Global NDVI-LST Correlation: Temporal and Spatial Patterns from 2000 to 2024

Abstract

While numerous studies have investigated the NDVI-LST relationship at local or regional scales, existing global analyses are outdated and fail to incorporate recent environmental changes driven by climate change and human activity. This study aims to address this gap by conducting an extensive global analysis of NDVI-LST correlations from 2000 to 2024, utilizing multi-source satellite data to assess latitudinal and ecosystem-specific variability. The MODIS dataset, which provides global daily LST data at a 1 km resolution from 2000 to 2024, was used alongside MODIS-derived NDVI data, which offers global vegetation indices at a 1 km resolution and 16-day temporal intervals. A correlation analysis was performed by extracting NDVI and LST values for each raster cell. The analysis revealed significant negative correlations in regions such as the western United States, Brazil, southern Africa, and northern Australia, where increased temperatures suppress vegetation activity. A total of 38,281,647 pixels, or 20% of the global map, exhibited statistically significant correlations, with 80.4% showing negative correlations, indicating a reduction in vegetation activity as temperatures rise. The latitudinal distribution of significant correlations revealed two prominent peaks: one in the tropical and subtropical regions of the Southern Hemisphere and another in the temperate zones of the Northern Hemisphere. This study uncovers notable spatial and latitudinal patterns in the LST-NDVI relationship, with most regions exhibiting negative correlations, underscoring the cooling effects of vegetation. These findings emphasize the crucial role of vegetation in regulating surface temperatures, providing valuable insights into ecosystem health, and informing conservation strategies in response to climate change.

1. Introduction

Understanding the complex interplay between vegetation dynamics and climatic variables is critical for monitoring and predicting ecosystem responses to climate change. The Normalized Difference Vegetation Index (NDVI) and land surface temperature (LST) are two widely used satellite-derived parameters that have provided valuable insights into vegetation health, land cover changes, and climatic interactions across diverse ecosystems [1,2,3]. NDVI serves as a proxy for vegetation greenness and biomass, while LST reflects the surface thermal environment influenced by land cover and atmospheric conditions. The global spatiotemporal relationship between NDVI and LST has garnered significant attention in recent decades [4,5,6], given its relevance to understanding vegetation–climate feedback, assessing ecosystem health, and devising sustainable land management practices.

Numerous studies have explored the intricate relationships between NDVI and climatic factors, including temperature and precipitation, at varying spatial and temporal scales. For example, a study by Guha et al. [7] explored the spatial and temporal dynamics between LST and NDVI in India. Their analysis indicated a pronounced inverse correlation between LST and NDVI, though this relationship weakened in extreme temperature zones. Similarly, Hashim et al. [8] assessed urban heat island (UHI) effects and LST variations in Baghdad using Landsat datasets. Their study documented a temperature rise of 6.8 °C, with the hottest regions occurring in residential and barren landscapes, whereas lower temperatures were observed in water bodies and vegetated areas. Garai et al. [9] examined the interplay between rainfall, NDVI, and LST in India, employing MODIS datasets and Google Earth Engine (GEE). Their findings revealed a robust negative correlation between NDVI and LST, with an R² value of approximately 0.86.

Similarly, Jamei et al. [10] analyzed the effects of LULC changes on mean daytime LST in Melbourne during hot summers from 2001 to 2018 using GEE. Their analysis indicated that the mean daytime LST experienced a rise of 5.1 °C over the study period, with temperature increases of 5.5 °C observed in vegetated areas and 5.9 °C in built-up regions. Roy et al. [11] examined the relationship between LST and various landscape features in Sylhet Sadar Upazila. By incorporating NDVI, NDWI, NDBI, and NDBAI, they assessed vegetation, water bodies, built-up areas, and bare land. Their results indicated that NDVI exhibited a positive correlation with LST (+0.57), whereas NDBI had a negative correlation (−0.52). Ullah et al. [12] investigated the relationship between LST, NDVI, LULC, and topographic elements in Pakistan. Their study found that LST decreases with elevation, with built-up and bare soil areas exhibiting the highest temperatures, while NDVI increases with elevation. Naga Rajesh et al. [13] also explored long-term relationships between NDVI and hydrological-climatic variables in India, and reported a strong negative correlation (R = −0.91) between LST and NDVI.

Seasonal and diurnal variations in the NDVI-LST relationship have also been investigated. For instance, Sun et al. [14] conducted an in-depth analysis of the correlation between NDVI and LST across North America, demonstrating that the direction and strength of the relationship depend on the season and time of day. While NDVI positively correlated with LST during winter, a strong negative correlation was observed during the growing season, emphasizing the thermal regulation by vegetation in warmer months. Guha et al. [15] also investigated the seasonal dynamics of NDVI and LST in Raipur, India, revealing an inverse relationship between the two variables. Song et al. [16] examined interannual fluctuations in global LST and NDVI, linking these patterns to air pollution and vegetation cover changes. Kikon et al. [17] demonstrated the inverse relationship between LST and NDVI in Kohima, India, highlighting the potential of NDVI as an indicator of urban thermal dynamics.

Guha et al. [18] also assessed the seasonal variation of LST and its relationship with NDVI across different LULC types in India. The research indicated that land surface temperature was highest in barren and urbanized regions, whereas vegetated areas exhibited the lowest temperatures. The relationship between LST and NDVI fluctuated across seasons, displaying a pronounced negative correlation during the monsoon and post-monsoon periods, a moderate inverse association in the pre-monsoon season, and a weak negative correlation in winter. LULC types significantly influenced the correlation strength, with green vegetation showing a strong negative correlation, urban areas and bare land a moderate negative correlation, and water bodies exhibiting a nonlinear relationship.

Further advancements in understanding NDVI trends have been achieved through the integration of climate variables and human activity. Liu et al. [19] reported a global greening–browning–greening trend from 1982 to 2012, with temperature playing a diminished role in influencing NDVI over time. Their study suggested a weakened temperature effect on vegetation growth in certain regions, while precipitation continued to exhibit a positive impact on NDVI. Julien et al. [20] introduced the Yearly Land Cover Dynamics (YLCD) method, utilizing NDVI and LST data to characterize vegetation–climate interactions on an annual basis, thereby offering a novel approach to bioclimatic monitoring. Similarly, Yang et al. [21] employed advanced statistical methods, including Theil–Sen trend analysis and boosted regression trees, to examine NDVI trends from 1982 to 2015. Their results revealed widespread greening, particularly in the Sahel, Europe, India, and southern China, while browning trends were localized in regions such as Canada and Central Africa. Rainfall emerged as the dominant factor influencing vegetation evolution, underscoring the importance of water availability in sustaining ecosystem productivity. Recent studies have further emphasized the role of vegetation structure, including height and canopy characteristics, in modulating LST. Yu et al. [22] and He et al. [23] underscored the cooling effects of tree height and canopy cover, with significant negative correlations between tree canopy height and LST observed globally. Similarly, Shaik et al. [24] demonstrated that vegetation height significantly influences surface temperatures, with higher canopies associated with cooler surface conditions. These studies provide critical insights into the role of vegetation structure in alleviating urban heat stress and enhancing ecosystem resilience to climate change.

Despite extensive research on the association between NDVI and LST, most have been conducted at local or regional scales, focusing on specific ecosystems or climatic zones. These studies provide valuable insights into vegetation–climate interactions but lack a comprehensive global perspective. Furthermore, the few global-scale analyses available are outdated and do not reflect the significant environmental transformations of the past two decades, driven by climate change, land use changes, and intensified human activities. To address these gaps, this study conducts a global correlation analysis of NDVI and LST from 2000 to 2024, leveraging multi-source satellite data and robust statistical methods. By analyzing 24 years of data across diverse ecosystems and climatic zones, this research aims to identify regions where the NDVI-LST relationship is statistically significant and assess the latitudinal variability of these relationships. Unlike previous studies that primarily focused on localized trends, this study provides a comprehensive global-scale perspective, offering a more extensive understanding of how vegetation influences surface temperature across different environmental settings. The findings contribute to a more nuanced understanding of vegetation’s role in regulating LST on a global scale, with implications for climate adaptation, ecosystem management, and land use planning. By filling this critical knowledge gap, this research enhances the scientific foundation for assessing vegetation–climate feedback and provides a valuable resource for future studies addressing the complexities of worldwide ecological transformation.

2. Materials and Methods

2.1. Selection of MODIS LST Data

According to Patel et al. [3], the United States Geological Survey’s (USGS) Landsat program is the most predominant tool used to capture land use/land cover (LULC) data, accounting for 57% of the studies. MODIS follows closely with 12%, while studies integrating two or more tools for LULC data collection comprise another 12%. In the context of LST data, Phan et al. [2] analyzed 529 articles sourced from 159 journals in the Scopus database, spanning from 2009 to 2018. The findings indicate that the number of publications on MODIS LST data applications has been consistently rising year on year. These publications cover a broad array of topics, ranging from environmental and agricultural sciences to biological, social science, and even medicine, reflecting the versatility and widespread applications of MODIS LST data.

In this study, we utilized the Moderate Resolution Imaging Spectroradiometer (MODIS) MOD11A1 Version 6.1 (V6.1) LST product to analyze global temperature trends from 2000 to 2024. MODIS was chosen for its extensive temporal coverage, global reach, and high-quality retrievals of LST under clear-sky conditions. The MOD11A1 product provides daily LST and emissivity values at a spatial resolution of 1 km, making it well-suited for global-scale analysis. The dataset includes both daytime and nighttime LST bands, enabling a comprehensive assessment of temperature variations across different times of the day. The LST values in MOD11A1 are derived from the MOD11_L2 swath product, which aggregates and averages temperature readings under clear-sky conditions. To ensure data consistency, beyond 30° latitudes, where multiple observations may occur per day, the product provides an averaged LST value across all valid clear-sky readings for each pixel. This approach minimizes cloud contamination and improves the reliability of the dataset.

2.2. Thermal Infrared Observations and Quality Control

MODIS retrieves LST using thermal infrared (TIR) observations captured in bands 31 (10.78–11.28 µm) and 32 (11.77–12.27 µm), which are specifically designed for detecting land surface temperatures. The dataset includes multiple quality control layers that provide essential metadata on retrieval accuracy, cloud contamination, and atmospheric corrections. The thermal infrared retrieval algorithm applies radiative transfer models to account for emissivity variations and atmospheric interferences, ensuring that the extracted LST values maintain high accuracy.

2.3. LST Data Acquisition and Processing

The MOD11A1 dataset was accessed from NASA’s Land Processes Distributed Active Archive Center (LP DAAC) at the USGS EROS Center (https://lpdaac.usgs.gov/products/mod11a1v061/, accessed on 20 December 2024). The dataset’s daily temporal resolution allows for continuous monitoring of global temperature patterns, making it particularly valuable for detecting seasonal and interannual trends in surface temperature. For our analysis, we extracted MODIS LST data for the daytime observations, which were considered more relevant for examining the interaction between surface temperature and vegetation changes. Nighttime temperatures were excluded to maintain consistency with previous large-scale studies on LST-NDVI correlations, which primarily focus on the daytime thermal response of land surfaces. The combination of high spatial resolution (1 km), daily observations, and extensive temporal coverage (2000–2024) makes MODIS MOD11A1 an ideal dataset for our global-scale analysis. By leveraging this dataset, we aimed to assess spatiotemporal variations in LST and explore its correlation with NDVI across different climatic zones and land cover types.

2.4. Selection of MODIS NDVI Data

NDVI is a widely used remote sensing-derived parameter for assessing vegetation health, land cover changes, and ecosystem dynamics. It is extensively applied in environmental, agricultural, and ecological studies due to its ability to quantify vegetation greenness and monitor plant phenology over time. The increasing use of NDVI in research is reflected in bibliometric analyses, such as Xu et al. [1], who highlighted the significant expansion of NDVI-related studies from 1985 to 2021, driven by advancements in remote sensing technologies and improved data accessibility. For this study, we used the MODIS MOD13A2 Version 6.1 dataset to analyze NDVI trends globally from 2000 to 2024. MODIS was selected due to its high temporal resolution, consistent data availability, and global coverage, making it particularly suitable for large-scale vegetation monitoring. The MOD13A2 product provides vegetation indices at a spatial resolution of 1 km with a 16-day temporal frequency, ensuring the availability of high-quality, cloud-free observations for trend analysis.

The dataset includes two primary vegetation indices: NDVI and the Enhanced Vegetation Index (EVI). NDVI, which serves as a continuity index for the NOAA-AVHRR-derived NDVI, was chosen for this study due to its extensive historical usage and compatibility with previous vegetation–climate studies. The MOD13A2 product applies a rigorous data selection process, choosing the best available pixel values every 16 days based on criteria such as low cloud cover, minimal sensor view angles, and optimal pixel quality. This methodology minimizes atmospheric contamination and ensures that only the most reliable NDVI values are incorporated into the analysis.

2.5. NDVI Data Acquisition and Processing

The NDVI dataset was obtained from NASA’s Land Processes Distributed Active Archive Center (LP DAAC) at the USGS EROS Center and accessed via GEE using the Earth Engine Snippet (ee.ImageCollection(“MODIS/061/MOD13A2”)). The combination of global coverage, frequent temporal sampling, and rigorous data quality control makes MOD13A2 a highly valuable dataset for long-term vegetation monitoring. For our analysis, we extracted NDVI values corresponding to the spatial extent and temporal range of the MODIS LST dataset, ensuring a one-to-one comparison of both parameters. This allowed us to examine the spatiotemporal relationship between NDVI and LST across diverse biomes and climate zones, contributing to a comprehensive global-scale assessment.

2.6. Correlation Analysis

To assess the relationship between NDVI and LST from 2000 to 2024, we conducted a correlation analysis at the raster cell level. Each raster cell was treated as an independent unit, and for each cell, we extracted corresponding NDVI and LST values over the study period, resulting in a time series of 24 data points per variable. The correlation analysis aimed to identify whether vegetation greenness, as represented by NDVI, was statistically associated with variations in land surface temperature.

2.7. Handling Missing Data and Data Cleaning

Remote sensing datasets often contain missing values due to cloud cover, sensor malfunctions, and temporal inconsistencies in satellite observations. In our dataset, missing values were identified at various time points, primarily due to cloud cover affecting NDVI retrievals or atmospheric interference in LST measurements. Instead of interpolating missing values, which could introduce biases, we opted to exclude time points with missing observations from the correlation analysis. To maintain statistical reliability while preserving spatial coverage, we selected a minimum threshold of 10 valid data points per cell as a balanced compromise. Increasing this threshold (e.g., to 15 or 20) would significantly reduce the number of valid pixels, leading to incomplete spatial representation and potential geographic bias in the results.

Additionally, we considered statistical best practices for sample size determination in correlation analyses. Bujang [25] demonstrated that the required sample size for Pearson’s correlation varies based on effect size and confidence interval width, ranging from 12 to 143 (at a 95% confidence level). Similarly, Kendall’s Tau-b correlation requires 8 to 65, while Spearman’s rank correlation requires 16 to 149. These findings suggest that a sample size of at least 10 observations provides a reasonable statistical basis for estimating meaningful correlations, particularly in the context of remote sensing data where missing values are common. This threshold led to the removal of 509,925 cells from the final analysis due to missing values (NA) or having valid data, accounting for 0.26% of the total study area.

2.8. Computing Correlation Coefficients

Pearson correlation analysis was performed to quantify the strength and direction of the relationship between NDVI and LST for each raster cell. Pearson correlation was chosen because it effectively measures the linear association between two continuous variables and is widely used in remote sensing studies. The resulting correlation coefficients were mapped globally to assess spatial patterns and latitudinal variations in NDVI-LST relationships. By employing a rigorous approach to data selection, missing value handling, and correlation computation, this study ensures a reliable and comprehensive assessment of the NDVI-LST relationship at a global scale. The findings contribute to a deeper understanding of how vegetation dynamics influence surface temperature across different ecosystems and climatic regions.

3. Results

Figure 1 presents the correlation map of LST-NDVI categorized into six classes (a) and the distribution of significant versus non-significant pixels (b). As depicted in Figure 1a, a substantial portion of the map is dominated by classes exhibiting negative correlations between LST and NDVI. These negative correlations, represented in red, are predominantly concentrated in the western regions of the United States, Brazil, southern Africa, northern Australia, and countries such as Türkiye, Greece, Bulgaria, and Kazakhstan. This spatial pattern aligns with expectations, as negative LST-NDVI correlations often indicate areas where increasing temperatures suppress vegetation activity. However, not all pixels on the map exhibit statistically significant correlations at a p-value threshold of <0.05. To address this, we generated a separate classification of significant pixels, which is illustrated in Figure 1b. This visualization isolates areas where correlations are statistically robust, providing a clearer understanding of the regions where the relationship between LST and NDVI is most pronounced and reliable.

Figure 2a illustrates the latitudinal distribution of significant LST-NDVI correlations, revealing distinct patterns along the latitudinal gradient. The figure highlights two prominent peaks in the distribution. The first peak occurs between latitudes −40 and 0, indicating a notable concentration of significant correlations in the Southern Hemisphere’s tropical and subtropical regions. The second, more pronounced peak spans latitudes 30 to 55 in the Northern Hemisphere, encompassing temperate zones where the majority of significant correlations are observed. These peaks suggest that the relationship between LST and NDVI varies with latitude.

Our detailed analysis revealed that a significant proportion of the global LST-NDVI correlation map—specifically, 38,281,647 pixels (Figure 2b), equivalent to 20% of the entire map—exhibits statistically significant correlations. These significant pixels are distributed across four primary classes, highlighting distinct trends in the relationship between LST and NDVI. Class 1 ((−1)–(−0.6)) contains 11,032,462 pixels, Class 2 ((−0.6)–(−0.2)) encompasses 18,098,538 pixels, Class 5 ((0.2)–(0.6)) includes 7,333,981 pixels, and Class 6 ((0.6)–(1)) accounts for 1,816,665 pixels (Figure 2b). Analyzing the nature of these correlations, we find that 19.6% of the significant pixels demonstrate a positive relationship, indicating areas where higher LST is associated with increased NDVI. In contrast, the remaining 80.4% of the significant pixels reveal a negative correlation, suggesting that elevated temperatures may suppress vegetation activity or that vegetation loss leads to increased surface temperatures (Figure 2b).

4. Discussion

Our study reveals key insights into the spatial and latitudinal patterns of the relationship between LST and NDVI. Globally, approximately 20% of the map exhibits statistically significant correlations, amounting to 38,281,647 pixels. Among these, the majority (80.4%) show negative correlations, highlighting regions where increased surface temperatures likely suppress vegetation activity, or vegetation loss exacerbates surface warming. These negative correlations are predominantly observed in western parts of the United States, Brazil, southern Africa, northern Australia, and areas like Türkiye, Greece, Bulgaria, and Kazakhstan. Conversely, a smaller but notable proportion (19.6%) of significant correlations are positive, indicating regions where higher temperatures may enhance vegetation productivity. The latitudinal distribution of significant correlations further reveals two distinct peaks, with one spanning latitudes −40 to 0 in the Southern Hemisphere’s tropical and subtropical regions and a larger peak observed between latitudes 30 and 55 in the temperate zones of the Northern Hemisphere. These findings underscore the heterogeneous nature of LST-NDVI interactions across different biomes and climatic zones, highlighting the influence of geographical and ecological factors in shaping temperature–vegetation dynamics.

The relationship between vegetation patterns and LST has been widely explored, with many studies emphasizing the cooling effects of cohesive vegetation cover [26,27,28,29,30]. However, contrasting evidence suggests that these effects are not universally consistent. For instance, research by Li et al. [31], Maimaitiyiming et al. [32], and Bao et al. [33] observed negative effects, suggesting that in certain contexts or environments, aggregated vegetation might not consistently lead to LST reduction indicates that in certain contexts, aggregated vegetation might not always result in LST reductions.

Notably, studies focusing on urban green spaces and the spatial configuration of vegetation have yielded findings that align with our observations. For instance, Rahimi et al. [34] demonstrated that expanding green spaces can significantly reduce LST, a trend supported by the work of Guo et al. [35] and Li et al. [36]. Similarly, Masoudi et al. [37] and Masoudi et al. [38] reported negative correlations between area-based landscape metrics and LST in various urban environments. Guo et al. [35] further revealed inverse relationships between LST and some area-related landscape metrics, findings that resonate with the broader implications of this study. These varied results underscore the complexity of the LST–vegetation relationship and the need to consider contextual factors when interpreting these dynamics.

The NDVI-LST relationship is influenced by several confounding factors, including land use change, irrigation, and elevation effects, which require consideration for a more accurate interpretation. Land use changes, such as urbanization, deforestation, and agricultural expansion, significantly impact the NDVI-LST relationship by altering surface temperatures and vegetation cover [39,40]. Deforestation, particularly in tropical regions, exposes bare soil, increasing surface temperatures and strengthening negative correlations [41]. Conversely, large-scale afforestation and reforestation efforts, such as those in China, may mitigate warming and alter LST-NDVI relationships [42,43]. Elevation and topographic features, including aspect, can modulate thermal regimes, reinforcing the interplay of vegetation structure and environmental factors in shaping LST [44,45].

Our findings also suggest that the observed positive correlations between LST and NDVI in certain regions may be influenced by variations in vegetation structure, including tree height, canopy cover, and vegetation density, which play crucial roles in regulating LST. Recent studies have shown that tree shade, quantified through spatial tools such as the hillshade function, has a cooling effect on LST, with its effectiveness depending on the ratio of tree canopy to impervious surfaces, as demonstrated in Tampa and New York City [46]. Larger and taller vegetation patches exhibit stronger cooling effects, while increased building cover and height tend to elevate LST, highlighting the importance of vegetation height in mitigating surface temperatures [47]. Additionally, tree canopy temperature strongly correlates with ground and air temperatures beneath, emphasizing trees’ ability to regulate microclimatic conditions, as evidenced in Hong Kong [48].

Moreover, studies utilizing advanced tools like LiDAR and multispectral imagery reveal that vertical vegetation structures and land cover composition significantly influence LST patterns, particularly in ecosystems with complex topographies, such as Brazil’s Atlantic Forest [44,49]. Although most studies focus on urban areas, our results extend these insights to global patterns. For example, the strong negative correlation between tree height and LST observed in temperate regions aligns with findings from urban studies, such as those by Jung et al. [50] and Zhang et al. [51] which demonstrate significant cooling effects of taller forests.

The unexpected positive correlations between LST and NDVI, though less common (19.6% of significant pixels), provide intriguing insights into complex ecological dynamics. These positive relationships suggest areas where higher temperatures are associated with increased vegetation productivity, challenging the typical expectation of negative LST-NDVI correlations. Several factors may explain this phenomenon. In temperate or colder regions, increased temperatures can extend growing seasons, promoting vegetation growth by reducing temperature-related constraints [52]. Similarly, warmer temperatures within optimal ranges can enhance photosynthesis, particularly in ecosystems where water and other resources are not limiting factors. Shifts in vegetation composition, favoring heat-adapted or drought-resistant species, might also contribute to higher NDVI values despite rising temperatures. Additionally, positive correlations in irrigated or managed landscapes may reflect human interventions, such as irrigation or fertilization, sustaining or boosting vegetation productivity despite elevated surface temperatures. For example, Yang et al. [53]’s study found that irrigation in China leads to an average reduction of 1.15 K in daytime land surface temperature across irrigated regions.

The spatial distribution of these positive correlations, particularly in Class 5 (0.2–0.6) and Class 6 (0.6–1) pixels, aligns with biogeographical zones where temperature increases may temporarily benefit vegetation. For instance, in higher latitudes or historically cooler climates, moderate warming might reduce thermal limitations on plant growth. In arid or semi-arid regions, these correlations may indicate interactions with rainfall patterns, where temperature increases during rainy seasons enhance vegetation growth responses to water availability [54]. A recent investigation examined how the relationship between land surface temperature (LST) and the Normalized Difference Vegetation Index (NDVI) fluctuates spatially and seasonally across Europe. Using long-term MODIS datasets from 2000 to 2017, the study revealed substantial variability in the LST–NDVI association across different biomes. During the spring, significant positive correlations were observed in approximately 80% of the European region. However, this pattern changed in the summer, with most biomes—except those in the northernmost areas—displaying a negative correlation between LST and NDVI [52].

In Mongolia, research conducted by Karnieli et al. [55] found that the strength and direction of the LST–NDVI relationship differed based on biome type and latitude. Strong negative correlations were detected in desert and desert steppe ecosystems, whereas high-altitude regions such as the taiga and mountainous areas exhibited positive correlations. On a global scale, Schultz et al. [54] demonstrated that dryland ecosystems, including deserts, savannas, and woodlands—where water availability is a limiting factor—tend to show an inverse relationship between LST and NDVI. In contrast, equatorial regions dominated by broad-leaved humid evergreen forests, which are primarily constrained by energy availability, exhibited a positive correlation. Similarly, Lambin et al. [56] conducted a large-scale analysis across Africa and identified a predominantly negative LST–NDVI relationship across all biomes, except for evergreen forests, where the correlation shifted to positive.

5. Conclusions

This study provides a detailed analysis of the spatial and temporal variations in the NDVI-LST relationship, uncovering both anticipated and unexpected patterns across different biomes and climatic zones. The accuracy of these relationships may be affected by factors such as cloud cover, humidity, wind dynamics, and land cover variations, which influence MODIS data quality and introduce uncertainties in correlation estimates [57,58,59]. Furthermore, the 1 km spatial resolution of MODIS LST may not adequately capture localized temperature fluctuations caused by microclimatic conditions or topographic complexity [60,61]. Future studies could integrate higher-resolution thermal datasets, such as those from Landsat, to refine these analyses and improve the accuracy of temperature–vegetation interactions at local scales.

However, the widespread prevalence of negative correlations (80.4%) underscores the essential role of vegetation in regulating surface temperatures through evaporative cooling and shading effects. Despite the robustness of our analysis, several limitations must be considered. These insights have significant policy implications for climate-resilient land management. Local governments can use this information to design strategies that enhance vegetation cover in areas experiencing strong negative correlations, such as urban reforestation initiatives, green infrastructure development, and afforestation programs. In agricultural regions, sustainable water management and irrigation planning can help mitigate heat stress and maintain vegetation productivity despite rising temperatures. Additionally, conservation efforts in tropical and temperate forests should prioritize protecting existing vegetation to sustain their cooling effects and ecosystem services.

Future research should focus on exploring seasonal variations in NDVI-LST relationships, integrating additional environmental variables such as soil moisture and precipitation, and assessing how extreme climate events impact temperature–vegetation interactions. Machine learning approaches and high-resolution remote sensing data could further refine our understanding of the underlying mechanisms driving these patterns. By addressing these knowledge gaps, we can better inform land use policies and conservation strategies aimed at enhancing ecosystem resilience to climate change.