2.1. Study Area
Hubei Province, located at the core of China and surrounded by Henan, Shaanxi, Jiangxi, Hunan, Anhui, and Chongqing, acts as a crucial internal transport nexus (
Figure 1). It spans from longitude 108°21′ E to 116°07′ E and latitude 29°05′ N to 33°20′ N. The province features a diverse range of landscapes and climates, from the fertile Jianghan Plain to its mountainous western regions, which are important for agriculture and ecotourism. Hubei is abundant in water resources, which play a vital role in its extensive agricultural and fisheries sectors, especially in the Jianghan Plain’s central and eastern parts, which enjoy favorable hydrothermal conditions and long frost-free periods, making them key agricultural areas. Additionally, the western areas, covered in forests and grasslands, are important for biodiversity and environmental sustainability efforts.
The terrain in Hubei is predominantly forested and cultivated, covering more than 80% of the region’s land and showcasing its rich forestry resources. Urban development is gradually integrating with these natural areas, creating a dynamic shift in land use patterns. The province’s diverse geography includes mountains, hills, and plains, offering a variety of terrain and soil types, ranging from yellow-brown and cinnamon soils in the north to red soils in the southeast and yellow soils in the southwest, alongside the prevalent paddy and alluvial soils on the plains.
2.2. Data Source
MODIS, installed on NASA’s EOS satellites Terra and Aqua (NASA, Washington, DC, USA), launched in 1999 and 2002, plays a significant role in Earth observation. This instrument captures data across 36 spectral bands, ranging from visible to thermal infrared light [
27]. With spatial resolutions from 250 m in the visible range to 1 km for other bands, providing comprehensive global monitoring every 1 to 2 days and aiding in the study of large-scale environmental changes across the atmosphere, oceans, and land surfaces [
28]. The normalization of the Vegetation Health Index for each year in the calculations is mainly based on the characteristics of MODIS satellite data in order to more accurately reflect the areas with the healthiest vegetation [
29].
Using a dataset of 335,709 Landsat images (U.S. Geological Survey, Reston, VA, USA), a research team from Wuhan University constructed the first Landsat-based CLCD for the period from 1985 to 2019 [
30]. Training samples were gathered by combining stable samples extracted from the Chinese Land Use/Cover Dataset (CLUD) with visual interpretation samples from satellite time-series data, Google Earth, and Google Maps (Google LLC, Mountain View, CA, USA). Several temporal indices were constructed using all available Landsat data, which were then input into a Random Forest classifier to obtain classification results. Furthermore, a post-processing method integrating spatiotemporal filtering and logical reasoning was proposed to enhance the spatial and temporal consistency of the CLCD. The land cover change map from 2012 to 2022 compiled in the CLCD is shown in
Figure 2.
2.3. Land Cover Vegetation Health Index
The VCI is an index that uses remote sensing data to assess vegetation conditions. It calculates by comparing the current observed vegetation index, such as the Normalized Difference Vegetation Index (NDVI), against historical maximum and minimum values at the same location. The purpose of the VCI (1) is to measure the changes in vegetation condition over a specific period relative to the past, making it widely used for monitoring drought and assessing vegetation health and productivity in environmental and agricultural applications.
The NDVI represents the current Normalized Difference Vegetation Index value. NDVImin refers to the minimum NDVI value recorded historically for the location, while NDVImax denotes the maximum NDVI value observed in the records. These values are essential for calculating indices such as the Vegetation Condition Index (VCI), which assess vegetation health by comparing the current NDVI to its historical range.
The calculation formula for the TCI is shown in Formula (2).
In the context of temperature analysis, Land Surface Temperature (LST) refers to the current observed surface temperature at a specific location. LSTmin represents the minimum temperature recorded during the observation period at the same location, whereas LSTmax denotes the maximum temperature observed. These metrics are pivotal in studies involving thermal dynamics of the earth’s surface, providing foundational data for assessing thermal variability and extremities over time.
The VHI evaluates the overall health of vegetation by integrating the VCI and the TCI. The calculation formula for the VHI is defined as VHI = α × VCI + (1 − α) × TCI, Formula (3); the TCI’s calculation formula is shown in (2). α is the weight coefficient used to balance the contributions of the two indices in the VHI. Following the recommendations of Kogan et al. [
31], this study assigns equal weights to both the VCI and TCI (α = 0.5). The VHI is extensively applied in drought monitoring, vegetation management, and climate change studies. It assists researchers and policymakers in precisely analyzing vegetation conditions across different regions, providing robust data support for formulating related environmental policies and strategies.
The LCVHI is an indicator used to assess the comprehensive impact of different land cover types on vegetation health. This index is calculated using the following Formula (4):
In vegetation health assessment, the Land Cover Vegetation Health Index (LCVHI) employs different weighting factors to comprehensively analyze vegetation health under various land cover types. The weight
βj (
Table 1) for each land cover type indicates its relative importance in the overall assessment.
Ten experts with extensive experience in the fields of ecology, environmental science, and land management were invited to participate in the scoring. These experts have extensive research and practical experience in their respective fields.
Experts independently rated the ecological importance (EI) and environmental sensitivity (ES) of each land cover type based on their expertise and experience. The scoring system ranges from 0 to 1, with 0 indicating the lowest importance or sensitivity and 1 indicating the highest importance or sensitivity (
Appendix A) (
Table 1). In this study, we assigned equal weight to ecological importance (EI) and environmental sensitivity (ES) to ensure a comprehensive and balanced evaluation. Ecological importance reflects the long-term and broad functions of land cover types within the ecosystem, such as carbon storage, biodiversity maintenance, and climate regulation. Environmental sensitivity, on the other hand, assesses the response capacity and vulnerability of these types to environmental changes. By giving both EI and ES equal weight, we ensure that our assessment does not favor one aspect over the other, thus achieving a more balanced and comprehensive evaluation.
To assess the consistency of the expert rating data, we used the statistical software SPSS (version 25) to calculate the Cronbach’s Alpha coefficient. Specifically, we input the ratings of ecological importance (EI) and environmental sensitivity (ES) for each land cover type from 10 experts into the software, which automatically generated a Cronbach’s Alpha coefficient of 0.997. This coefficient, well above the commonly accepted reliability threshold of 0.9, indicates a very high level of consistency and reliability in the expert ratings. This result confirms the reliability of the research data and ensures the scientific rigor and validity of the subsequent analysis.
Based on the above ratings, we calculated the comprehensive score (CS) for each land cover type as follows (
Table 2), using Formulas (5) and (6):
To ensure that the sum of all weight coefficients equals 1, we standardized the comprehensive scores. The formula for the standardized weight coefficients is as follows:
According to practical application needs, we have made appropriate adjustments to the standardized weight coefficients to ensure their rationality and operability in practical applications, ensuring that their sum equals 1. The adjusted weight coefficients (
β) were thus obtained (
Table 3).
In assessing vegetation health, monthly calculations of the TCI and VCI were utilized to derive the VHI. This index was employed to generate drought condition maps by analyzing VHI values from each July during the study period. Additionally, a decade-long time-series analysis of the VHI for Hubei Province identifies overall trends and changes in vegetation health, with a particular focus on the initial and final years, using the LCVHI.
Related research indicates that low VCI and TCI values are often closely associated with deteriorating vegetation health, and these indices are used to monitor the impact of drought, extreme temperatures, and other environmental stresses on vegetation health [
32]. These indices are combined to form the VHI, which is used to more accurately assess vegetation health under various environmental stresses [
33]. By incorporating land cover data, the LCVHI can more accurately reflect ecological pressure across different land types [
34]. To determine the threshold for “significant ecological stress”, this study references the threshold settings of the VCI and TCI from the relevant literature and, combined with expert consultation, categorizes LCVHI values into five levels, defining areas with a level 5 rating as significant ecological stress zones.
The Natural Breaks Classification, also known as the Jenks optimization method, is an effective statistical data classification method particularly suited for data with natural clusters. When applied to the study of VHI changes, this method offers several benefits. It maximizes inter-class differences by optimizing variance between classes, making data points within each category relatively similar and accentuating significant differences between categories. This approach helps to clearly identify and interpret VHI changes, ensuring that the categorization of LCVHI values into different levels reflects the intrinsic structure and patterns of the data.
This threshold is based on the analysis of historical data and the ecosystem’s response to environmental pressure [
35].