclimatic variability.

NDVI-based land-cover maps (Figs. 3–6) illustrate a persistent decline in dense forest cover accompanied by a concurrent expansion of moderate forest and shrub–grassland classes. The quantitative evolution of land-cover categories is summarized in Table 3 and illustrated in Fig. 2. Dense forest cover decreased continuously from 3,122.39 km² (70.05%) in 1988 to 908.35 km² (20.38%) in 2023, representing a net loss of approximately 71% over the study period. This trend indicates widespread canopy thinning and forest degradation rather than abrupt or localized deforestation

In contrast, moderate forest cover increased substantially from 495.59 km² to 2,119.60 km², corresponding to an increase of approximately 327%, suggesting large-scale conversion of dense forest into lower-canopy forest formations. Similarly, shrub–grassland cover expanded from 536.56 km² to 1,283.73 km², an increase of nearly 139%, reflecting progressive forest degradation, secondary succession, and the expansion of disturbed landscapes. Areas classified as bare soil, rock, sand, snow, and cloud exhibited temporal variability but declined overall from 302.94 km² in 1988 to 145.66 km² in 2023, potentially indicating partial vegetation recovery in previously exposed zones. Water bodies showed no detectable change throughout the study period. The total geographical area remained nearly constant at approximately 4,457 km² across all time slices, confirming that the observed trends represent actual land-cover transitions rather than classification or geometric artifacts.

The observed spatio-temporal patterns of forest cover change are closely linked to anthropogenic drivers operating in the district. Historical records document the rapid expansion of plywood and timber industries in Lohit and adjacent districts during the 1990s, resulting in large-scale extraction of commercially valuable species such as Dipterocarpus gracilis (hollong) and Terminalia myriocarpa (hollock) (Roychowdhury, 1992). In addition, traditional jhum (shifting) cultivation, practiced primarily by the Tai Khampti and Singpho communities, has contributed to forest clearance and delayed regeneration, particularly where fallow periods have shortened (FAO, 2020).

Population growth and agricultural expansion have further intensified land-use conversion, while infrastructure development, including road construction, hydropower projects, and settlement expansion has accelerated habitat fragmentation and landscape discontinuity (Dash et al., 2007). These pressures are compounded by illegal logging, forest fires, and encroachment, which collectively exacerbate forest degradation and increase the likelihood of long-term ecological transformation (Aymonier et al., 2010).

Overall, the spatio-temporal analysis reveals a clear shift from dense forest dominance toward structurally simplified forest and non-forest vegetation types. The results underscore sustained forest degradation driven by interacting human activities and environmental stressors, highlighting the need for integrated forest management approaches to arrest further decline and support long-term ecosystem recovery.

Spatio-temporal analysis of Normalized Difference Vegetation Index (NDVI) and Landsat-derived classified imagery from 1988 to 2023 reveals pronounced changes in vegetation density, composition, and spatial distribution across Lohit District. The observed trends reflect the combined influence of sustained anthropogenic pressure and hydroclimatic variability operating over the past three and a half decades.

Spatio-temporal analysis of rainfall data derived from Indian Meteorological Department (IMD) gridded datasets for Lohit District over the period 1988–2023 reveals pronounced temporal variability and clear long-term changes in precipitation patterns across the district. Rainfall characteristics were assessed using annual minimum and maximum rainfall values, which effectively capture interannual extremes and provide insight into long-term hydroclimatic trends.

Annual rainfall statistics (Table 4) indicate a consistent and statistically significant declining trend over the 35-year study period. Maximum annual rainfall decreased from 3,456.97 mm in 1988 to 2,179.38 mm in 2023, corresponding to an overall reduction of approximately 37%. In contrast, minimum annual rainfall exhibited a more pronounced decline, decreasing from 3,092.21 mm to 1,011.67 mm, representing an approximate 67% reduction. The sharper decline in minimum rainfall suggests an increasing frequency and severity of low-precipitation years, which may intensify moisture stress during critical periods of vegetation growth and regeneration. Trend detection using the Mann–Kendall non-parametric test confirms that both minimum and maximum rainfall series exhibit statistically significant negative trends (Mann, 1945; Kendall, 1975).

Temporal co-variation of rainfall and surface temperature is illustrated in Fig. 7, which highlights increasing interannual rainfall variability alongside a progressive decline in overall precipitation magnitude in recent decades. The widening range between minimum and maximum rainfall values in later years reflects heightened hydroclimatic variability, providing important climatic context for interpreting concurrent changes in vegetation condition and land surface temperature observed across the district.

Spatial rainfall distribution maps (Figs. 8–11) further demonstrate marked spatio-temporal heterogeneity in precipitation across the district. The maps show a gradual contraction of high-rainfall zones (> 3,000 mm) and an expansion of moderate- to low-rainfall areas, particularly after 2000. These spatial patterns indicate a redistribution of rainfall intensity across the landscape rather than uniform decline.

Overall, the rainfall analysis reveals a significant long-term decline in both minimum and maximum precipitation, accompanied by increasing interannual variability. These spatio-temporal trends provide a critical climatic context for interpreting observed changes in vegetation dynamics and forest cover across the district.

Spatio-temporal analysis of Land Surface Temperature (LST) derived from Landsat 5 TM and Landsat 8 TIRS datasets indicates a clear warming tendency across Lohit District during the period 1988–2023. Annual minimum and maximum LST values were examined to characterize long-term thermal variability and changes in surface energy conditions. Summary statistics for selected years are presented in Table 5.

The LST records show an overall increase in surface thermal conditions over the study period. Maximum LST values increased from approximately 25.83°C in the late 1980s to 26.37°C in 2023, with the highest values exceeding 30°C observed during the mid-study years (2008–2013). These elevated maximum temperatures indicate enhanced daytime surface heating, particularly during years characterized by reduced vegetation cover and lower rainfall.

Minimum LST values also exhibit a long-term warming tendency, reflected in a reduction in the frequency and intensity of extremely low surface temperatures in recent years. Although interannual variability remains evident, the overall shift toward higher minimum LST values suggests progressive nighttime warming and reduced surface cooling, consistent with broader regional warming trends.

Temporal co-variation between rainfall and LST is illustrated in Fig. 12, which highlights an inverse pattern whereby periods of declining rainfall correspond to elevated surface temperatures. This temporal correspondence provides additional context for understanding concurrent changes in vegetation conditions observed during the same period.

Spatial distribution maps of LST (Figs. 13–16) further demonstrate pronounced spatio-temporal heterogeneity in surface temperature across the district. Over time, higher-temperature zones expand spatially, while cooler zones associated with dense forest cover contract. Increasing thermal contrasts between forested and non-forested areas are particularly evident in later years, reflecting the influence of land-cover change on surface energy balance.

Figure 15: Land Surface Temperature of Lohit District of the Year 2008 and 2013

Figure 16: Land Surface Temperature of Lohit District of the Year 2018 and 2023

Overall, the spatio-temporal LST analysis reveals a sustained increase in both daytime and nighttime surface temperatures across Lohit District, accompanied by increasing spatial heterogeneity. These thermal trends provide an important climatic context for interpreting observed changes in rainfall patterns and forest cover dynamics during the study period.

To quantify the relationships between vegetation dynamics and hydroclimatic variables, Spearman’s rank correlation analysis was conducted between NDVI, rainfall, and Land Surface Temperature (LST) for the period 1988–2023 (Table X).

The correlation matrix demonstrates strong and statistically significant linkages between vegetation greenness, precipitation, and surface thermal conditions. The positive association between NDVI and rainfall (ρ = +0.69, p < 0.05) indicates that vegetation productivity and canopy conditions in Lohit District are strongly regulated by moisture availability. In monsoon-dominated forest ecosystems, precipitation is a primary control on photosynthetic activity, biomass accumulation, and regeneration processes (Myneni et al., 1997; Zhou et al., 2016). Adequate rainfall enhances soil moisture availability and sustains canopy vigor, whereas reductions—particularly in minimum rainfall—can intensify moisture stress during critical growing periods. Such hydroclimatic stress is likely to constrain forest recovery and amplify degradation in areas already experiencing canopy thinning.

The strong negative correlation between NDVI and LST (ρ = − 0.75, p < 0.05) provides robust evidence that increasing surface temperatures are associated with declining vegetation vigor. Reduced canopy cover diminishes shading and evapotranspirative cooling, resulting in higher sensible heat flux and elevated land surface temperatures (Kalnay & Cai, 2003; Mildrexler et al., 2011). Elevated LST, in turn, exacerbates soil moisture depletion and imposes thermal stress on vegetation, thereby suppressing growth and regeneration. Similar NDVI–LST relationships have been widely documented across forested regions, highlighting the role of intact vegetation cover in regulating surface energy balance and mitigating warming (Peng et al., 2014).

The negative association between rainfall and LST (ρ = − 0.62, p < 0.05) further emphasizes the tight coupling between atmospheric moisture availability and surface thermal regimes. Reduced precipitation limits latent heat flux and enhances surface heating, leading to higher LST values, particularly in degraded or sparsely vegetated landscapes (Bonan, 2008; Zhou et al., 2016). This relationship underscores how hydroclimatic variability not only directly affects vegetation growth but also indirectly amplifies thermal stress through land–atmosphere feedback.

Taken together, the relationships presented in Table 6 provide quantitative evidence of a reinforcing climate–vegetation feedback in Lohit District. Declining rainfall and increasing surface temperatures act synergistically to weaken canopy structure, constrain vegetation recovery, and accelerate forest degradation. As dense forests transition toward more open or simplified states, their capacity to buffer local climate diminishes, potentially transforming them from regulators of surface temperature into contributors to surface warming. These findings highlight the importance of conserving remaining dense forest cover and promoting canopy restoration to enhance ecosystem resilience under ongoing climatic change.

Anthropogenic pressures have acted synergistically with climatic stress to intensify forest degradation. Historical expansion of timber and plywood industries during the 1990s resulted in extensive extraction of commercially valuable species such as Dipterocarpus gracilis and Terminalia myriocarpa, leading to canopy opening and fragmentation (Roychowdhury, 1992). Traditional jhum (shifting) cultivation, particularly under shortened fallow cycles, has further limited forest recovery and promoted secondary vegetation dominance (Ramakrishnan, 1992; Tripathi & Barik, 2003). In addition, infrastructure development, including roads and hydropower projects—has increased landscape fragmentation and accessibility, reinforcing degradation processes (Dash et al., 2007).

The interaction between sustained human disturbance and increasing hydroclimatic variability explains the persistence of degraded forest states and the limited recovery of dense forest cover observed in recent decades. These findings underscore the vulnerability of Eastern Himalayan forests to compound pressures arising from land-use change and climate variability (Chapin et al., 2008; IPCC, 2021).

Overall, the Discussion demonstrates that forest dynamics in the district are governed by a synergistic interaction between vegetation condition, rainfall variability, surface thermal regimes, and anthropogenic disturbance, emphasizing the need to consider integrated climate–vegetation–land-use linkages in assessments of forest degradation and resilience.

The observed interactions between forest degradation, declining rainfall, and rising surface temperature highlight the necessity of climate-responsive and ecosystem-based forest management approaches. Restoration strategies should prioritize native, site-adapted, and drought-tolerant species to enhance resilience under increasing hydroclimatic variability. Maintaining and restoring canopy continuity is particularly important for reducing surface warming and stabilizing local microclimates (Peng et al., 2014).

Strengthening community-based forest management is essential in regions where traditional land-use practices remain integral to livelihoods. Improved fallow management, regulated timber extraction, and promotion of agroforestry systems can reduce pressure on natural forests while sustaining local economies (Ramakrishnan, 1992; FAO, 2020). In addition, micro-watershed development and soil–water conservation measures can mitigate moisture stress and reduce erosion in degraded landscapes.

Long-term satellite-based monitoring of vegetation indices, rainfall, and LST should be institutionalized to support early warning systems and adaptive management. Integrating remote sensing diagnostics with ground-based hydrometeorological observations can provide a robust framework for tracking ecosystem health and guiding policy interventions in the Eastern Himalayan region.