Impacts of Land Use Intensity on Ecological Quality Dynamics in the Central Yunnan Plateau Lake Basins, China

Abstract

Land use intensification profoundly impacts ecological quality, with this dynamic relationship being particularly pronounced in China’s Central Yunnan Plateau Lake Basin (CYP-LBs), an ecologically fragile area of significant socioeconomic value. Despite the critical importance of their interaction, existing research has largely overlooked their dynamic interplay—especially within plateau lake basins. To address this gap, this study employs the Remote Sensing Ecological Index (RSEI) to assess the ecological quality dynamics of CYP-LBs from 2005 to 2025 and its association with land use intensity (LUI), revealing spatiotemporal patterns of ecological quality evolution and its linkage to land use. Results indicate that CYP-LBs maintained overall moderate ecological quality (average RSEI ~0.50), exhibiting an initial increase followed by decline, peaking at 0.5519 in 2015. The center of gravity for ecological quality shifted eastward in most watersheds, with Moran’s I index consistently above 0.50 indicating significant spatial autocorrelation. The LUI showed an overall upward trend, with high-intensity areas primarily concentrated in lakeshore zones (e.g., eastern Dianchi Lake, Xingyun Lake) and exhibiting regional expansion over time. RSEI and LUI generally showed a negative correlation, but positive correlations emerged in localized areas of eastern and northern Dianchi Lake due to concurrent urbanization and ecological restoration efforts. Among land types, grasslands and forests were identified as the primary drivers influencing ecological quality changes in CYP-LBs. These findings provide crucial scientific basis for integrated conservation, land use optimization, and sustainable development in ecologically fragile plateau lake basins.

1. Introduction

Land use intensification plays a critical role in shaping ecological quality, directly impacting ecosystem sustainability, which is essential for both human well-being and regional development. In the lake basins of the Central Yunnan Plateau (CYP-LBs), China, where land use patterns are changing rapidly, these effects are especially significant due to the region’s unique ecological and socio-economic importance. A healthy ecological environment is essential for sustainable human development, playing a critical role in ensuring ecological security and stability for human society [1,2,3]. Yunnan Province, located in the southwestern part of China, is home to a wide distribution of lakes. It contains nine lakes larger than 30 km², primarily found in central, southern, and northwestern Yunnan [4,5,6]. The lakes of the Yunnan Plateau are crucial for regulating the climate within their watersheds, providing freshwater for adjacent urban and rural areas, serving as tourist attractions, and supplying irrigation water for agriculture [7,8,9]. The CYP-LBs play an important role in climate regulation, freshwater supply, and agricultural support, while increasingly facing pressures from urbanization and land development. Understanding how land use intensity (LUI) drives changes in ecological quality within these lake basins is essential for developing effective conservation strategies and ensuring environmental resilience. Therefore, studying the ecological quality of plateau lake basins and evaluating the dynamic effects of LUI on ecological quality is of significant importance.

Internationally, research into ecological quality assessment commenced relatively early, with evaluations of ecosystem service functions and land use beginning in the 1970s [10,11]. By the 1980s, assessments focused on ecosystem health [12,13], progressing to the current comprehensive evaluations of ecological and environmental quality [14,15]. In China, the predominant methodology for ecological quality assessment is index-based evaluation [16,17,18,19,20]. Numerous such indicator evaluation methods exist, differing in the number and type of indicators employed [21,22,23]. The absence of a unified evaluation standard, coupled with complex model calculations and difficulties in data acquisition, precludes objective assessment of current ecological quality evaluations. Xu proposed the Regional Remote Sensing Ecological Index (RSEI) [3]. By avoiding subjective human factors, the model achieves strong objectivity and can be applied to regions with diverse geographical conditions [24]. Land use intensity (LUI) indicates the degree of human disturbance to ecosystems and can serve as a factor for assessing the rationality of land use [25]. Changes in LUI directly impact the sustainability of land systems. Land-use change has increasingly been recognized in recent years as a key driver of ecosystem service degradation [26,27]. Research on LUI has made significant progress, particularly in its applications to ecosystem service trade-offs, landscape dynamics monitoring, and climate adaptation strategies [28,29]; moreover, it can also quantify the impacts of human activities on land, including the pressures on ecosystems caused by urban expansion [30] and industrialization [31]. However, extant literature has predominantly focused on plains or temperate regions, with insufficient exploration of the relationship between land-use intensity and ecological quality in CYP-LBs. As an indicator measuring the degree of human disturbance, LUI is closely linked to ecological quality. Particularly in the CYP-LBs, where urbanization coexists with ecological conservation, the interplay between these factors holds significant research value. Existing studies predominantly focus on the spatiotemporal variations in the RSEI or the independent assessment of LUI, neglecting the dynamic interaction between the two. This oversight is particularly pronounced in ecologically fragile high-altitude lake basins, potentially limiting the scientific rigor of regional ecological management decisions.

The central Yunnan region, known for its concentration of highland lakes, forms the heart of the Central Yunnan Urban Agglomeration. Moreover, as the five major plateau lake basins constitute a core component of the Central Yunnan Urban Cluster, possessing unique topographical conditions and ecological service functions, it plays a pivotal role in advancing socio-economic development and ecological civilization construction [32]. Consequently, this study examines the five major lake basins of central Yunnan (Dianchi, Yangzonghai, Fuxian lakes, Xingyun lakes, and Qilu lakes). Utilizing data from 2005, 2010, 2015, 2020, and 2025, it employs RSEI for quantitative ecological quality assessment while analyzing LUI to identify its variation patterns and driving factors. The findings aim to provide scientific grounds for ecological conservation and integrated management within the Central Yunnan Plateau lake basins, offering crucial reference for the region’s sustainable development.

2. Study Area Overview and Data Sources

The study area is primarily distributed in the central-southern part of the Yunnan Central Region, as shown in Figure 1. The CYP-LBs are located in a subtropical monsoon climate zone. As Yunnan Province’s most economically dynamic region, this area plays a vital role as a hub for economic and cultural exchange with South and Southeast Asia. It also serves as a key junction for both the Belt and Road Initiative and the Yangtze River Economic Belt [33]. Driven by its unique geopolitical advantages, the region has experienced accelerated urbanization and intensive land development, leading to pronounced habitat degradation and elevated ecological vulnerabilities that have attracted substantial research and policy concern [34].

Sub-basins were delineated and their boundaries determined using the SWAT model. The four 30 metre-resolution Landsat image sets and DEM data employed in this study were primarily sourced from the U.S. Geological Survey (https://www.usgs.gov/). Land use data utilized the 30 m annual China land cover dataset analyzed by Yang Jie and Huang Xin [35]. The MODIS series of data is from Google Earth Engine (https://code.earthengine.google.com/).

3. Methodology

3.1. Extraction and Fractionation of CYP-LBs

The Soil and Water Assessment Tool (SWAT), developed by the USDA [36], is a distributed hydrological model widely applied to simulate and forecast the impacts of land-use management and climate change on watershed hydrology at the basin scale. Therefore, this study employed the SWAT hydrological distribution model within ArcGIS 10.2 tools, utilizing 30 m-resolution DEM digital elevation data to delineate distinct sub-runoff areas. Based on the characteristics of the surrounding river network, the CYP-LBs scope was further defined.

3.2. Construction of Remote Sensing-Based Ecological Index (RSEI)

RSEI is a recent ecological index for assessing environmental quality using remote sensing data [3]. The RSEI is built from four ecologically relevant components—Greenness, Wetness, Heat, and Dryness. Their calculation formulas are detailed in Table 1.

$$\mathrm{RSEI}=\mathrm{f}(\mathrm{G}\mathrm{r}\mathrm{e}\mathrm{e}\mathrm{n}\mathrm{n}\mathrm{e}\mathrm{s}\mathrm{s},\mathrm{W}\mathrm{e}\mathrm{t}\mathrm{n}\mathrm{e}\mathrm{s}\mathrm{s},\mathrm{D}\mathrm{r}\mathrm{y}\mathrm{n}\mathrm{e}\mathrm{s}\mathrm{s},\mathrm{H}\mathrm{e}\mathrm{a}\mathrm{t})$$

(1)

Table 1. Calculation equations of single indicators of the RSEI model.

Components	Index	Equation
Greenness	NDVI	$\mathrm{NDVI}=({\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}-{\mathrm{b}}_{\mathrm{R}})/({\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}+{\mathrm{b}}_{\mathrm{R}})$	(2)
Wetness	WET	$\mathrm{WET}=\mathrm{u}{\mathrm{b}}_{\mathrm{B}}+\mathrm{v}{\mathrm{b}}_{\mathrm{G}}+\mathrm{w}{\mathrm{b}}_{\mathrm{R}}+\mathrm{x}{\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}+\mathrm{y}{\mathrm{b}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}1}+\mathrm{z}{\mathrm{b}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}2}$	(3)
Dryness	NDBSI	$\mathrm{SI}=\left[\left({\mathrm{b}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}1}+{\mathrm{b}}_{\mathrm{R}}\right)-\left({\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}+{\mathrm{b}}_{\mathrm{B}}\right)\right]/\left[\left({\mathrm{b}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}1}+{\mathrm{b}}_{\mathrm{R}}\right)+\left({\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}+{\mathrm{b}}_{\mathrm{B}}\right)\right]$	(4)
		$\mathrm{IBI}={\displaystyle \frac{2{\mathrm{b}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}1}/\left({\mathrm{b}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}1}+{\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}\right)-\left({\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}/\left({\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}+{\mathrm{b}}_{\mathrm{R}}\right)+{\mathrm{b}}_{\mathrm{G}}/\left({\mathrm{b}}_{\mathrm{G}}+{\mathrm{b}}_{\mathrm{B}}\right)\right)}{2{\mathrm{b}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}1}/\left({\mathrm{b}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}1}+{\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}\right)+\left({\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}/\left({\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}+{\mathrm{b}}_{\mathrm{R}}\right)+{\mathrm{b}}_{\mathrm{G}}/\left({\mathrm{b}}_{\mathrm{G}}+{\mathrm{b}}_{\mathrm{B}}\right)\right)}}$	(5)
		$\mathrm{NDBSI}=(\mathrm{I}\mathrm{B}\mathrm{I}+\mathrm{S}\mathrm{I})/2$	(6)
Heat	LST	$\mathrm{LST}=[\epsilon \mathrm{B}({\mathrm{T}}_{\mathrm{S}})+\left(1-\epsilon \right)\mathrm{L}\mathrm{d}]\tau +\mathrm{L}\mathrm{u}$	(7)
		$\mathrm{B}({\mathrm{T}}_{\mathrm{S}})=[\mathrm{L}-{\mathrm{L}}_{\mathrm{U}\mathrm{P}}-\tau (1-\epsilon )\mathrm{L}\mathrm{d}]/\tau \epsilon$	(8)
		${\mathrm{T}}_{\mathrm{S}}={\mathrm{K}}_2/\mathrm{l}\mathrm{n}\left({\mathrm{K}}_1/\mathrm{B}\left({\mathrm{T}}_{\mathrm{S}}\right)+1\right)$	(9)

Table 1. Calculation equations of single indicators of the RSEI model.

Components	Index	Equation
Greenness	NDVI	$\mathrm{NDVI}=({\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}-{\mathrm{b}}_{\mathrm{R}})/({\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}+{\mathrm{b}}_{\mathrm{R}})$	(2)
Wetness	WET	$\mathrm{WET}=\mathrm{u}{\mathrm{b}}_{\mathrm{B}}+\mathrm{v}{\mathrm{b}}_{\mathrm{G}}+\mathrm{w}{\mathrm{b}}_{\mathrm{R}}+\mathrm{x}{\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}+\mathrm{y}{\mathrm{b}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}1}+\mathrm{z}{\mathrm{b}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}2}$	(3)
Dryness	NDBSI	$\mathrm{SI}=\left[\left({\mathrm{b}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}1}+{\mathrm{b}}_{\mathrm{R}}\right)-\left({\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}+{\mathrm{b}}_{\mathrm{B}}\right)\right]/\left[\left({\mathrm{b}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}1}+{\mathrm{b}}_{\mathrm{R}}\right)+\left({\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}+{\mathrm{b}}_{\mathrm{B}}\right)\right]$	(4)
		$\mathrm{IBI}={\displaystyle \frac{2{\mathrm{b}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}1}/\left({\mathrm{b}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}1}+{\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}\right)-\left({\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}/\left({\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}+{\mathrm{b}}_{\mathrm{R}}\right)+{\mathrm{b}}_{\mathrm{G}}/\left({\mathrm{b}}_{\mathrm{G}}+{\mathrm{b}}_{\mathrm{B}}\right)\right)}{2{\mathrm{b}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}1}/\left({\mathrm{b}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}1}+{\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}\right)+\left({\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}/\left({\mathrm{b}}_{\mathrm{N}\mathrm{I}\mathrm{R}}+{\mathrm{b}}_{\mathrm{R}}\right)+{\mathrm{b}}_{\mathrm{G}}/\left({\mathrm{b}}_{\mathrm{G}}+{\mathrm{b}}_{\mathrm{B}}\right)\right)}}$	(5)
		$\mathrm{NDBSI}=(\mathrm{I}\mathrm{B}\mathrm{I}+\mathrm{S}\mathrm{I})/2$	(6)
Heat	LST	$\mathrm{LST}=[\epsilon \mathrm{B}({\mathrm{T}}_{\mathrm{S}})+\left(1-\epsilon \right)\mathrm{L}\mathrm{d}]\tau +\mathrm{L}\mathrm{u}$	(7)
		$\mathrm{B}({\mathrm{T}}_{\mathrm{S}})=[\mathrm{L}-{\mathrm{L}}_{\mathrm{U}\mathrm{P}}-\tau (1-\epsilon )\mathrm{L}\mathrm{d}]/\tau \epsilon$	(8)
		${\mathrm{T}}_{\mathrm{S}}={\mathrm{K}}_2/\mathrm{l}\mathrm{n}\left({\mathrm{K}}_1/\mathrm{B}\left({\mathrm{T}}_{\mathrm{S}}\right)+1\right)$	(9)

In all calculation formulas, b_k represents the band data of Landsat series imagery (including red, green, blue, near-infrared, etc.). The u, v, w, x, y, and z denotes parameters of Landsat series image bands. The ε represents the surface specific emissivity, B(T_S) is the blackbody thermal radiation brightness, L_d and L_up denote atmospheric downward and upward radiation, respectively, τ is the thermal infrared band transmittance, and K₁ and K₂ are pre-launch calibration constants [37].

In this study, the RSEI is constructed via principal component analysis [38]. Given that each indicator possesses distinct units and numerical ranges, it is necessary to normalize the values of the four indicators within the range [0, 1] prior to performing PCA as a single unit. To mitigate errors arising from inconsistent scales and noise, data points with confidence intervals falling within the 2% to 98% range are selected for PCA analysis.

3.3. Land Use Intensity

Land use intensity (LUI) is widely applied to examine how natural and anthropogenic factors shape land use patterns, capturing the degree of land exploitation in a given area [39]. To better adapt to global-scale analysis and ensure coverage of a broader geographic area, the calculation scale of the LUI was expanded to a 0.5° grid, with the specific calculation formula shown in Equation (10).

$$\mathrm{LUI}={\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}\frac{{\mathrm{A}}_{\mathrm{i}}}{{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{A}}_{\mathrm{i}}}}\times {\mathrm{W}}_{\mathrm{i}}}$$

(10)

Here, A_i denotes the area of land use category, n is the total number of categories, and W_i is the LUI coefficient for category. In this study, W_i is determined according to the degree of human disturbance and the documented hydrological effects of each land type. Following Chen et al. [40], we set (W_i = 1) for unused land; (W_i = 2) for forest, shrubland, grassland and wetland; (W_i = 3) for cropland (moderate disturbance); and (W_i = 4) for construction land.

To assess the dynamic changes within individual land use categories, the land use dynamics index was selected to calculate the degree of change for each category during the study years [41]:

$$\mathrm{K}={\displaystyle \frac{{\mathrm{A}}_{\mathrm{l}\mathrm{a}\mathrm{s}\mathrm{t}}-{\mathrm{A}}_{\mathrm{f}\mathrm{i}\mathrm{r}\mathrm{s}\mathrm{t}}}{{\mathrm{A}}_{\mathrm{l}\mathrm{a}\mathrm{s}\mathrm{t}}}}\times {\displaystyle \frac{1}{\mathrm{N}}}\times 100\%$$

(11)

where K represents the degree of dynamic change in a single land category, A_last and A_first represent the area of this land category in the final year and the initial year, respectively. N represents the length of the research period, measured in years.

3.4. Ecological Quality Unit Delineation

Based on existing research, the calculation of ecosystem service values exhibits sensitivity to the scale of the evaluation unit [42]. By comparing the ecological resilience levels of Shenyang City in 2020 across five different scales—0.5 km × 0.5 km, 1 km × 1 km, 1.5 km × 1.5 km, and 2 km × 2 km—it was found that the 1.5 km × 1.5 km grid scale more accurately and clearly represents the ecological resilience of the study area. Therefore, considering the actual conditions of the study area, the CYP-LBs was thus divided into a 1.5 km × 1.5 km grid. Grids whose center points fell outside the study area were excluded, ultimately yielding 2081 grid cells designated as ecological risk assessment units.

3.5. Center of Gravity Migration Model

The ecological quality center of gravity represents a spatial point where ecological forces are balanced in all directions. The center-of-gravity model is used to quantify the migration distance of this point [43].

$${\mathrm{SDE}}_{\mathrm{x}}=\sqrt{{\displaystyle \frac{{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}{({\mathrm{x}}_{\mathrm{i}}-\overline{\mathrm{X}})}^2}}{\mathrm{n}}}}$$

(12)

$${\mathrm{SDE}}_{\mathrm{y}}=\sqrt{{\displaystyle \frac{{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}{({\mathrm{y}}_{\mathrm{i}}-\overline{\mathrm{Y}})}^2}}{\mathrm{n}}}}$$

(13)

The SDEx and SDEy denote the centroid x-coordinate and centroid y-coordinate, respectively; $\overline{\mathrm{X}}$ and $\overline{\mathrm{Y}}$ represent the arithmetic mean centers of the geographic features; and n denotes the number of geographic features.

3.6. Local Spatial Autocorrelation

Spatial autocorrelation is categorized into global spatial autocorrelation and local spatial autocorrelation. Through z-tests, this generates LISA (Local Spatial Auto-correlation) aggregation maps, revealing the precise locations of spatial clustering or dispersion of variables within study units and their neighborhoods, thereby identifying areas significantly influencing global correlations [44].

$$\mathrm{I}={\displaystyle \frac{({\mathrm{x}}_{\mathrm{i}}-\overline{\mathrm{x}}){\displaystyle \sum _{\mathrm{j}=1}^{\mathrm{m}}{\mathrm{W}}_{\mathrm{i}\mathrm{j}}({\mathrm{x}}_{\mathrm{i}}-\overline{\mathrm{x}})}}{\frac{1}{\mathrm{n}}{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}{({\mathrm{x}}_{\mathrm{i}}-\overline{\mathrm{x}})}^2}}}$$

(14)

The I denotes the local spatial autocorrelation value, n represents the number of spatial units, W_ij is the spatial weighting matrix between study area i and study area j, and (x_i − x) and (x_j − x) denote the deviations between the observed values and the mean value in the i-th and j-th spatial units within the unit-level regional unit.

3.7. Random Forest Model

A random forest model was constructed to investigate the ecological contribution of different land-use change levels to CYP-LBs, while balancing interpretability and large-sample characteristics in the analysis [45]. Each model comprised 100 binary trees, with each tree having 1 leaf node and a random seed value of 42. The accuracy of the random forest models was evaluated using four statistical metrics: the coefficient of determination (R²), root mean square error (RMSE), bias, and mean absolute error (MAE).

4. Results

4.1. Temporal Variation in Ecological Environment Quality in CYP-LBs

Table 2. Principal component analysis of each index of RSEI in the CYP-LBs.

Years	Index	PC1	PC2	PC3	PC4
2005	NDVI	0.20	0.08	−0.08	−0.03
	WET	0.37	−0.08	0.02	0.01
	LST	−0.04	0.06	−0.04	0.05
	NDBSI	−0.11	0.10	0.11	0.00
	Eigenvalue	0.19	0.03	0.02	0.00
	Percent eigenvalue	79.01	11.53	8.28	1.18
2010	NDVI	0.24	0.08	−0.09	−0.03
	WET	0.35	0.07	0.00	−0.03
	LST	−0.11	0.08	−0.06	0.05
	NDBSI	−0.14	0.10	0.11	0.00
	Eigenvalue	0.21	0.03	0.02	0.00
	Percent eigenvalue	77.94	11.21	9.10	1.75
2015	NDVI	0.26	0.07	−0.09	−0.02
	WET	0.23	0.14	−0.04	0.02
	LST	−0.24	0.02	0.02	−0.04
	NDBSI	−0.17	0.10	0.12	0.00
	Eigenvalue	0.21	0.04	0.02	0.00
	Percent eigenvalue	76.85	13.20	9.10	0.85
2020	NDVI	0.18	−0.12	−0.06	0.03
	WET	0.06	0.13	−0.12	0.01
	LST	−0.40	−0.07	−0.01	0.01
	NDBSI	−0.01	0.08	0.10	0.04
	Eigenvalue	0.20	0.04	0.03	0.00
	Percent eigenvalue	73.51	15.40	10.22	0.88
2025	NDVI	0.26	−0.08	0.10	0.02
	WET	0.10	0.13	−0.05	0.03
	LST	−0.29	−0.09	−0.01	0.03
	NDBSI	0.20	−0.10	−0.12	0.00
	Eigenvalue	0.18	0.04	0.03	0.00
	Percent eigenvalue	72.24	16.50	10.80	0.46

displays the principal component analysis outcomes for RSEI component indicators across five periods. Results show that only the first principal component exhibits consistently stable positive and negative loadings for all ecological factors. NDVI and WET carry positive weights, whereas the other indicators are negative, demonstrating that greenness and wetness positively contribute to ecological quality. The principal component PC1 exhibited the most significant eigenvalue contribution rate across all years, consistently surpassing 72.00%. This finding indicates that the first principal component encapsulates the majority of information characteristics and can serve as the basis for quantitatively characterizing the ecological index in the CYP-LBs. Consequently, the initial principal component was selected for the calculation of the remote sensing ecological index. The results were subsequently categorized into five ecological grades [15] based on established classification standards: The scale ranges from Poor I (0.00–0.2), Fair II (0.20–0.35), Moderate III (0.35–0.55), Good IV (0.55–0.75), and Excellent V (0.75–1.00), with higher values indicating superior ecological quality. The results of this study are presented in Figure 2.

Overall, the ecological quality of lake basins on the Central Yunnan Plateau exhibited an upward trend. The average RSEI peaked in 2015 at approximately 0.55, with values hovering around 0.50 in other years. Areas with poor or very poor ecological quality were predominantly distributed within the urban agglomerations surrounding the lakes, corresponding to the actual pattern of urban development along the lakeshores. To investigate the trends in ecological quality across different years for each watershed, the areas corresponding to each RSEI grade within each watershed were statistically analyzed (Figure 3). The areas classified as RSEI grades II, III, and IV occupy the largest proportions within all river basins, while other grades cover relatively smaller areas. Within the Dianchi Lake basin, the proportion of areas rated at grade IV or below exhibits a W-shaped trend. In 2015, areas classified as good or excellent accounted for 42.79% of the total basin area, indicating a favorable ecological environment. The Fuxian Lake basin exhibits an ecological quality trend where areas classified as poor first decreased, then increased, before decreasing again. In 2015, areas with poor ecological quality covered approximately 12.44 km², accounting for 2.71% of the basin’s total area, indicating an overall favorable ecological quality. Subsequently, ecological quality deteriorated, with areas of medium or lower ecological quality accounting for 37.51% of the total study area by 2020, representing a 20.63% increase compared to 2015. Excessive human disturbance and the conversion of ecological land have led to varying degrees of ecological degradation. The Qilu Lake basin exhibits generally favorable ecological quality due to its lower elevation in the south and predominant land use types of woodland and arable land. In 2015, areas classified as moderate or above covered 284.36 km², accounting for 78.42% of the study area’s total extent. Both the Xingyun Lake and Yangzonghai basins exhibited an ecological quality trend of initial improvement followed by deterioration. In 2010, the Xingyun Lake basin had approximately 15.31 km² of poor ecological quality areas, accounting for 4.35% of the CYP-LBs. Subsequently, ecological quality gradually deteriorated, with poor quality areas increasing to 37.69 km² by 2025. The ecological quality of the Yangzonghai watershed reached its optimum in 2015, with areas classified as good or better covering 37.83% of the total study area, representing an increase of 10.40% to 16.03% compared to other years.

4.2. Spatial Variation in Ecological Environment Quality in CYP-LBs

During the study period, the center of gravity of the Dianchi Lake basin shifted overall towards the southwest (Figure 4). Between 2005 and 2025, the longitude moved from 102.7745° E to 102.7733° E, and the latitude shifted from 24.9721° N to 24.9698° N, representing a relatively minor overall displacement. The ecological quality centers of the Fuxian Lake, Xingyun Lake, and Yangzonghai Lake basins shifted southeastward during the study period. Within the Fuxian Lake basin, the center’s latitude shifted from 24.5912° N to 24.5855° N, positioning it over the upper-middle lake area. This indicates a relatively concentrated ecological quality grade within this region. The ecological quality center of the Qilu Lake basin moved eastward, with minimal shift between 2020 and 2025. This indicates limited spatial variation in ecological quality within the region and a relatively stable overall distribution.

All years satisfied the significance threshold, with global Moran’s I values consistently above 0.50, confirming strong positive spatial autocorrelation in ecological quality throughout the region (Figure 5). The “high–high” aggregation zone shows an “N”-shaped periodic trend from 2005 to 2025. Their coverage and trend mirrored those of high-quality ecological zones, primarily occurring in the northern Dianchi Lake basin, southern Qilu Lake basin, and the junction between Fuxian Lake and Dianchi Lake basins. These areas predominantly feature forested land with relatively high elevations and superior ecological quality. Areas with “low–low” aggregation exhibited an increasing trend followed by a decrease between 2005 and 2025. Most of these areas are located in the northwestern part of Dianchi Lake, the southern part of Fuxian Lake, and the western side of Xingyun Lake. The terrain in these areas is relatively gentle, and the ecological quality is generally at a medium level or below due to the impact of human activities.

4.3. Spatial Variation in Land Use Degree in CYP-LBs

Spatial autocorrelation analysis of CYP-LBs land use intensity revealed significant results for all years, with Moran’s I indices consistently exceeding 0.50. This indicates a pronounced positive spatial correlation in ecological quality across the study area. As illustrated in the results map (Figure 6), the basin exhibits predominant ‘High–High’ (HH) and ‘Low–Low’ (LL) clustering patterns. ‘HH’ clusters predominantly occur along lake margins, such as those encircling Xingyun Lake and Qilu Lake. Within the Dianchi Lake basin, HH clusters concentrate in eastern areas like Wuhua District and Xishan District of Kunming City, characterized by relatively flat terrain, robust economic development, dense populations, and high levels of anthropogenic disturbance. Over time, HH clusters in the Dianchi Lake basin gradually expanded northwards. In the Fuxian Lake basin, HH clusters predominantly occurred in Chengjiang County, north of Fuxian Lake. LL clusters across all basins were largely confined to peripheral areas, primarily constrained by topography. These regions featured relatively intact landscape patches, stable ecosystems, and minimal human disturbance. Overall, between 2005 and 2025, land use intensity in the CYP-LBs exhibited a trend of shifting from localized high-intensity clusters towards regional expansion. This was particularly pronounced in the Dianchi, Xingyun Lake, and Qilu Lake basins, where high-intensity land use was most conspicuous.

4.4. Relationship Between RSEI and LUI in CYP-LBs

4.4.1. Land Cover Under Different Ecological Quality Grades

As illustrated in Figure 7, the CYP-LBs is predominantly characterized by arable land, woodland, grassland, and impervious surfaces. Among these, woodland and arable land constitute the primary landscape matrix types within the study area. Cultivated land area initially decreased and then rebounded, hitting a minimum of 1503.59 km² in 2015. Ecological quality in these areas was mostly rated as poor or moderate. Forest land area exhibited a sustained upward trend throughout the study period, reaching approximately 1702.64 km² by 2025. In these regions, ecological quality was mainly good or excellent, with excellent grade areas covering about 716.16 km² in 2015—equivalent to 44.85% of the total forest land. Grassland area peaked in 2010 at approximately 714.80 km², predominantly classified as poor ecological quality. Built-up land area exhibited a sustained upward trend throughout the study period, reaching an impervious surface area of approximately 241.51 km² by 2025, with ecological quality predominantly rated as poor. As a key core development area in Yunnan Province, the Central Yunnan Plateau region has seen accelerating urbanization and intensified exchanges between urban clusters and surrounding cities, driving increased demand for built-up land. Areas with high ecological quality initially expanded before declining, primarily comprising forested land and cultivated land. Conversely, areas with poor ecological quality predominantly featured cultivated land and grassland, where environmental carrying capacity has diminished due to human disturbance.

4.4.2. LUI Spatiotemporal Variation and Correlation Analysis

Spatial analysis of pixel changes in the LUI sequence over time reveals a gradual increase in land use intensity for CYP-LBs. Significant intensification occurred primarily in the northern regions of Dianchi Lake, the northern part of Fuxian Lake, and the areas surrounding Xingyun Lake and Qilu Lake. In Kunming’s northern Dianchi Lake region, areas such as the southwestern part of Guandu District and Wuhua District—key components of the city—experienced markedly heightened land use intensity due to accelerated urbanization, encompassing urban expansion, residential development, and commercial projects. Conversely, during the same period, land use intensity exhibited a pronounced downward trend in Panlong District (northern Dianchi Lake basin), Chenggong District and southeast Guandu District (eastern Dianchi Lake basin), and Huaining County (eastern Fuxian Lake basin).

To reveal the spatial correlation characteristics between RSEI and LUI, this study analyzed time-series data spanning 2005 to 2025. Analysis of Figure 8 indicates that RSEI and LUI predominantly exhibit a negative correlation, meaning that stronger land use intensity correlates with poorer ecological quality. This highlights the crucial role of natural vegetation cover in enhancing the ecological quality of CYP-LBs. A positive RSEI and LUI relationship mainly occurs in Kunming’s urban zones east and north of the Dianchi basin. As the primary city center, LUI shows fluctuating increases and decreases. From 2005 to 2025, cropland along Dianchi’s northern shore steadily contracted, whereas forest and grassland expanded markedly. Such protective measures probably boosted RSEI levels.

4.4.3. RSEI Response to Changes in Different Land Types

We utilized land use dynamics to calculate the dynamic change levels of each land category from 2005 to 2025, investigating which land categories primarily influenced RSEI variations. As shown in Figure 9b, the predicted RSEI values based on land use dynamics exhibit a significant linear correlation. The regression line closely approximates a 1:1 straight line, with the model R² value reaching 0.85. The model’s RMSE is notably low at 1.73. This demonstrates that the RF model exhibits high accuracy and reliability in explaining how changes in land use categories influence RSEI.

The bar chart illustrates the feature importance scores derived from a random forest model assessing the influence of land use dynamics on the Remote Sensing Ecological Index (RSEI) over the 2005–2025 period in the study area (Figure 9a). Notably, Grassland and Forest emerged as the predominant drivers, both attaining identical importance scores exceeding 0.240, collectively accounting for the highest contribution to RSEI variability. This equivalence underscores their comparable roles in exerting either positive or negative influences on ecosystem health, likely mediated through critical biophysical processes such as evapotranspiration, soil stabilization, and biodiversity maintenance. Cropland ranked third with a score of 0.206, representing a modest decline yet retaining substantial explanatory power. This position highlights the transitional yet non-negligible impact of agricultural landscapes within the land-use matrix. The relatively low impervious surface ratio of 0.110 indicates its limited direct impact. This may stem from the localized effects of urban expansion on the urban heat island effect or runoff, which gradually dissipate across the broader study area compared to the extensive natural cover.

5. Discussion

Using remote sensing-based ecological indices, this research assesses the ecological conditions of the CYP-LBs, examines long-term trends in ecological quality across the study area, and analyzes the underlying driving factors. To ensure the reliability of the four indicators of RSEI, independent verification was conducted with MODIS products as the reference. The NDVI, LST, WET and NDBSI inverted by Landsat were compared with those inverted by MOD13A1, LST inverted by MOD11A2 and WET and NDBSI inverted by MOD09A1 in 2020. Five hundred random points were selected within the CYP-LBs range for regression analysis (Figure 10). It was found that all the indicators exceeded r > 0.70. These results confirmed the reliability of the four indicators in the RSEI construction of the study area.

Between 2005 and 2025, the ecological quality of CYP-LBs first improved and then deteriorated. These results are generally consistent with findings from previous studies [46,47]. The RSEI decreased from 0.5066 in 2005 to 0.4944 in 2025, consistent with the current scenario of increasing urban impervious surfaces and accelerated urbanization. However, this process also exhibited fluctuations. Between 2005 and 2015, CYP-LBs ecological quality improved from 0.5066 to 0.5519, indicating a marked enhancement in environmental quality. The CYP-LBs constitutes the political, economic and cultural heartland of Yunnan Province. The CYP-LBs are ecologically fragile plateau lake basins. They form a key part of the upper Jinsha River system under the Grain-for-Green Program. The region suffers from water shortages and rock desertification in the Dianchi basin. Qilu Lake faces severe eutrophication. These issues create strong pressure on ecological management [48,49]. Since 2007, Yunnan Province has vigorously implemented conservation and management initiatives for its nine major plateau lakes, undertaking environmental projects including afforestation, returning farmland to forests, ecological restoration, and nature reserve development [50]. Concurrently, the province deepened its ‘Lake Revolution’ approach, adopting tailored strategies for each lake and implementing categorized management measures. Implementation of these projects and policies has promoted ecological environment improvement. The RSEI values for the Xingyun Lake and Qilu Lake basins have declined markedly, falling from 0.4698 and 0.5486, respectively, in 2005 to 0.4285 and 0.4915 by 2025. This correlates with significant vegetation degradation and the predominance of agricultural land use within these basins. Future efforts should prioritize increasing forest coverage. The RSEI values for the Fuxian Lake and Yangzonghai Lake basins showed an upward trend, rising from 0.5000 and 0.4384 in 2005 to 0.5208 and 0.4900 in 2025, respectively. This improvement is associated with relatively intensive ecological restoration efforts. Notably, Fuxian Lake basin management has been included among ten exemplary Chinese ecological restoration cases, demonstrating significant vegetation recovery outcomes [47]. Future efforts should continue to uphold green development principles for high-quality growth, safeguarding ecosystem stability and health.

In the CYP-LBs, land use intensity and ecological quality dynamics show distinct variations, with grasslands and forests exerting the greatest influence among land cover types. Since the twentieth century, regions characterized by high land use intensity, primarily driven by the expansion of croplands and urban sprawl, have experienced the most significant decline in RSEI values. In contrast, forests and grasslands make the greatest positive contributions to ecological quality, with RSEI values remaining above 0.5 even under moderate human pressure. In the karst landforms of central Yunnan, grasslands play a crucial role in maintaining hydrological balance. There, thin soil layers and high permeability amplify the ecological consequences of vegetation removal [51]. Although the area of urban expansion constitutes only 3% of the total study area, its fragmentation effect has led to disproportionately high ecological damage, reducing the connectivity between forest patches and weakening their collective regulatory functions [52]. Preserving and rehabilitating forest and grassland systems represent the efficient approaches to sustaining ecological quality in the CYP-LBs. These two land cover types exhibit high baseline RSEI values, strong resistance to moderate disturbance, and significant contributions to regional recovery. Spatial correlation analysis between RSEI and LUI reveals an overall negative relationship between the two, consistent with the detrimental effects of intensive land use practices—such as urbanization and agricultural intensification—on ecological quality. However, the positive relationship noted along Dianchi Lake’s eastern and northern shores reflects the simultaneous advance of urbanization and ecological recovery. Thus, improving construction land efficiency, refining land use configurations to support urban expansion, and securing mutual benefits for economic progress and environmental protection are essential.

Future strategies should scientifically determine the scale of urban development, uphold an ecological-first development philosophy, manage and protect lake aquatic ecosystems, and strengthen ecological restoration and governance. Nevertheless, this study relies solely on the generic RSEI framework and does not incorporate indicators tailored to the distinctive karst topography of the central Yunnan Plateau, characterized by high permeability, rapid subsurface drainage, and vulnerability to debris flows and landslides triggered by intense rainfall. Consequently, localized ecological risks may be underestimated. Future work should develop a karst-adapted RSEI by integrating rainfall-induced hazard susceptibility, and multi-temporal high-resolution imagery to better capture the region-specific drivers of ecological degradation in plateau lake watersheds.

6. Conclusions

This study employs remote sensing ecological indices to assess and monitor the ecological quality of the CYP-LBs from 2005 to 2025, while investigating its relationship with land use intensity. The conclusions are as follows:

(1)

Over the 20-year period, the ecological quality of CYP-LBs exhibited an overall trend of initial improvement followed by decline. The average RSEI peaked in 2015 at approximately 0.5519, with the largest areas within the study region classified as ecological quality grades II, III, and IV.

(2)

During the study period, the centers of gravity for ecological quality shifted eastward to varying degrees across individual basins, while the overall distribution remained relatively stable. Ecological quality within the study area exhibited a significant positive spatial correlation. ‘High–High’ clusters were predominantly distributed in the Dianchi Lake basin in the north and the Qilu Lake basin in the south, while ‘low–low’ clusters were largely concentrated in the northwestern part of Dianchi Lake and the southern part of Fuxian Lake.

(3)

From 2005 to 2025, land use intensity in the CYP-LBs exhibited significant spatial correlation. High–high clusters clustered along lakeshores, with notably increased high-intensity use in eastern Dianchi Lake, Xingyun Lake, and Qilu Lake. Low–low clusters predominantly occurred at basin margins, exhibiting ecological stability, with the overall trend expanding from localized to regional scales.

(4)

Within the study area, RSEI and LUI predominantly exhibited negative correlations, indicating poorer ecological quality with higher land use intensity. Regional positive correlations emerged on the eastern and northern shores of Dianchi Lake, reflecting the intertwined processes of urbanization and ecological restoration.