关键词: 季节性遥感生态指数（RSEI-AP）, 主成分分析, 大气污染物, 生态环境质量, 高污染负荷

Abstract:

[Objective] Accurate assessment of regional ecological quality is increasingly critical under the dual pressures of intensified human activities and complex atmospheric pollution, especially in regions with high pollution loads and marked seasonal variations. Traditional remote sensing-based ecological indices, such as the Remote Sensing Ecological Index (RSEI), integrates greenness, wetness, dryness, and heat via Principal Component Analysis (PCA) to provide rapid and objective evaluation. However, the RSEI relies on a fixed suite of indicators and fails to dynamically incorporate atmospheric pollutant stressors, leading to assessment biases in heavily polluted areas where Particulate Matter (PM_2.5) and Ozone (O₃) exert significant seasonal ecological stress. Taking Shijiazhuang, a typical developing city characterized by a composite pollution pattern with PM_2.5 dominance in winter-spring and O₃ dominance in summer-autumn, as the research area, for the first time this research constructed a Seasonally Adaptive Remote Sensing Ecological Index with Atmospheric Pollutants (RSEI-AP), with the overarching goal of developing a more responsive and accurate method for evaluating ecological quality under high pollution loads. [Methods] The research utilized Landsat 8/9 Level-2 Collection 2 Tier 1 imagery (2023) to derive 4 baseline RSEI indicators: the Normalized Difference Vegetation Index (NDVI), the Wetness component (WET) from tasseled cap transformation, the Normalized Difference Built-up and Soil Index (NDBSI), and Land Surface Temperature (LST). Pre-processing procedures strictly adhered to the CFMASK algorithm for cloud masking and atmospheric correction to ensure the fidelity of surface reflectance data prior to indicator calculation. To capture the seasonally varying atmospheric pollution stress, high-resolution gridded PM_2.5 and O₃ concentration datasets (China High PM_2.5 and China High O₃, 1 km resolution) were acquired and processed. These datasets feature high accuracy, with ten-fold cross-validation R² of 0.92 and RMSE of 10.76 µg∙m⁻³ for PM_2.5, and R² of 0.89 with RMSE of 15.77 µg∙m⁻³ for O₃. According to the typical pollution regimes, PM_2.5 was integrated as an ecological stress factor for the winter-spring season (representing the heating period and stagnant weather), while O₃ was integrated for the summer-autumn season (when photochemical reactions are intense). This temporal stratification acknowledges that the ecological damage mechanisms differ fundamentally between particulate matter deposition in colder months and oxidative stress from photochemical smog in warmer months. All indicators were standardized and water bodies were masked. PCA was then performed separately for each season on the five input layers (NDVI, WET, NDBSI, LST, and the respective pollutant). The first Principal Component 1(PC1), which captured 80.05% of the variance for winter-spring and 71.42% for summer-autumn, was extracted and normalized to produce seasonal RSEI-AP images. To synthesize an annual composite RSEI-AP that reflects cumulative ecological impact, an exhaustive traversal weight optimization algorithm was applied; this algorithm iterated through all possible weight combinations (α for winter-spring and β for summer-autumn, with α+β=1) to identify the optimal linear combination, ensuring that the annual index retained the maximum ecological information. In the annual integration, the PC1 collectively accounted for 75.7% of the total variance, highlighting its efficiency in representing the coupled ecological and pollution signals. For validation, a conventional annual RSEI was also generated from the same seasonal imagery without pollutants, using an equivalent synthesis framework. Performance was assessed through spatial pattern comparison. [Results] Spatially, the annual RSEI-AP revealed a clear and stable “superior in the west, moderate in the center, and inferior in the east” pattern. The western Taihang Mountain area appeared as a continuous high-value zone (RSEI-AP≥0.6, reaching “excellent” and “extremely excellent” levels), confirming its role as the critical ecological barrier for the region. The distinct gradient captured by the RSEI-AP aligns closely with the topographic transition from mountainous terrain to alluvial plains, reflecting the constraints of physical geography on both pollution diffusion and ecological conservation. The central piedmont plain exhibited moderate ecological quality, while the eastern plain and urban core showed extensive low-value clusters. Specifically, the “poor” category (RSEI-AP<0.2) accounted for 15.37% of the total area, an increase of 4.36 percentage points over the RSEI, primarily distributed in regions suffering combined PM_2.5 and O₃ stress. Meanwhile, the proportion of “good” and above grades increased by 3.46 percentage points, demonstrating improved discrimination of high-quality ecological zones. In contrast, the annual RSEI produced a relatively homogenized pattern with blurred spatial gradients, underestimating the ecological disadvantage of the heavily polluted eastern plain and overestimating the quality in some vegetated yet ozone-stressed areas. Spatial autocorrelation analysis using Global Moran’s I returned a value of 0.97 (p<0.01), confirming extremely strong spatial clustering of ecological quality. Hot spot analysis (Getis-Ord Gi*) further identified statistically significant hot spots concentrated in the western mountains and cold spots clustered in the central-eastern plains and along major transportation corridors, reinforcing the “hot in the west, cold in the east” dichotomy and highlighting the spatial imprint of intensive human activities and pollution exposure. Seasonal RSEI-AP results provided nuanced insights: the winter-spring RSEI-AP, with PM_2.5 incorporated, expanded the “poor” category to 17.26%—more than double that of the corresponding RSEI—sharply delineating the degraded zone in the eastern plain. The summer-autumn RSEI-AP, by incorporating O₃, revealed that the “poor” area surged from 0.51% in the RSEI to 8.85%, a more than 16-fold increase, and the “moderate” area expanded from 24.87% to 56.28%, effectively unmasking the ozone stress that was hidden by flourishing vegetation. These seasonal dynamics underscore the necessity of a season-adaptive approach for accurate ecological assessment. [Conclusion] Overall, the proposed RSEI-AP successfully integrated seasonally dominant atmospheric pollutants as dynamic stress factors, overcoming the limitations of the static RSEI. It provided a more accurate depiction of the true ecological background, restored the ecological prominence of the Taihang Mountain barrier, and precisely identified pollution-induced degradation hotspots. This method offers a universal and transferable technical framework for ecological assessment in high-pollution-load regions worldwide with seasonal pollution signatures and holds significant theoretical and practical value for advancing remote sensing-based ecological monitoring from static evaluation toward dynamic, precise, and multi-scenario early warning systems. Moreover, it lays a foundation for future multi-scenario simulations, long-term ecological dynamic monitoring, and early warning of ecological degradation under combined pollution stress.

Key words: RSEI-AP, principal component analysis (PCA), atmospheric pollutants, ecological environment quality, high pollution load