Long-Term Spatiotemporal Assessment of Land-Use Change, Drought Stress, and Vegetation Resilience in Alabama’s Black Belt: Implications for Sustainable Agricultural Resource Management

Abstract

Climate-induced drought and intensifying land-use pressures threaten ecosystem services and agricultural productivity, particularly in regions with distinctive soil and ecological characteristics. Alabama’s Black Belt, defined by its clay-rich soils and shaped by a legacy of plantation agriculture, uneven land tenure, and persistent socioeconomic disadvantage, is increasingly vulnerable to these interacting stressors. This study analyzes long-term (2000–2023) spatiotemporal patterns of Land Use Land Cover (LULC) change and vegetation response to drought to inform sustainable resource management. Multi-temporal Landsat imagery and National Land Cover Database (NLCD) products were used to quantify LULC dynamics. At the same time, vegetation condition and moisture stress were assessed using the Normalized Difference Vegetation Index (NDVI) and Normalized Difference Moisture Index (NDMI). Drought conditions were evaluated using the Standardized Precipitation Index (SPI) and the Standardized Precipitation Evapotranspiration Index (SPEI), which incorporates temperature-driven evaporative demand. Results indicate substantial landscape change, including declines in deciduous forest (−17.78%) and pasture/hay (−13.17%), alongside increases in medium-intensity developed land (+20.25%) and evergreen forest (+10.62%). Declining NDVI and NDMI values indicate increasing vegetation stress, particularly during prolonged droughts. Vegetation response exhibited a weak relationship with SPI (R = 0.37) but a stronger association with SPEI (R = 0.59), underscoring the importance of accounting for atmospheric water demand. These findings highlight the growing vulnerability of Black Belt ecosystems to coupled climate and land-use pressures and provide insights to strengthen climate-resilient agricultural management.

1. Introduction

Sustainable management of agricultural landscapes necessitates an in-depth understanding of the interplay between Land Use Land Cover (LULC) changes driven by human decisions and climate variability, and how these factors collectively influence ecosystem services such as soil moisture regulation, hydrologic buffering, carbon storage, habitat provision, and agricultural stability [1].

Since 2000, global drought frequency and severity have risen by approximately 30%, adversely impacting over two billion people and leading to an estimated 20% decline in crop yields in rainfed systems across sub-Saharan Africa and South Asia. Simultaneously, land-use changes, especially the conversion of natural ecosystems into agricultural and urban areas, have accelerated dramatically, with an annual loss of around 4.9 million hectares of forest from 2010 to 2020. These interlinked challenges highlight the critical need for integrated assessments that explore the dual influences of land transformation and climatic variability on vegetation resilience and ecosystem functioning. Remote sensing and Geographic Information Systems (GISs) have become vital for systematically analyzing LULC dynamics across different spatial and temporal scales [2,3,4,5].

LULC dynamics are significantly shaped by anthropogenic activities that continually reconfigure natural landscapes, leading to both immediate and long-term ecological ramifications. Numerous investigations have focused on the impacts of urbanization and agricultural expansion on forest ecosystems, elucidating how these transformations disrupt hydrological cycles and degrade essential ecosystem services vital for biodiversity and ecological equilibrium. The conversion of native forests to agricultural land imposes substantial biophysical changes with extensive ramifications for water and energy cycles, crucial for ecosystem functionality and sustainability [5].

LULC changes represent a persistent and dynamic process influenced by both human actions and natural phenomena. Transformations such as urban development, agricultural intensification, and deforestation exert considerable pressure on local and regional ecosystems, affecting water quality, biodiversity, microclimates, and hydrological processes. To track and analyze these complex dynamics, researchers increasingly leverage advanced geospatial technologies, particularly remote sensing and GIS. Remote sensing facilitates multi-temporal, large-scale imagery that enables comprehensive detection and analysis of landscape transformations over time [6].

The Alabama Black Belt serves as a strategically significant landscape for investigating coupled land-use and climate interactions. This region, characterized by fertile clay soils and a long-standing agricultural history predominantly featuring cropland and forest systems, is currently undergoing significant transitions influenced by urban expansion, changing agricultural practices, and increasing climate variability. Alterations in precipitation patterns, including shifts in seasonal distribution, intensity, and frequency, present considerable challenges for agricultural productivity and water resource management. As changes in land cover affect surface albedo, evapotranspiration rates, and soil moisture retention, ongoing LULC modifications are poised to alter local hydrologic and microclimatic conditions, thereby exacerbating the region’s sensitivity to rainfall variability [7].

Satellite and remote sensing data are crucial for validating and calibrating models related to atmospheric and land cover processes. By offering continuous and extensive observations of Earth’s surface and atmosphere, these technologies ground model outputs, whether pertaining to land use changes, atmospheric composition, or climate trends, enhancing their reliability for scientific research, policy formulation, and resource management [8,9,10]. The synergy of multi-sensor remote sensing and GIS provides a robust framework for assessing LULC dynamics, integrating optical and radar data with historical GIS datasets to facilitate detailed classification and analysis [11,12,13].

Numerous studies have investigated LULC dynamics and drought impacts in the Southeastern United States. However, few have integrated these aspects over extended time frames while considering both precipitation-centric and temperature-sensitive drought metrics. For example, Preetha et al. [14] explored the short-term impacts of LULC changes and climate variability on water quality in Alabama’s Cahaba River watershed, revealing significant interactions that altered sediment and nutrient loads. Govender et al. 2022 [15] highlighted advancements in remote sensing for hydrological modeling related to LULC and climate variability, while also noting persistent gaps in evapotranspiration estimation. While valuable, these studies tend to isolate water quality issues from vegetation responses or fail to quantify the interactive effects of LULC and climate on vegetation health. Moreover, many regional assessments lean on precipitation-only drought indices, risking an underestimation of drought stress in a warming climate, where atmospheric water demand is an increasing factor.

In a related context, Barakat et al. utilized ASTER and Sentinel-2A data to evaluate forest cover dynamics in Morocco, reporting a net increase in wooded areas over time [16]. This study focuses on the Alabama Black Belt, classifying LULC changes using the Maximum Likelihood method, an established supervised classification algorithm known for its statistical rigor and accuracy [17,18]. Drought conditions were examined using the Standardized Precipitation Index (SPI), conceptualized by McKee, Doesken, and Kleist in 1993 [19], which remains a cornerstone metric for assessing wet/dry periods purely based on precipitation data [20]. Landsat satellites, along with their distinct spatial, spectral, and temporal resolutions, provide critical data for monitoring vegetation dynamics [21], utilizing the Normalized Difference Vegetation Index (NDVI) derived from surface reflectance as a key indicator for evaluating vegetation health and drought impacts [4,22].

This research aims to address existing gaps through three key advancements:

• Integration of LULC with Longitudinal Drought Metrics: We integrate LULC trajectories spanning 2000–2023 with both the SPI (reflecting precipitation anomalies) and the Standardized Precipitation Evapotranspiration Index (SPEI), which accounts for temperature-driven evaporative demand. This dual-index framework allows for an assessment of which metric more effectively characterizes vegetation stress under evolving climatic conditions.
• Distinguishing Greenness Declines from Moisture Stress: We employ paired analyses of NDVI and the Normalized Difference Moisture Index (NDMI) to evaluate vegetation conditions. By analyzing these indices collaboratively, we can better discern whether reductions in vegetation are due to shifts in biomass or are primarily attributable to moisture stress, resulting in more nuanced ecological interpretations.
• Spatially Explicit Assessment of Land–Climate Interactions: Using multi-temporal Landsat imagery and a consistent supervised classification framework, we quantify LULC transitions across the entire Alabama Black Belt at high spatial resolution. Additionally, we investigate the co-variability of these transitions with SPI and SPEI across multiple temporal scales, leading to a spatially explicit understanding of where and how land-use and climate drivers interact to influence vegetation resilience.

Through this integrated assessment, the aim is to establish a robust evidence base that informs climate-resilient land and water management strategies. This includes promoting drought-tolerant cropping systems, forest retention or afforestation initiatives, riparian buffer restoration, and enhancing demand-side water governance. Ultimately, this study contributes to broader efforts to sustain agricultural productivity and ecosystem services in regions facing compounded land–climate stressors.

2. Materials and Methods

2.1. Study Area

The Alabama Black Belt region serves as a critical case study for examining LULC dynamics and rainfall variability. Pedologically defined by its fertile, dark clay soils (Vertisols and Alfisols), the Black Belt has a deep agricultural history. While opportunities in manufacturing, capital, and development have been historically limited, climate-related changes may amplify modern pressures, including urban development and evolving agricultural practices.

The area experiences notable rainfall variability, with potential shifts in precipitation patterns, intensity, and timing that have significant implications for its agriculturally dependent economy and water resources. The interplay between LULC changes and rainfall variability is an important environmental concern, as land cover modifications can alter surface albedo, evapotranspiration rates, and soil moisture, which in turn affect local weather and hydrologic regimes [7,23].

The Black Belt forms part of a broader corridor of fertile soils extending from Texas to Virginia and is often described by the paradoxical phrase “the richest soil and the poorest people” in the United States. The term “Black Belt” originally referred to the region’s exceptionally fertile, dark-colored soils, which attracted settlers during the 1820s and 1830s and facilitated the development of a widespread cotton plantation economy [24]. These plantations thrived due to the soil’s productivity and the labor of enslaved populations, as illustrated in Figure 1.

Alabama contains approximately 23.0 million acres of timberland, ranking third nationally in commercial forestland. Including reserved timber regions such as wildlife refuges and wilderness areas, the total forestland expands to 23.1 million acres [26]. Of this, 93.1% are privately owned, while only 6.9% is public land. The forest industry manages about 6.5% of timberland, whereas non-industrial private owners control 86.6%. Historical data from the 1972 Forest Inventory and Analysis (FIA) survey indicate that forest industries previously held a larger share, 19.7% of timberland acreage [27]. Ownership of croplands and forests is generally concentrated due to longstanding social and political structures, including racial exclusion, unequal land tenure systems, and chronic underinvestment [28]. Within the Black Belt, African Americans represent 26% of Alabama’s population but own only 4% of non-industrial private forest acreage, much of which lies in this region. These ownership patterns have remained largely unchanged in recent years.

Ecologically, Alabama’s forests comprise 47% softwood stands dominated by pines, 41% hardwood stands featuring oaks and other species, and 12% mixed hardwood/pine stands. Softwood acreage has steadily increased over time. Despite these natural resources, the Black Belt faces severe socioeconomic challenges, including a poverty rate of 34.9%, nearly double the state average of 18.8% and far above the national average of 13.3%. Residents often experience shorter life expectancies, lower educational attainment, and reduced per capita income compared to other Alabama counties [29].

2.2. Materials

All LULC analyses in this study were conducted using Landsat-derived NLCD products at a spatial resolution of 30 m, obtained from the U.S. Geological Survey (USGS). Vegetation indices, namely the Normalized Difference Vegetation Index (NDVI) and Normalized Difference Moisture Index (NDMI), were generated from Landsat TM, ETM+, and OLI surface reflectance collections to ensure temporal continuity and comparability across sensor generations. Meteorological drought indicators were derived from CHIRPS precipitation data for the Standardized Precipitation Index (SPI), while the Standardized Precipitation Evapotranspiration Index (SPEI) was computed from climatic water-balance estimates (precipitation minus PET). Importantly, no MODIS land-cover products were used in any quantitative component of the analysis [30]. To quantify LULC dynamics across the Alabama Black Belt from 2000 to 2023, we used multi-temporal Landsat imagery consistent with NLCD classification standards. These datasets were acquired through the USGS Earth Explorer platform (https://earthexplorer.usgs.gov; accessed 18 August 2025) and include observations from the Landsat 5 Thematic Mapper (TM), Landsat 7 Enhanced Thematic Mapper Plus (ETM+), and Landsat 8 Operational Land Imager (OLI) sensors. Images were selected based on acquisition timing, cloud coverage, and radiometric quality to maintain consistency and minimize temporal bias in the derived land-cover assessments [31]. Meteorological drought conditions were evaluated using long-term, high-resolution precipitation records from CHIRPS, a quasi-global rainfall product that integrates satellite-based observations with in situ station measurements. SPI values were computed by fitting probability distribution to historic precipitation and standardizing deviations from the long-term mean. The CHIRPS product, with an approximate spatial resolution of 5 km and more than four decades of continuous daily observations, provides robust temporal depth for establishing climatic baselines across multiple timescales (e.g., 3-month and 12-month). This temporal flexibility enables precise characterization of drought frequency, duration, and intensity within the Black Belt region [32].

2.3. Methods

The methodology employed to analyze land cover changes and vegetation dynamics is comprehensively illustrated in Figure 2. This approach effectively integrates several key components: remote sensing data preprocessing, detailed image classification, and rigorous GIS-based spatial analysis.

In selecting the appropriate Landsat imagery for this study, careful consideration was given to the availability of high-quality images specific to the study area [33,34]. The imagery was obtained from the USGS Earth Explorer platform, a widely utilized resource for accessing satellite images, and subsequently processed using the Google Earth Engine (GEE) platform.

• Preprocessing Steps:
  Preprocessing is a critical phase, encompassing several essential techniques aimed at preparing the remote sensing data for analysis. The preprocessing included:

-

Radiometric Correction: This step was performed using the Radiometric Calibration Extension, which adjusts the pixel values in the images to compensate for various factors such as sensor inconsistencies and variations in atmospheric conditions that can lead to inaccurate readings.

-

Atmospheric Correction: This was implemented using the Dark Object Subtraction (DOS) model through the Semi-Automatic Classification Plugin (SCP). The atmospheric correction is particularly vital, as it ensures that the information extracted from the images is not distorted due to atmospheric variability, thus enhancing the reliability of the data.

To define the study area based on basin boundaries precisely, an image mosaic technique was employed within the ArcGIS Pro 3.6 GIS software [35,36]. This procedure involved stitching together multiple images to create a comprehensive and seamless representation of the landscape.

2.3.1. Pre-Processing of Remote Sensing Data

Radiometric and atmospheric corrections are essential preprocessing steps that enhance the quality and analytical reliability of remotely sensed data. These corrections calibrate raw imagery to account for sensor-related inconsistencies and atmospheric effects, thereby improving data accuracy for subsequent analyses [37]. Landsat images were processed using the Google Earth Engine (GEE) platform and corrected for geometric and radiometric distortions to ensure temporal consistency across acquisition [38]. Radiometric correction compensated for variations in solar geometry, sensor calibration, and viewing geometry, while atmospheric correction was performed using the Dark Object Subtraction (DOS) algorithm implemented through the Semi-Automatic Classification Plugin (SCP) to minimize atmospheric interference [21].

To maintain radiometric consistency throughout the multi-temporal Landsat time series (2000–2023), all images were processed into the Landsat Surface Reflectance product (Collection 2 Level 2), produced by the USGS. This product implements uniform atmospheric correction and cross-sensor calibration algorithms to mitigate systematic discrepancies among Landsat 5 TM, Landsat 7 ETM+, and Landsat 8 Operational Land Imager (OLI).

▪

Atmospheric Correction: For Landsat 5 TM and Landsat 7 ETM+, the LEDAPS (Landsat Ecosystem Disturbance Adaptive Processing System) algorithm was deployed. LEDAPS applies the 6S radiative transfer model to adjust for atmospheric scattering and absorption, utilizing ancillary climate data to incorporate water vapor, ozone, and aerosol optical depth. Conversely, for Landsat 8 OLI, the LaSRC (Land Surface Reflectance Code) algorithm was employed, which utilizes a similar radiative transfer methodology but is tailored for the specific spectral bands of the OLI sensor and includes a comprehensive cloud-masking routine.

▪

Cross-Sensor Calibration: The Collection 2 Level 2 product harmonizes surface reflectance values across sensors through spectral band-pass adjustment factors, determined from overpass coincidences and pseudo-invariant calibration sites. This calibration process ensures that reflectance values for corresponding spectral bands (e.g., red, near-infrared, shortwave infrared) are directly comparable between TM, ETM+, and OLI, thereby eliminating biases that might otherwise be misinterpreted as land cover changes or trends in vegetation.

▪

Cloud and Shadow Masking: All images underwent screening for clouds and cloud shadows utilizing the CFMask algorithm, which is part of the Collection 2 processing workflow. Pixels identified as cloud or shadow were excluded from further classifications and vegetation index computations to prevent erroneous signals.

By implementing this standardized preprocessing workflow, we ensured that observed temporal variations in LULC and vegetation indices genuinely reflect actual ground-level changes, rather than artifacts associated with sensor discrepancies. We selected 5–8 images per year from the June–September growing season with cloud cover limited to less than 10%, thus assuring consistent phenological and atmospheric conditions throughout the 24-year study period.

2.3.2. Image Classification

All reported LULC statistics and mapped outputs in this study are derived exclusively from NLCD Landsat-based products, following established national methodologies and mapping protocols [39,40]. The NLCD annual product employs a 16-class categorical system based on the Anderson Level II classification scheme, enabling consistent land-cover mapping across time and space [41]. NLCD classes and their standardized color conventions (

Table 1. RGB Color Values for Land Cover [42].

Pixel Value	Color Table RGB Value	Land Cover Class
250	0, 0, 0	No Data
11	70, 107, 159	Open Water
12	209, 222, 248	Perennial Ice/Snow
21	222, 197, 197	Developed, Open Space
22	217, 146, 130	Developed, Low Intensity
23	235, 0, 0	Developed, Medium Intensity
24	171, 0, 0	Developed, High Intensity
31	179, 172, 159	Barren Land (Rock/Sand/Clay)
41	104, 171, 95	Deciduous Forest
42	28, 95, 44	Evergreen Forest
43	181, 197, 143	Mixed Forest
52	204, 184, 121	Shrub/Scrub
71	223, 223, 194	Grassland/Herbaceous
81	220, 217, 57	Pasture/Hay
82	171, 108, 40	Cultivated Crops
90	184, 217, 235	Wood Wetlands
95	108, 159, 184	Emergent Herbaceous Wetlands

) were applied directly to quantify land-cover distributions, transitions, and long-term landscape dynamics within Alabama’s Black Belt.

Exploratory Maximum Likelihood (ML) classification tests were conducted on representative Landsat scenes solely to evaluate spectral separability among selected land-cover classes. These pilot analyses, documented in Supplementary Methods S1, were not used to generate the final LULC maps or statistical results. All reported LULC change statistics and mapped outputs are therefore derived exclusively from NLCD products, ensuring transparency and consistency in data sources and analytical procedures [42].

The NLCD product characterizes the dominant land-cover conditions for each mapping year by classifying landscape features into generalized categories reflecting both natural and anthropogenic origins. The underlying Anderson framework of the NLCD was established to:

• Ensure alignment with the land-cover classification systems employed by U.S. federal agencies.
• Facilitate clear distinctions between land-cover categories that can be interpreted from multispectral remote sensing data.
• Uphold a logical hierarchical structure across various thematic levels.

RGB color assignments for each NLCD class are detailed in Table 1 and visually illustrated in Figure 3, which aids in ensuring consistent interpretation and facilitates comparison across different spatial and temporal dimensions [42].

The classification process assigned pixels within predefined polygons to derive homogeneous spectral signatures, each corresponding to a distinct land-cover class. Training samples were validated using visual interpretation of Google Earth imagery and available field survey data [43].

2.3.3. Accuracy Assessment

Accuracy assessment is essential for validating classification results and quantifying errors, which commonly arise from spectral similarity among land-cover classes [44]. Evaluation was conducted using a confusion matrix that compares classified outputs with reference data [45]. This framework enables the calculation of key performance metrics, including overall accuracy, producer’s accuracy, user’s accuracy, and the Kappa coefficient, which accounts for agreement due to chance. Collectively, these measures support the identification of omission and commission errors and provide a robust basis for assessing classification reliability and limitations [46].

To account for spatial dependence and land-cover heterogeneity, correlation analyses were supplemented with advanced spatial methods, including global Moran’s I (queen contiguity, 999 permutations) for NDVI, NDMI, SPI, and SPEI; spatial lag and spatial error models relating annual NDVI to SPI and SPEI using row-standardized weights; geographically weighted regression with adaptive bi-square kernels optimized via AICc; spatial block cross-validation (50 km blocks); and NLCD class-stratified correlations. Class-specific 95% confidence intervals were estimated using block bootstrap procedures, and partial correlations were computed with NLCD class controls.

The accuracy assessment is based on reference data that reliably represent actual land cover conditions, serving as the benchmark for evaluating classification performance. Errors were quantified using a confusion matrix generated in ENVI 5.3, structured as a square contingency table in which rows represent classified outputs and columns represent reference land cover classes. Misclassifications are identified in the off-diagonal elements, enabling systematic evaluation of classification errors and the calculation of standard accuracy metrics [47,48,49].

From the confusion matrix, key performance measures—including overall accuracy, producer’s accuracy, user’s accuracy, and the Kappa coefficient—were derived. The Kappa statistic, ranging from 0 to 1, quantifies agreement between classified and reference data beyond chance, with values greater than 0.80 indicating strong agreement, values between 0.40 and 0.80 reflecting moderate agreement, and values below 0.40 indicating low classification reliability [50,51,52].

2.3.4. NDVI Index

Landsat imagery was used to assess temporal variations in vegetation cover from 2000 to 2023, accounting for the vegetation and soil characteristics of the study area. Vegetation condition was quantified using the Normalized Difference Vegetation Index (NDVI), a spectral indicator widely applied to evaluate vegetation health and density based on surface reflectance properties [53].

NDVI is calculated from the near-infrared (NIR) and red bands of the electromagnetic spectrum [54,55,56] using the following expression:

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

(1)

For Landsat 5 MSS, the near-infrared (NIR) and red bands correspond to bands 7 (0.8–1.1 µm) and 5 (0.6–0.7 µm), respectively. For Landsat TM and ETM+, NIR and red wavelengths are represented by bands 4 and 3, while for Landsat 8 OLI, they correspond to bands 5 (0.85–0.88 µm) and 4 (0.64–0.67 µm). These spectral bands are used to compute the Normalized Difference Vegetation Index (NDVI), which ranges from −1 to 1. Higher NDVI values indicate dense, healthy vegetation, whereas lower values reflect sparse or non-vegetated surfaces [57].

NDVI values below zero typically represent water or bare surfaces. Values between 0.15 and 0.30 denote moderate vegetation density, commonly associated with agricultural areas, while values greater than 0.30 indicate high vegetation density, characteristic of forested landscapes.

2.3.5. NDMI Index

The Normalized Difference Moisture Index (NDMI) is a widely used remote sensing indicator for assessing vegetation water status and soil moisture, particularly in drought-prone regions. It is derived from satellite measurements of near-infrared (NIR) and shortwave infrared (SWIR) reflectance (Equation (2)) [58].

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

(2)

NDMI exploits the contrasting spectral responses of vegetation, where healthy plants exhibit high NIR reflectance and low SWIR reflectance due to water absorption. Consequently, higher NDMI values indicate greater vegetation moisture and vigor, while lower values reflect moisture stress, senescence, or drought-related decline [59].

2.3.6. SPI Index

The Standardized Precipitation Index (SPI), developed by McKee et al. (1993) [19], is a widely used indicator for quantifying meteorological drought. It evaluates precipitation anomalies across multiple time scales by standardizing observed precipitation relative to long-term historical records, enabling the assessment of both short-term and long-term drought conditions. Owing to its robustness and applicability across diverse climatic regions, SPI is commonly employed in hydrological and agricultural drought monitoring [60]. As noted by Bordi et al. (2011) [61], the SPI is favored for its reliability and ease of use across diverse climatic zones. In this study, SPI values were computed in Python 3.12 using CHIRPS precipitation data spanning 2000–2023. Drought conditions were identified when SPI values were ≤−1.0, indicating significant precipitation deficits (

Table 2. SPI index classes [62].

Degree of Drought	SPI Classes
Exceptional wet	SPI > 2
Severe wet conditions	1 < SPI < 2
Medium wet condition	0 < SPI < 1
Slight drought	−1 < SPI < 0
Severe drought	−2 < SPI < −1
Exceptional drought	SPI < −2

) [61]. The SPI is defined as:

$$SPI={\displaystyle \frac{P-Pm}{\sigma p}}$$

(3)

where P is the observed precipitation for a given period (mm), P_m is the long-term mean precipitation, and σ_p is the corresponding standard deviation.

2.3.7. SPEI Index

The Standardized Precipitation Evapotranspiration Index (SPEI) is a comprehensive drought metric that integrates precipitation and potential evapotranspiration (PET) to characterize drought intensity and duration. By accounting for atmospheric water demand, SPEI provides a more robust representation of drought conditions than precipitation-only indices, particularly under variable and warming climatic regimes [63]. SPEI is based on the standardized climatic water balance (Equations (4) and (5)).

$$SPEI={\displaystyle \frac{D-Dm}{\sigma D}}$$

(4)

$$D=P-PET$$

(5)

where D represents the difference between precipitation (P) and PET, D_m is the long-term mean water balance, and $\sigma D$ is its standard deviation.

PET was calculated using the Hargreaves–Samani method, which requires only temperature data and extraterrestrial radiation. Temperature data (daily minimum and maximum) were obtained from the PRISM Climate Group (4 km resolution). The Hargreaves–Samani equation is:

PET = 0.0023 × RA × (T_mean + 17.8) × (T_max − T_min)^0.5

where RA is extraterrestrial radiation (converted from solar radiation), Tmean is the mean daily temperature, and T_max and T_min are the daily maximum and minimum temperatures. This method was selected for its robustness in data-sparse regions and its demonstrated performance in the southeastern United States. The 3-month SPEI was calculated using the SPEI package in R (v1.7), fitting a log-logistic distribution to the climatic water balance (P − PET) for each calendar month.

SPEI is computed by fitting the climatic water balance (precipitation minus PET) to an appropriate probability distribution and standardizing the results to enable comparison across temporal scales. Positive SPEI values indicate wet conditions, whereas negative values represent increasing drought severity, as defined by established thresholds (moderate: SPEI < −1.0; severe: SPEI < −1.5; extreme: SPEI < −2.0).

In Alabama’s Black Belt, SPEI provides critical insight into the combined effects of precipitation variability and temperature-driven evapotranspiration. Although the region’s clay-rich soils possess high water-holding capacity, prolonged dry periods intensify vegetation water stress. Analysis of SPEI at multiple temporal scales (e.g., 3- and 12-month) therefore captures both short-term agricultural drought and longer-term hydrological deficits.

3. Results and Discussion

LULC classification maps were derived from multi-temporal Landsat NLCD data spanning 2000–2023, providing a consistent representation of landscape dynamics. Classification was performed using the Maximum Likelihood algorithm, a statistically robust supervised approach that assigns pixels to land-cover classes based on spectral probability. The resulting maps, presented in Figure 4, provide a consistent and reliable depiction of landscape dynamics across the study period.

For each target year in the study area, NLCD land cover maps were generated based on Landsat imagery with a spatial resolution of 30 m. This approach allows for a consistent, year-on-year evaluation of landscape alterations from 2000 to 2023. The extracted annual NLCD layers were harmonized using the Anderson Level II classification system (refer to Table 1) and subsequently clipped to the boundaries of the Alabama Black Belt. These layers were then utilized to calculate land cover class extents and transitions. The resulting time series delivers a detailed and spatially explicit record of LULC dynamics, thereby supporting the quantitative analyses presented in

Table 3. Data source: Landsat-based NLCD (30 m). Areas and percent changes computed for the study-area boundary for 2000–2023.

Land Cover Types	Hectares_2000	Hectares_2023	Change_ha	% Change
Background/No Data	3,581,141.04	3,581,141.04	0	0
Open Water	67,963.77	68,096.07	132.3	0.19
Developed, Open Space	124,515.27	126,036.54	1521.27	1.22
Developed, Low Intensity	32,676.84	34,085.07	1408.23	4.31
Developed, Medium Intensity	10,492.38	12,616.83	2124.45	20.25
Developed, High Intensity	3302.46	4180.86	878.4	26.6
Barren Land	6389.64	6615.54	225.9	3.54
Deciduous Forest	442,049.58	363,455.19	−78,594.39	−17.78
Evergreen Forest	889,341.75	983,798.64	94,456.89	10.62
Mixed Forest	655,043.13	625,555.71	−29,487.42	−4.5
Shrub/Scrub	250,503.39	260,432.19	9928.8	3.96
Grassland/Herbaceous	113,975.1	193,803.66	79,828.56	70.04
Pasture/Hay	665,392.68	577,787.31	−87,605.37	−13.17
Cultivated Crops	77,294.79	83,507.4	6212.61	8.04
Woody Wetlands	660,494.7	643,453.74	−17,040.96	−2.58
Emergent Herbaceous Wetlands	33,435.72	49,446.45	16,010.73	47.89

and informing the spatiotemporal patterns discussed in the subsequent sections.

In the lower right corner of each map, the legend presents a standardized set of color-coded land-cover classes, including forest types (evergreen, deciduous, and mixed), shrubland, savanna, wetlands, grassland, urban areas, agricultural lands, barren surfaces, water bodies, and unclassified areas.

This multi-decadal dataset enables systematic analysis of land-cover transitions from 2000 to 2023, supporting the identification of key processes such as deforestation, reforestation, urban expansion, agricultural conversion, and hydrological change. Examination of the spatial and temporal patterns captured in these maps provides critical insight into the environmental and anthropogenic drivers shaping regional landscape dynamics.

LULC maps were generated from multi-temporal Landsat imagery using the Maximum Likelihood (ML) supervised classification algorithm, which assigns land-cover classes based on spectral probability distributions. Quantitative outcomes of the classification are summarized in Table 3.

Table 3 summarizes land-cover changes in the study area between 2000 and 2023 in terms of area and percentage change. All developed land categories expanded during this period, with medium-intensity development showing the largest increase (20.25%; 2124.45 ha). Evergreen forest and cultivated cropland also increased by 10.62% and 8.04%, respectively, indicating a shift in dominant land-use patterns.

Classification accuracy was high, with overall accuracies of 90% for 2000 and 93% for 2023 and corresponding Kappa statistics of 0.84 and 0.87, reflecting strong agreement beyond chance. Changes exceeding class-specific uncertainty thresholds, such as the 17.78% decline in deciduous forest and the 10.62% increase in evergreen forest, are therefore considered robust. In contrast, minor changes, including the 0.19% increase in open water, fall within expected classification error and should be interpreted cautiously.

Deciduous forest and pasture/hay experienced substantial losses, with deciduous forest declining by 78,594.39 ha (−17.78%), underscoring a significant reduction in this land-cover type. Overall, the results indicate a clear trend toward urban expansion and agricultural development, often at the expense of forested and pasture landscapes, highlighting the need for sustainable land-management strategies.

Forests and croplands remain central to the ecological and socioeconomic sustainability of the Alabama Black Belt. Forests support biodiversity, carbon sequestration, hydrological regulation, and soil conservation while providing important economic benefits through timber and non-timber products. Croplands, in turn, are essential for food security and rural livelihoods in a region historically shaped by intensive agriculture [25,64].

Classification accuracy for the years 2000, 2010, and 2023 were evaluated using confusion matrices that compared mapped land-cover classes with ground-truth reference data. Overall accuracies of 90%, 82%, and 93%, along with Kappa coefficients of 0.8376, 0.7266, and 0.8745, indicate substantial to near-perfect agreement and demonstrate the robustness of the classification results. These metrics confirm the reliability of the land-cover datasets for analyzing spatial and temporal dynamics and for supporting environmental monitoring and land-management decisions [65].

Given the ecological importance of these land-cover types, continued monitoring of land-use change, particularly the conversion of forested areas to croplands, remains essential. Remote sensing provides an effective means of quantifying forest loss and improving understanding of land-use dynamics to inform sustainable land-management strategies.

Figure 5 presents a forest-loss map for Alabama’s Black Belt, illustrating year-by-year changes in forest cover from 2000 (yellow) to 2023 (red). The widespread presence of loss indicators across the region reflects substantial forest decline over this period. Statewide data indicate that Alabama lost approximately 3.45 million hectares of tree cover between 2001 and 2024, representing a 38% reduction since 2000 [66]. Although losses within the Black Belt may differ in magnitude, these data confirm a consistent pattern of persistent forest reduction. Primary drivers include timber harvesting, agricultural expansion, and urban development, with resulting impacts on biodiversity, carbon sequestration capacity, and essential ecosystem services such as air and water quality [67].

Figure 6 examines temporal changes in cropland extent across the study area from 2000 to 2023. The results reveal cyclical fluctuations, with notable expansions around 2003, 2008, and 2020–2021. These periods likely reflect phases of agricultural intensification linked to broader economic and environmental drivers influencing regional land-use dynamics [68].

Figure 7 illustrates the spatial distribution of the Normalized Difference Vegetation Index (NDVI) across the study area. High NDVI values (>0.6) are primarily associated with forested landscapes, including both deciduous and evergreen forests, while moderate values (0.3–0.6) correspond to pasture, hay, and cropland areas. Low NDVI values (<0.3) are predominantly observed in urbanized and barren regions. Beyond these spatial patterns, Figure 7 reveals a clear temporal decline in vegetation greenness across the three analyzed periods. Notably, areas with NDVI values below 0.4 have expanded, particularly in the western and southern portions of the study area, indicating a substantial reduction in vegetation conditions within agricultural and transitional landscapes over the past two decades.

This trend is driven by the combined effects of recurrent droughts, land-use transformation, and expanding agricultural activity in the Black Belt region. Declines in high-NDVI areas are closely aligned with reductions in NDMI, indicating increasing moisture stress and ecosystem vulnerability. As a core indicator of photosynthetic activity and vegetation condition, the sustained decline in NDVI highlights the need for adaptive land-management strategies to mitigate drought impacts.

Within Alabama’s Black Belt, NDMI is particularly effective for assessing vegetation responses to drought due to the region’s clay-rich soils and sensitivity to moisture availability. Analysis of multi-temporal NDMI derived from Landsat and Sentinel-2 imagery reveals pronounced spatial and temporal variability in vegetation moisture across the region from 2000 to 2023 (Figure 8).

NDMI patterns generally correspond with LULC distributions; however, they provide complementary information on vegetation moisture conditions that are not fully captured by NDVI alone. Areas exhibiting similar NDVI values may show substantial variation in NDMI due to differences in soil moisture availability, vegetation water content, and drought stress. To address concerns regarding seasonality, all NDMI maps were generated for a consistent growing-season period (June–September) each year, ensuring that observed differences reflect interannual moisture dynamics rather than seasonal effects. Figure 8 illustrates a clear shift from predominantly moderate moisture conditions (NDMI 0.2–0.4) in 2000 to widespread low moisture levels (NDMI < 0.2) by 2020, particularly across the central and western portions of the study area.

This spatial trend underscores the cumulative effects of recurrent drought and evolving precipitation regimes on vegetation health. The progressive expansion of low-moisture conditions highlights increasing vulnerability across the Black Belt’s ecosystems and agricultural landscapes. These patterns are consistent with long-term NDMI trends and drought indicators, including SPEI, reinforcing the need for adaptive and region-specific water management strategies.

The precipitation values illustrated are normalized pixel-level ranges intended for spatial visualization rather than raw annual totals. The CHIRPS-derived annual precipitation totals for the study region align with climatological expectations for Alabama, as shown in Figure 9, indicating considerable variability throughout the study period, which significantly impacts vegetation dynamics in the Black Belt region. In 2000, precipitation predominantly fell within the moderate-to-high range (100–160 mm), fostering a stable and robust vegetation cover. However, by 2010, a marked decline in rainfall was recorded, especially affecting central and southern sections of the region, coinciding with severe drought conditions noted in historical climate datasets. This substantial reduction likely contributed to the observed decrease in NDVI values, correlated with a rise in drought severity indices (Standardized Precipitation Index, SPI), pointing to widespread stress on plant life.

By 2020, precipitation levels exhibited signs of partial recovery, with an increased proportion of areas returning to moderate and high categories, suggesting improved soil moisture availability and potential vegetation recovery. These observations underscore the inherent relationship between precipitation variability and vegetation responses, emphasizing the urgent need for adaptive land management strategies that address the ongoing challenges associated with climate variability.

Figure 10 provides a detailed visualization of annual precipitation variability in the study area over 24 years. The associated SPI values range from −2 to +2, with the blue bars representing wet periods (SPI ≥ 0) and red bars denoting dry periods (SPI < 0). The SPI is a recognized metric for monitoring meteorological drought conditions, and the temporal variations in SPI yield critical insights into the hydrological processes at play in the region.

The SPI values depicted in Figure 10 are calculated using a 3-month accumulation period, standardized to June–July–August to ensure consistency with the vegetation indices. The probability distribution was fitted to the reference period of 1981–2010 for each calendar month individually, ensuring that SPI values adequately reflect deviations from the historical average for the specified 3-month window. This methodology adheres to standard SPI practices and facilitates meaningful year-to-year comparisons of drought conditions.

The analysis of land cover dynamics throughout the study period revealed a clear pattern of alternating wet and dry cycles, highlighting the significant climatic variability influencing these transitions. Dry spells, characterized by negative values on the Standardized Precipitation Index (SPI), were associated with reduced soil moisture levels and increased vegetation stress, leading to lower Normalized Difference Vegetation Index (NDVI) values key indicators of vegetation health that ultimately hinder agricultural productivity. The intensification of these dry periods can exacerbate land degradation, particularly in areas converting from forest ecosystems to agricultural fields. In contrast, wet periods promote enhanced vegetative growth and potential increases in crop yields; however, excessive moisture can elevate the risks of flooding and soil erosion, thereby necessitating the implementation of careful land management practices to alleviate these threats.

Moran’s I analysis indicated significant positive spatial autocorrelation in both NDVI and drought indices (p < 0.01). Spatial regression models, such as SAR and SEM, confirmed a positive correlation between NDVI and the Standardized Precipitation-Evapotranspiration Index (SPEI) after accounting for spatial dependence, with effect sizes surpassing those associated with SPI. Within-class analyses revealed that the NDVI–SPEI relationship exhibited its highest strength in forest and pasture/hay categories, while cropland data showed greater variability, aligning with practices such as crop rotation and management strategies. Geographically Weighted Regression (GWR) surfaces underscored spatial non-stationarity, demonstrating heightened sensitivity of SPEI in the western Black Belt and along riparian corridors.

The scatter plot titled “Correlation between SPI and NDVI” visually represents the relationship between SPI—indicating precipitation anomalies—and NDVI across the study area over a designated timeframe (refer to Figure 11). In this plot, the x-axis represents SPI values, with negative values indicating arid conditions and positive values reflecting wetter conditions. The y-axis displays NDVI values, a widely utilized metric for assessing vegetation health and density.

Each data point represents a paired SPI–NDVI observation, while the red line denotes the fitted linear regression. The regression Equation (6) indicates a weak positive relationship between precipitation conditions and vegetation greenness. Although NDVI increases marginally with higher SPI values, the shallow slope of the regression line reflects the limited strength of this association [69,70].

$$\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}=0.021\ast \mathrm{S}\mathrm{P}\mathrm{I}+0.695$$

(6)

The observed correlation coefficient (0.3695) indicates a weak positive relationship between SPI, representing precipitation variability, and NDVI, a proxy for vegetation condition. This suggests that, over the study period, precipitation anomalies alone had a limited direct influence on vegetation health. Instead, other factors—such as land-use change, soil characteristics, and anthropogenic influences—likely played a more substantial role in shaping vegetation dynamics.

Figure 12 illustrates the relationship between NDVI and SPEI in Alabama’s Black Belt. The scatter plot presents paired NDVI–SPEI observations, along with a fitted regression line that summarizes the overall trend (Equation (7)).

$$\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}=0.358\ast \mathrm{S}\mathrm{P}\mathrm{E}\mathrm{I}+0.6955$$

(7)

The correlation coefficient of 0.587 signifies a moderate to strong positive relationship between the Standardized Precipitation-Evapotranspiration Index (SPEI) and the Normalized Difference Vegetation Index (NDVI). The positive slope of the regression line indicates that increased SPEI values, indicative of wetter conditions, correspond with heightened NDVI values, suggesting enhanced vegetation health driven by optimal moisture conditions. Most NDVI measurements are concentrated between 0.5 and 0.8; however, during extreme drought conditions (e.g., SPEI < −2), NDVI values markedly decline below 0.4, revealing significant vegetation stress. Despite the moderate correlation, the SPEI accounts for approximately 34.5% of the variability in NDVI, implying that additional factors such as soil characteristics, land-use changes, and various management practices likely account for the residual variability.

This comprehensive analysis underscores the sensitivity of vegetative greenness to climatic water balance, stressing the necessity of integrating drought indices with remote sensing metrics to develop a holistic understanding of drought dynamics and their ecological impacts. The observed trends clearly illustrate that flora in the Black Belt region exhibits heightened sensitivity to deficits in moisture availability, particularly during extreme drought, which can have profound effects on local ecosystems and biodiversity.

Significant landscape transformations were identified, including a 17.78% decrease in deciduous forest cover and a 13.17% reduction in pasture and hay lands. In contrast, urban areas expanded substantially, with medium and high-intensity developed land increasing by 20.25% and 26.6%, respectively. Additionally, a 10.62% rise in evergreen forests was noted, likely reflecting shifts in forest composition due to ecological succession or management interventions.

Spatiotemporal analysis of NDVI and Normalized Difference Moisture Index (NDMI) data indicated a consistent decline in both vegetation vigor and moisture availability over 24 years, coinciding with recurrent drought events identified through Standardized Precipitation Index (SPI) and SPEI analyses. The relationship between drought indices and vegetation response yielded valuable insights; while the correlation between SPI, which solely considers precipitation, and NDVI was weak (R = 0.37), a much stronger correlation with SPEI, incorporating temperature variations and evapotranspiration, was recorded at (R = 0.59). This finding emphasizes the critical need to factor in the climatic water balance when assessing drought impacts on vegetation, rather than relying exclusively on precipitation data.

Land cover change identification and classification were achieved using the Maximum Likelihood (ML) classification method, which processed multi-temporal satellite imagery to generate detailed LULC maps. These maps offered quantitative metrics delineating the spatial extent of various land cover types in hectares for both 2000 and 2023, highlighting absolute changes and percentage variations over the two-decade span and providing deep insights into the intricate dynamics of landscape transformation [71].

The findings indicate a significant increase in developed land areas, particularly those categorized as medium to high intensity, which reflect urban expansion and infrastructure development driven by population growth and regional economic activities. Conversely, there have been marked declines in natural land covers, most notably deciduous forests and pasture/hay lands, with reductions of 17.78% and 13.17%, respectively. This shift suggests a transition from natural and semi-natural land uses to more intensive development practices.

On the other hand, certain land cover types showed substantial growth. Evergreen forest areas increased by 10.62%, likely due to targeted reforestation initiatives or revised forest management practices. Grasslands and herbaceous areas saw an extraordinary expansion of 70.04%, likely influenced by factors such as land abandonment, conservation efforts, or natural ecological regeneration. Emergent herbaceous wetlands experienced a substantial increase of 47.89%, indicating possible alterations in hydrological dynamics or land management strategies focused on wetland restoration.

To thoroughly assess the state of vegetation in this changing landscape, it is vital to analyze the interactions between precipitation patterns and vegetation health, utilizing SPI and NDVI datasets. The SPI acts as a metric for climatic variability, while NDVI represents vegetation response. However, correlation analysis between SPI and NDVI revealed only a weak positive relationship, suggesting that alterations in land use practices may exert a more profound influence on vegetation dynamics in the study area than precipitation alone. This highlights the intricate interplay of ecological responses to both climatic and anthropogenic factors in Alabama’s Black Belt region.

Integrating SPEI and NDVI data enhances our comprehension of environmental dynamics in the region. SPEI effectively captures variability in the climatic water balance by incorporating precipitation and temperature, while NDVI remains a reliable indicator of vegetation health and vigor. Notably, despite the strong positive correlation noted between SPEI and NDVI, the data indicates that shifts in land use practices significantly impact vegetation dynamics more than precipitation variability.

The analytical framework developed in this study consists of three core elements: supervised LULC classification, integrated analysis of NDVI and NDMI trends, and dual drought assessment using SPI and SPEI. Designed for methodological transferability, the framework can be readily applied to agricultural regions experiencing similar land-use and climate pressures. Its modular structure allows each component to be implemented independently, enabling flexibility in response to local data availability and research objectives.

A key strength of the framework is its reliance on open-access, globally available datasets, including Landsat Collection 2, CHIRPS precipitation data, and PRISM climate products. This approach enhances reproducibility and broad applicability while reducing financial and technical barriers. The framework is well-suited for regions with socio-ecological characteristics comparable to the Alabama Black Belt, such as the Mississippi Delta, the Brazilian Cerrado, and the South African Highveld, where historical land-use legacies interact with increasing drought frequency, making them priority settings for integrated environmental monitoring.

For effective adaptation of this framework to local contexts, several steps are essential. First, local calibration of the classification schemes must be conducted to ensure that the LULC categories reflect the specific agricultural practices and ecological conditions of the region. Second, the validation of drought indices, such as SPI and SPEI, should be performed against existing regional climate records to ensure accuracy and reliability. Finally, participatory engagement with land managers and stakeholders is crucial to tailoring early warning systems and establishing conservation priorities that are contextually relevant and practically applicable.

By offering a modular and data-driven template, this study paves the way for operationalizing comprehensive drought assessment of vegetation health and resilience. Ultimately, this framework supports sustainable agricultural resource management across a variety of geographic and socio-economic landscapes, translating scientific insights into actionable strategies for local communities.

Extensive research in ecological studies has documented the existence of a temporal lag phenomenon, where vegetation exhibits a delayed response to changes in precipitation. This lag implies that the cumulative effects of water stress on vegetation health amplify over time, resulting in observable impacts only after prolonged environmental changes [72].

A comprehensive investigation into aridity in Bekasi, Indonesia, further supports this phenomenon, providing a detailed analysis of the correlation between SPI, which assesses precipitation anomalies, and NDVI, which reflects vegetation health and density. The study highlights a robust positive association in regions characterized by high NDVI values, indicating dense and healthy vegetation correlating with similarly high SPI values, denoting adequate precipitation levels [73]. In contrast, areas with low NDVI values often coincide with elevated SPI values, suggesting that vegetation health is influenced by factors beyond mere precipitation levels.

Temporal analyses of drought indices such as SPEI and SPI reveal strong correlations with NDMI and NDVI anomalies. Notably, years of severe drought, specifically 2007, 2011, and 2016, correspond with significant declines in both vegetation moisture content and greenness, underscoring the vulnerability of rainfed agricultural systems in Alabama’s Black Belt region. The correlation between NDMI and SPI yields a strong coefficient (r ≈ 0.82, p < 0.001), reinforcing the critical influence of precipitation variability on vegetation stress [74].

4. Summary and Conclusions

This study presents a comprehensive, multi-decade analysis of the interplay between land-use change and drought stress in Alabama’s Black Belt, characterized by its rich clay soils, intricate land-use history, and persistent socioeconomic vulnerabilities. By integrating remote sensing-derived LULC assessments with vegetation indices (NDVI and NDMI) and dual drought indicators (Standardized Precipitation Index − SPI and Standardized Precipitation-Evapotranspiration Index − SPEI) over 24 years (2000–2023), we provide a holistic view of how these interacting factors influence vegetation resilience and ecosystem services.

The LULC analysis uncovers significant reductions in deciduous forest and pasture/hay areas, accompanied by an increase in developed land and evergreen forest, highlighting trends of urban expansion and shifts in forest composition. Concurrently, region-wide decreases in NDVI and NDMI indicate persistent vegetation stress, particularly during prolonged drought episodes. Notably, the correlation of vegetation response is stronger with SPEI (R = 0.59) compared to SPI (R = 0.37), emphasizing the critical role of atmospheric water demand, influenced by temperature-driven evaporative pressure, in mediating drought impacts on vegetation. These results suggest that relying solely on precipitation metrics may underestimate the agricultural and ecological drought risks in a warming climate.

Three key contributions are derived from this research. First, the comparative analysis of precipitation-only (SPI) versus temperature-sensitive (SPEI) drought indicators reveals a more pronounced response of vegetation health to the climatic water balance denoted by SPEI. This underscores the necessity of incorporating atmospheric water demand into assessments of agricultural and ecological drought risk under escalating temperatures. Second, the concurrent examination of NDVI and NDMI differentiates between diminished photosynthetic activity and genuine water stress, with long-term declines in both indices during multi-year droughts (notably in 2007, 2011, and 2016) confirming compounded stress on vegetation that cannot be adequately captured through a singular index. Third, the spatially explicit LULC classification delineates substantial ecological transformations: a 17.78% decline in deciduous forests, a 13.17% decrease in pasture/hay, alongside a 20.25% increase in developed land and a 10.62% rise in evergreen forests. These shifts are indicative of ongoing urbanization, changing agricultural practices, and evolving forest dynamics, all of which affect the region’s resilience to climatic extremes.

Furthermore, it is vital to address local environmental challenges through collaboratively developed policies and programs that recognize communities as rights-bearing decision-makers, affirm their right to self-determination, and empower them to navigate the balance between environmental protection and economic growth based on localized priorities and experiences [73,75,76,77].

The implications for sustainable agricultural resource management are profound. The heightened sensitivity of vegetation to SPEI highlights the crucial need to factor temperature-driven evaporative demand into drought monitoring and early warning systems. The observed declines in deciduous forest and pasture/hay, coupled with urban encroachment, emphasize the necessity of safeguarding remaining service-rich landscape elements, including forested riparian corridors, wetlands, and contiguous forest blocks that facilitate hydrologic buffering and carbon sequestration.

In rainfed agricultural zones, which exhibit the strongest correlations between vegetation health and drought, the adoption of climate-resilient practices, such as cover cropping, conservation tillage, and diversified crop rotations, can mitigate moisture stress. Given the intersection of environmental vulnerabilities with entrenched socioeconomic disparities in the Black Belt, effective strategies must be co-developed with local communities, respecting their knowledge and rights, as highlighted in participatory governance frameworks.

Ultimately, sustaining agricultural productivity and ecosystem services in the Black Belt necessitates a dual focus on managing land change and climatic water balance. Planning instruments and policies that incorporate SPEI-informed early warnings, protect critical landscape elements (such as forests and wetlands), and incentivize climate-resilient agricultural practices can significantly diminish drought risk and bolster long-term socio-ecological resilience. Through collaborative efforts, researchers, practitioners, and communities can effectively tackle the ongoing and forthcoming challenges posed by the interplay of anthropogenic landscape modifications and climate-induced drought conditions.