Monitoring Agricultural Vegetation Health Under Climate Stress Using NDVI and LST Indices in the Sylhet Region ^†

Abstract

Agricultural ecosystems in northeastern Bangladesh are increasingly vulnerable to climate-induced stressors, particularly rising temperatures and seasonal droughts. While previous research has examined the climate’s impact on agriculture in broader contexts, no study has specifically investigated long-term seasonal vegetation and thermal dynamics in Sylhet. This study addresses this gap by assessing spatio-temporal variations in vegetation health under climate stress in the Sylhet region from 2005 to 2025 using remote sensing techniques. To investigate this problem, the study derived the Normalized Difference Vegetation Index (NDVI) and land surface temperature (LST) from Landsat satellite imagery and evaluated their seasonal behavior across the major cropping periods Rabi, Kharif I, and Kharif II. The relationship between vegetation health and surface temperature was examined using Pearson’s correlation matrix along with a statistical comparison to identify change patterns, transitions among vegetation and thermal stress classes, and the seasonal intensity of climate stress. The findings indicate that increased LST generally corresponds with reduced vegetation cover in lowland agricultural zones, whereas elevated areas with forest or tree covers show an opposite response. Distinct spatial hotspots of thermal stress and drought-prone zones were also identified, particularly during the dry Rabi season. These results highlight the idea that rising LST corresponds with declining NDVI values, indicating that increasing thermal stress and potential reductions in agricultural vegetation productivity and climate stress across Sylhet’s agricultural landscape have intensified markedly from 2005 to 2025, with clear seasonal differences in vulnerability. NDVI analysis reveals a consistent decline in vegetation health, while LST patterns show widespread transitions from moderate to high and severe thermal stress, particularly during the Kharif seasons. The observed NDVI decline under elevated LST conditions indicates reduced vegetation vigor and potential productivity within agricultural lands, rather than a direct reduction in cultivated areas, since NDVI primarily captures vegetation density and physiological condition. The strongest NDVI–LST inverse relationship occurs in Rabi and Kharif I, indicating vegetation’s cooling role, whereas this linkage weakens in Kharif II due to dominant monsoon-driven atmospheric controls.

1. Introduction

Bangladesh’s farmlands in the northeast are becoming more susceptible to climate change as temperatures are increasing, rainfall patterns are changing, and the frequency of severe weather patterns is having negative impacts on crop production and the general health of vegetation. Although the urgency of this issue is widely recognized, previous studies have primarily focused on general climate–vegetation interactions, with limited attention to long-term seasonal vegetation dynamics and their relationship with surface temperature in the Sylhet region [1]. Early remote sensing research demonstrated that linear combinations of red and near-infrared reflectance are strongly related to vegetation biomass, chlorophyll content, and green leaf area, forming the basis of the Normalized Difference Vegetation Index (NDVI) for monitoring vegetation condition [2]. Studies linking satellite-derived surface temperature and NDVI have also shown strong associations between canopy density and land surface temperature, indicating that vegetation cover plays a major role in regulating surface energy balance and evapotranspiration processes [3]. More recent research continues to build on these foundations, demonstrating that NDVI trends are closely tied to temperature and precipitation variations across diverse ecosystems, with human activities also influencing vegetation dynamics [4,5]. Advanced statistical approaches, including machine learning, enhance the detection of complex spatio-temporal vegetation patterns [6,7]. This study aims to investigate seasonal agricultural vegetation dynamics and evaluate how thermal stress influences vegetation conditions across cropping seasons in Sylhet from 2005 to 2025, with particular focus on the Rabi (mid-November to mid-March), Kharif I (mid-March to mid-July), and Kharif II (mid-July to mid-November) seasons [8,9]. Therefore, the present research uses an analytical framework that is based on satellite-derived NDVI as well as land surface temperature (LST) gathered from Landsat images. These results highlight the pressing need for the introduction of climate-resilient management measures that would support agricultural productivity in the Sylhet area. This study lays a critical foundation for guiding policy and practical interventions in response to ongoing climate change. In this study, climate stress refers to the combined effects of temperature variability, precipitation changes, and drought conditions on agricultural systems. In contrast, thermal stress specifically denotes heat-related impacts derived from LST. Therefore, thermal stress is treated as a measurable component of broader climate stress.

Sylhet District is located in the northeastern region of Bangladesh between latitudes 24°36′0″ N and 25°11′0″ N and longitudes 91°38′0″ E and 92°30′0″ E. It is bordered by the Indian states of Meghalaya, Tripura, and Assam, as well as the Bangladeshi districts of Moulvibazar, Habiganj, and Sunamganj. Covering an area of 3490 square kilometers, the district is characterized by hilly terrain forming part of the Khasi and Jaintia Hills [10]. The location and extent of the study area are shown in Figure 1.

Figure 1. Study area map of Sylhet District, Bangladesh.

The region contains fertile valleys, river basins, and forested hills, all of which contribute to its biodiversity and agricultural potential [11,12]. Sylhet experiences a humid subtropical climate with distinct seasonal variations, receiving substantial monsoon rainfall, where winters are mild and dry, and summers are hot and humid. The landscape is predominantly agricultural, with tea (Camellia sinensis) [13], rice (Oryza sativa) [14], and other crops well-adapted to the local climate [15].

2. Materials and Methods

Bangladesh’s agricultural seasons are Rabi, Kharif I, and Kharif II. Rabi has cool temperatures for wheat (Triticum aestivum) and mustard (Brassica juncea), and Kharif I coincides with monsoon for Aman rice (Oryza sativa L.), while Kharif II peaks during monsoon for Boro rice (Oryza sativa L.) [16,17,18,19]. The methodological framework for processing satellite data and analyzing the seasonal relationship between NDVI and LST in the Sylhet study area of Bangladesh is shown in Figure 2.

Figure 2. Methodological insights of NDVI and LST correlation; conceptualization based on [20].

2.1. NDVI and LST Derivation

Landsat images were used to derive NDVI and LST. Images were carefully selected to represent comparable stages of the main agricultural seasons in Sylhet. For the Rabi season (mid-November to mid-March), images acquired on 10 February 2005 from Landsat 5 and 17 February 2025 from Landsat 8 were used. For the Kharif I season (mid-March to mid-July), images from 2 June 2005 and 9 June 2025 were selected. Finally, for the Kharif II season (mid-July to mid-November), the analysis utilized images acquired on 8 October 2005 and 15 October 2025. These acquisition dates fall within the mid to late stages of each cropping season, when vegetation typically reaches strong canopy development. Selecting imagery from similar phenological phases helps ensure that the vegetation indices reflect actual seasonal vegetation conditions rather than early growth or post-harvest stages, thereby improving the reliability of the comparison between 2005 and 2025. NDVI, derived from red and near-infrared light reflectance, is widely used in remote sensing to represent photosynthetically active vegetation and can indicate vegetation density and vigor. Values near +1 represent dense vegetation, values around 0 indicate sparse vegetation or bare surfaces, and negative values generally correspond to water bodies or built-up areas. NDVI values derived from Landsat imagery between 2005 and 2025 were categorized into four vegetation classes: degraded/bare land, stressed, moderately healthy, and healthy vegetation. Using the Reclassify tool in ArcMap 10.8 with the Natural Breaks classification method, NDVI vegetation maps were generated (Figure 2) [9,12].

$$NDVI={\displaystyle \frac{NIR-Red}{NIR+Red}}$$

LST reflects the Earth’s surface radiative temperature and is derived from thermal infrared (TIR) satellite data. This study used Landsat 5 TM for 2005 and Landsat 8 OLI/TIRS for 2025 to extract NDVI and LST. Landsat 5 TM utilized Band 3 (red), Band 4 (NIR) for vegetation, and Band 6 (TIR) for LST, while Landsat 8 used Band 4 (red), Band 5 (NIR) for NDVI, and Band 10 (TIR) for LST. These bands allow consistent monitoring of vegetation and surface temperature over 2005 and 2025, enabling precise spatio-temporal analysis. LST was calculated from the Landsat imagery as follows:

• Converting Digital Number (DN) to Spectral Radiance

$$L_{\lambda }=\left({\displaystyle \frac{L_{max}-L_{min}}{Q_{calmax}-Q_{calmin}}}\right)\ast \left(DN-Q_{calmin}\right)+L_{min}$$

• Conversion of Radiance to Brightness Temperature (TB)

$$TB={\displaystyle \frac{K2}{\mathit{ln}\left(\left(\frac{K1}{L_{\lambda }}\right)+1\right)}}$$

• Calculation of NDVI

$$NDVI={\displaystyle \frac{NIR-Red}{NIR+Red}}$$

• Estimation of Proportion of Vegetation (PV)

$$PV={\left({\displaystyle \frac{NDVI-{NDVI}_{min}}{{NDVI}_{max}-{NDVI}_{min}}}\right)}^2$$

• Estimation of Land Surface Emissivity (ε)

$$\epsilon =0.004 \ast PV+0.986$$

• Calculation of LST (Kelvin)

$$LST={\displaystyle \frac{TB}{1+\left(\lambda \ast \frac{TB}{\rho }\right)\ast \mathit{ln}\left(\epsilon \right)}}$$

• Conversion of Kelvin to Celsius

$$LST °\mathrm{C}=LST-273.15$$

where L_λ = spectral radiance, L_max and L_min = sensor max and min radiance, DN = pixel digital number, Q_calmax and Q_calmin = max and min DN, TB = brightness temperature, K1 and K2 = sensor calibration constants, λ = thermal band wavelength, and ρ = Planck’s constant (h) ∗ speed of light (c)/Boltzmann constant (σ) [21].

2.2. Statistical Relation

A spatially systematic approach was employed to evaluate the relationship between NDVI and LST across the major agricultural seasons of Sylhet. A fishnet grid was first generated to create uniform sampling units and subsequently clipped to the study area boundary. NDVI and LST layers for Rabi, Kharif I, and Kharif II were then integrated using the Extract Multi-Values to Points tool in ArcMap, allowing each grid-cell centroid to receive corresponding seasonal NDVI and LST values. The resulting attribute tables were exported to Excel for statistical analysis. Pearson’s correlation coefficients were computed to quantify the strength and direction of NDVI–LST relationships for each season [22]. Pearson’s correlation matrix was employed to understand the relationship between NDVI and LST across agricultural seasons.

3. Results

Thermal stress across Sylhet shows a clear intensification for over a 20-year life span in all three agricultural seasons. While 2005 is dominated by low-to-moderate stress, the 2025 maps reveal widespread transitions toward high and severe thermal stress, particularly in Kharif I (Figure 3). The expansion of red and yellow zones indicates a substantial rise in seasonal thermal load, highlighting increasing vulnerability of the region’s agricultural landscape to thermal stress (Figure 4).

3.1. Spatio-Temporal Transition of NDVI and LST in Agricultural Seasons

Both NDVI and LST exhibit significant seasonal and spatial variability influenced by climate and land-use changes, reducing vegetation and increasing surface temperatures.

Figure 3 shows a consistent NDVI decline across Sylhet with big regional differences. Eastern areas, especially Zakiganj and Kanaighat, saw the sharpest drop, shifting from moderately healthy vegetation in 2005 to widespread degradation by 2025. The administrative center shows moderate but increasing stress, while southern regions such as Balaganj and Fenchuganj now contain more stressed vegetation than before. Northern Sylhet, including Gowainghat and Companiganj, remains the most stable, with mostly healthy vegetation and only slight weakening. Overall, the east is the main hotspot, the center and south show moderate decline, and the north is relatively resilient. The transition matrix reflects the same downward trend. In Rabi, moderately healthy vegetation dominated at 18.11%, while recovery from degraded to healthy vegetation was only 0.27%. In Kharif I, the main shift was from healthy to moderately healthy vegetation at 11.21%, again with minimal recovery of 1.13%. In Kharif II, weakening intensified, led by a 15.98% shift from healthy to moderately healthy vegetation, while the smallest shift was 1.17% from stressed to degraded or bare land, as shown in Table 1.

Table 1. Temporal transition of NDVI across seasons.

Change in Vegetation Dynamics in m2	Rabi	Kharif I	Kharif II
Degraded/Bare Land–Degraded/Bare Land	108,554,314.7	121,103,898.4	77,586,346.38
Degraded/Bare Land–Stressed Vegetation	190,675,028.8	77,436,922.64	67,620,429.91
Degraded/Bare Land–Moderately Healthy Vegetation	28,680,613.47	40,794,187.7	93,568,166.7
Degraded/Bare Land–Healthy Vegetation	10,307,458.81	44,976,903.64	145,331,921.8
Stressed Vegetation–Degraded/Bare Land	67,592,197.85	279,814,097.6	50,043,397.01
Stressed Vegetation–Stressed Vegetation	607,149,211.8	421,024,056.1	176,842,668.3
Stressed Vegetation–Moderately Healthy Vegetation	373,381,691.3	244,972,221.3	187,676,918.5
Stressed Vegetation–Healthy Vegetation	61,927,643.8	269,245,680.8	141,882,051.5
Moderately Healthy Vegetation–Degraded/Bare Land	34,914,661.48	141,860,591.8	66,549,715.22
Moderately Healthy Vegetation–Stressed Vegetation	271,678,686.7	436,899,308.7	549,566,998.4
Moderately Healthy Vegetation–Moderately Healthy Vegetation	683,291,573.5	401,499,157	407,443,432.7
Moderately Healthy Vegetation–Healthy Vegetation	316,263,907.6	129,673,950.1	127,259,418.5
Healthy Vegetation–Degraded/Bare Land	17,559,061.19	66,320,470.02	34,439,176.83
Healthy Vegetation–Stressed Vegetation	82,744,176.85	233,799,381	358,212,739.9
Healthy Vegetation–Moderately Healthy Vegetation	205,629,079.1	448,000,081.2	682,289,008.5
Healthy Vegetation–Healthy Vegetation	309,673,016.4	92,599,937.77	283,794,023.9

Figure 3. NDVI analysis of the Sylhet region over the agricultural seasons.

Figure 3. NDVI analysis of the Sylhet region over the agricultural seasons.

Overall, these seasonal changes show a clear decline in vegetation vigor, with most transitions moving toward moderately healthy classes and very limited recovery to healthy vegetation. This pattern reflects rising ecological vulnerability and reduced vegetation resilience to seasonal climatic stress, especially during monsoon-driven cropping cycles. The NDVI transitions further clarify how vegetation health shifts.

Figure 4. LST analysis of the Sylhet region over the agricultural seasons.

Figure 4. LST analysis of the Sylhet region over the agricultural seasons.

Figure 4 shows a clear rise in thermal conditions across Sylhet from 2005 to 2025 in the Rabi, Kharif I, and Kharif II seasons, with strong regional contrasts. Sylhet Sadar exhibits the sharpest increase, shifting from low–moderate LST to high–severe levels, driven by rapid urbanization and vegetation loss. Dakshin Surma follows a similar pattern, moving into high thermal stress. Beanibazar and Zakiganj also warm rapidly, transitioning from moderate or low–moderate temperatures to widespread high–severe stress, marking them as major hotspots. Kanaighat shows a moderate rise with fewer severe zones. Golapganj, Fenchuganj, and Balaganj experience moderate increases, with expanding high-stress patches, but not at the scale of eastern Sylhet. Osmaninagar and Bishwanath show only low–moderate increases. The coolest and most stable areas remain Companiganj and Gowainghat, where LST rises only slightly within low–moderate ranges. In Rabi, the largest shift was moderate to low thermal stress at 22.54%, while severe-to-severe stress was minimal at 0.0035%, indicating little persistence of extreme winter conditions. In Kharif I, intensification peaks, with 29.10% transitioning from moderate to severe stress and only 0.0029% recovering from severe to low stress. In Kharif II, the dominant shift was moderate to high stress at 30.40%, while severe to low stress recorded 0%, showing no recovery once severe stress developed, as shown in Table 2.

Table 2. Temporal transition of LST across seasons.

Change in Thermal Stress in m2	Rabi	Kharif I	Kharif II
Low Thermal Stress–Low Thermal Stress	784,703.0199	24,470,608.3	51,236,126.39
Low Thermal Stress–Moderate Thermal Stress	836,911.0487	62,600,403.71	446,832,553.1
Low Thermal Stress–High Thermal Stress	687,982.3596	347,443,751.7	335,232,302.4
Low Thermal Stress–Severe Thermal Stress	175,516.3752	316,950,582	50,229,686.19
Moderate Thermal Stress–Low Thermal Stress	908,267,600.5	81,272,906.42	72,810,717.32
Moderate Thermal Stress–Moderate Thermal Stress	897,639,277.1	174,726,155.8	436,296,431
Moderate Thermal Stress–High Thermal Stress	584,752,491.5	918,544,083.3	1,249,634,669
Moderate Thermal Stress–Severe Thermal Stress	160,812,219.5	1,118,531,411	584,797,243.5
High Thermal Stress–Low Thermal Stress	78,234,742.13	4,067,881.024	2,300,291.285
High Thermal Stress–Moderate Thermal Stress	307,170,080.9	6,931,414.237	9,030,565.423
High Thermal Stress–High Thermal Stress	393,984,185.5	89,673,280.36	75,453,819.73
High Thermal Stress–Severe Thermal Stress	115,862,724	298,572,864.7	130,653,834.7
Severe Thermal Stress–Low Thermal Stress	293,200.8863	111,127.6611	0
Severe Thermal Stress–Moderate Thermal Stress	328,056.3035	518,263.305	1,234,215.77
Severe Thermal Stress–High Thermal Stress	197,715.0816	2,006,748.347	911,133.0514
Severe Thermal Stress–Severe Thermal Stress	142,741.6937	3,888,459.692	3,657,892.87

Overall, the seasonal transitions reveal a progressive trend toward higher thermal stress categories, with the most substantial increases consistently originating from moderate stress classes and escalating into high or severe stress. Conversely, transitions toward lower stress categories remain extremely limited, underscoring a growing thermal vulnerability across Sylhet’s agricultural landscape.

3.2. Relationship Between NDVI and LST in Different Agricultural Seasons

The NDVI–LST relationship varies notably across the agricultural calendar (Figure 5). In the Rabi season, vegetation cover displays a clear cooling influence, with healthier plant conditions corresponding to lower surface temperatures due to shading and evapotranspiration. This inverse relationship continues into Kharif I, where early monsoon vegetation still contributes meaningfully to surface temperature regulation. By Kharif II, however, the vegetation–temperature linkage becomes weak, as monsoon-driven factors such as cloud cover, humidity, and rainfall exert stronger control over thermal conditions.

Figure 5. Pearson’s correlation matrix over the agricultural seasons.

These patterns indicate that vegetation plays a dominant role in moderating surface temperature during drier or early monsoon periods, while atmospheric processes overshadow its influence during peak monsoon months.

4. Discussion

Studies in Dhaka and Mymensingh reported significant inverse correlations between NDVI and LST, where NDVI decreased, and LST increased over time due to urban expansion and loss of green space [23]. In Dhaka, NDVI and LST showed a strong negative correlation, and built-up indices had positive correlations with LST, reflecting the urban heat island effect [24]. Similarly, in Sylhet City, rapid urbanization increased LST, and Pearson correlation showed a strong negative relationship between NDVI and LST, while built-up areas correlated positively with LST [25]. In line with these findings, Sylhet shows a clear pattern of declining vegetation health alongside rising thermal stress between 2005 and 2025, with the strongest degradation occurring in Zakiganj and Kanaighat and greater resilience in northern uplands such as Gowainghat and Companiganj. Seasonal NDVI transitions indicate weakening vegetation across all periods, as healthy classes increasingly shift into moderately healthy or stressed states with minimal recovery from degraded land. LST patterns mirror this decline, with Sylhet Sadar, Dakshin Surma, Beanibazar, and Zakiganj experiencing the steepest warming and expanding high-severity thermal zones, especially during the monsoon. Correlation analysis further indicates that NDVI and LST exhibit a negative relationship across most agricultural seasons. During the Rabi season, NDVI and LST show a moderate negative correlation of r = −0.371, while during the Kharif I season, a similar inverse relationship is observed, with r = −0.344. However, the relationship becomes very weak during the Kharif II season, with r = −0.054, indicating that vegetation plays a reduced role in regulating surface temperature during the peak monsoon period. The correlation analysis indicates that several relationships among the seasonal vegetation and thermal variables are statistically significant. In particular, vegetation indices across the three cropping seasons show significant positive associations with one another, demonstrating that spatial patterns of vegetation condition remain relatively consistent throughout the agricultural cycle. Significant inverse relationships are also observed between vegetation condition and surface temperature during the Rabi and Kharif I seasons, suggesting that higher thermal conditions are linked with reduced vegetation health in these periods. However, the relationship between vegetation and temperature during the Kharif II season is weak and not statistically significant, likely due to the dominant influence of monsoon-driven moisture conditions. Therefore, it can be concluded that the interaction between LST and NDVI is generally significant in the dry and pre-monsoon seasons of the study area. The climatic conditions during the dates of satellite image acquisition provide important context for interpreting the observed NDVI and LST patterns. Monthly averages were calculated from daily temperature and precipitation data obtained from the NOAA to represent the climatic conditions corresponding to the satellite image acquisition periods.

For the Rabi season, the analysis used Landsat images acquired in February 2005 and 2025. February represents the late winter stage of the Rabi cropping cycle, when crops such as wheat, mustard, and vegetables approach maturity. The calculated NOAA climate averages indicate that in February 2005, the mean temperature was approximately 20.9 °C with very low precipitation of 0.058 mm, reflecting cool and relatively dry winter conditions. In February 2025, the average temperature increased slightly to about 21.5 °C, while rainfall remained extremely low at 0.0046 mm. Such climatic conditions typically support stable crop growth while maintaining relatively low land surface temperatures due to the cooler seasonal environment. The NDVI results correspond with this climatic background, where moderately healthy vegetation dominates, while the LST analysis shows mostly low-to-moderate thermal stress. Nevertheless, the slightly higher temperature in 2025 aligns with the observed gradual increase in thermal stress and vegetation weakening, suggesting that even during the cooler Rabi season, the region is experiencing subtle warming trends.

For the Kharif I season, images from June 2005 and 2025 were analyzed. June corresponds to the pre-monsoon to early monsoon transition period, typically characterized by high temperatures and increasing rainfall. According to the calculated NOAA monthly averages, June 2005 recorded an average temperature of approximately 28.1 °C with precipitation around 0.406 mm, while June 2025 showed a slightly higher average temperature of about 29.0 °C with precipitation around 0.467 mm. These conditions indicate a warm and humid environment associated with the onset of the monsoon, which increases evapotranspiration and surface heat accumulation. The climatic context strongly aligns with the LST results, where Kharif I shows the most pronounced increase in thermal stress, with substantial transitions from moderate to severe thermal stress categories. Similarly, the NDVI transition matrix indicates notable shifts from healthy to moderately healthy vegetation, suggesting that increasing seasonal heat combined with moisture variability during the early monsoon period contributes to vegetation stress and reduced canopy vigor.

For the Kharif II season, Landsat images from 8 October 2005 and 15 October 2025 were used. October typically represents the late monsoon or post-monsoon stage, when Aman rice crops approach maturity and vegetation cover is generally extensive. However, October climate data were not available, so monthly climatic averages for that period could not be calculated. Despite this limitation, the broader seasonal climate pattern of Sylhet suggests that temperatures remain warm while rainfall gradually declines after the monsoon peak, creating conditions that normally support dense crop canopies and moderate surface temperatures through evapotranspiration. Nevertheless, the NDVI results reveal noticeable vegetation weakening by 2025, particularly in eastern and central Sylhet. At the same time, the LST analysis indicates a clear shift from moderate to high thermal stress categories, implying that rising regional temperatures and vegetation degradation may reduce the natural cooling effect typically associated with dense vegetation during the late monsoon cropping period.

Overall, integrating the NOAA-derived climatic averages with the satellite observations provides important insight into the seasonal dynamics of vegetation health and surface temperature. The cool and dry conditions of February correspond with relatively lower thermal stress during the Rabi season, while the warmer and increasingly humid conditions of June align with the strong intensification of thermal stress observed during Kharif I. Although October climatic data were unavailable, the NDVI and LST patterns still suggest a growing thermal burden and declining vegetation resilience by 2025, highlighting the increasing vulnerability of Sylhet’s agricultural landscape to climate-driven thermal stress.

These results are consistent with broader evidence that areas with stronger vegetation cover maintain lower surface temperatures. These findings support broader evidence that areas with greater vegetation cover tend to maintain lower surface temperatures, highlighting the vulnerability of climate-sensitive agrarian landscapes to rising temperatures. The correlation analysis suggests that vegetation contributes to surface cooling primarily during drier or early monsoon periods, whereas atmospheric factors such as cloud cover, humidity, and rainfall exert stronger control over land surface temperature during peak monsoon months.

5. Limitations and Future Studies

Despite providing valuable insights into long-term vegetation dynamics and thermal stress patterns in the Sylhet region, several limitations should be acknowledged. The analysis relies exclusively on satellite-derived indicators without field-based validation. Although NDVI and LST derived from Landsat imagery are widely used proxies for vegetation health and surface temperature, the absence of ground-truth measurements limits the ability to directly verify the accuracy of the satellite-derived results. Integrating field observations, in situ vegetation measurements, or drone-based monitoring in future studies would improve the calibration and validation of remote sensing outputs. The study also focuses on general agricultural vegetation patterns and does not explicitly distinguish between specific crop species or include detailed phenological information. While the agricultural landscape of Sylhet includes crops such as rice, tea, and mustard, species-level vegetation data were not incorporated into the analysis. This constraint reduces the ability to interpret vegetation responses at the crop-specific level. Future research could incorporate species-level vegetation information and phenological datasets to enhance agronomic interpretation. In addition, the statistical analysis applied in this study remains relatively simple. The research primarily relies on Pearson’s correlation analysis and transition matrices to examine relationships between NDVI and LST across agricultural seasons. While these methods provide useful insights into general vegetation–temperature relationships, they do not account for spatial dependencies or complex temporal dynamics. The application of advanced analytical approaches, such as spatial autocorrelation analysis, time-series modeling, or machine learning techniques, could improve the robustness of future investigations. The study further relies on two remote sensing indicators, NDVI and LST, to assess vegetation health and thermal stress. Although these indicators are widely accepted and effective for vegetation monitoring, additional environmental variables such as soil moisture, precipitation variability, and anthropogenic land-use changes were not included in the analysis. Incorporating additional indices such as the Enhanced Vegetation Index (EVI), Soil Adjusted Vegetation Index (SAVI), and Vegetation Health Index (VHI) would provide a more comprehensive assessment of vegetation conditions. Another limitation relates to the spatial resolution of the imagery used in the analysis. Landsat data with a spatial resolution of 30 m are suitable for long-term regional analysis but may not capture finer-scale variations in vegetation patterns. The integration of higher-resolution datasets, such as Sentinel-2 imagery, could improve spatial accuracy and enable more detailed vegetation monitoring at the local scale. The research also primarily examines spatial patterns of vegetation and temperature change without explicitly incorporating socioeconomic drivers of land-use change or future climate scenario modelling. Integrating socioeconomic datasets and climate projections in future studies would provide a more comprehensive understanding of the mechanisms driving vegetation change in the region. Despite these limitations, the study provides a valuable long-term assessment of vegetation health and thermal stress dynamics in the Sylhet region and highlights the usefulness of remote sensing approaches for monitoring climate-related agricultural vulnerability in tropical monsoon environments.

6. Conclusions

This study demonstrates a clear and progressive decline in vegetation health across Sylhet over the past two decades, accompanied by a consistent intensification of land surface temperature during all agricultural seasons. Vegetation loss is most pronounced in areas exposed to strong monsoon dynamics, with Kharif II showing the greatest reduction in vegetation vigor and the weakest cooling influence from plant cover. By contrast, Rabi and Kharif I retain a stronger vegetation–temperature linkage, indicating that plant health still plays a substantial role in moderating surface heat during these periods. The combined NDVI and LST transitions reveal a landscape that is gradually shifting toward more stressed vegetation states and increasingly severe thermal conditions, highlighting reduced ecological resilience under changing climatic and land-use pressures. These patterns underscore the need for targeted policy interventions that strengthen vegetation recovery and minimize further thermal intensification. A progressive increase in thermal stress and declining vegetation vigor across Sylhet’s agricultural seasons can be observed. Kharif I emerges as the most vulnerable period, followed by Kharif II, while Rabi remains comparatively more stable due to stronger vegetation–temperature interactions. These findings highlight increasing climatic pressure on regional agriculture and emphasize the need for climate-resilient land management strategies. Priority actions should include preserving remaining natural vegetation, expanding agroforestry and urban greening initiatives, controlling unplanned urban expansion, and promoting climate-adaptive agricultural practices that maintain soil moisture and canopy cover. Strengthening these measures is essential to restoring vegetation stability, reducing seasonal thermal stress, and safeguarding Sylhet’s agricultural and ecological systems in the long-term. Future research should integrate higher temporal resolution, field-based measurements, and socioeconomic factors to better inform adaptive strategies across Bangladesh’s agroecosystems.