Abstract: Analyzing the impacts of road network expansion across different grades on environment quality is vital for guiding infrastructure planning and ecological conservation. We leveraged road data and Landsat imagery (2016 and 2023) alongside the remote sensing ecological index (RSEI) to assess RSEI responses to road networks with different grades within buffer zones in Quanzhou City. Spatial units were analyzed at different grid scales (1000 m×1000 m, 1500 m×1500 m, 2000 m×2000 m, 2500 m×2500 m, 3000 m×3000 m). The spatial heterogeneity of the relationships between road kernel density estimation (KDE) and RSEI were analyzed using bivariate spatial autocorrelation and geographically weighted regression (GWR). Results showed the total road length increased significantly from 2016 to 2023, with average KDE rising significantly, while the overall average RSEI declined. The areas with good ecological environment quality had a smaller increase compared to those with poor ecological environment quality, indicating local improvement coexisting with overall degradation in RSEI. As the buffer zone distance increased from 0 to 3000 m, the RSEI values increased gradually for all road grades, with impact thresholds approximately 900 m for national/provincial roads and 600-750 m for county/township roads. The latter showed greater RSEI fluctuation with distance. Multiscale bivariate spatial autocorrelation confirmed a significant negative KDE-RSEI correlation (significant negative Global Moran’s I), being strongest at the 1000 m×1000 m scale (Moran’s I=-0.476 in 2016, -0.553 in 2023). Moran’s I of ΔKDE-ΔRSEI was negative and decreased with increasing scales. By comparing multi-scale GWR models, 1000 m×1000 m and 1500 m×1500 m grids yielded optimal fitness (higher adjusted R², lower AIC). Overall, KDE negatively correlated with RSEI, exhibiting spatial heterogeneity. Small areas with positive effect emerged in Yongchun and Dehua counties (northwest part of the study area), while negative effects occurred in South Wing New City and along the Fuzhou-Xiamen High-Speed Railway. Such results were basically true for ΔKDE-ΔRSEI relationships. Areas with positive correlations increased notably from 2016 to 2023. Our results provide multi-scale spatial decision support for reconciling coastal urban development with ecological protection.

Key words: road network expansion, ecological environment quality, remote sensing ecological index, road kernel density, spatial autocorrelation, geographically weighted regression