With the acceleration of urbanization, problems such as urban ecological environment quality have become increasingly prominent. How to scientifically analyze and evaluate the spatial pattern of urban ecological environment changes and influential variables is a prerequisite for achieving green development and ecological priority new in urban planning. Our study was conducted on Pingtan Island, located in Fujian Province, China. First, we selected Landsat 8 OLI images in 2013, 2017, and 2021. Second, we extracted the remote sensing ecological index (RSEI) from these images and created RSEI maps to assess the spatial-temporal variations and spatial autocorrelation of the ecological environment condition in Pingtan Island. Third, the proportion of land-use types, road, and population density were selected as independent variable factors, RSEI as the dependent variable, least squares regression (OLS), geographically weighted regression (GWR), and multi-scale geographically weighted regression (MGWR) were used to establish global and local regression models. According to the regression coefficients of the model and its spatial distribution, the spatial heterogeneity between the ecological environment and the influencing factors was assessed. The results indicated that: (1) the mean value of the RSEI increased from 0.422 to 0.504 during 2013–2021, indicating that the overall ecological environment improved. (2) Based on the global Moran’s I value, the distribution of ecological environment quality was positively correlated. The local Moran’s I cluster map showed that the high-high cluster gradually extended to the northwest high-altitude region. Low-low clustering gradually extended to the more populous areas in the southeast. (3) The $R_{adj}^2$ of the MGWR model was 0.866, which was better than the results of the OLS model and GWR model, indicating that MGWR had obvious advantages in revealing the spatial heterogeneity between the ecological environment and the influencing factors. Importantly, the results indicate that population density, road density, and the proportion of cropland land and impervious surface in land-use types have varying degrees of negative effects on the urban ecological environment, with the impervious surface being more severe, followed by population density, while forest land in land-use types shows significant positive effects.

Semi-supervised graph convolutional networks (SSGCNs) have been proven to be effective in hyperspectral image classification (HSIC). However, limited training data and spectral uncertainty restrict the classification performance, and the computational demands of a graph convolution network (GCN) present challenges for real-time applications. To overcome these issues, a dual-branch fusion of a GCN and convolutional neural network (DFGCN) is proposed for HSIC tasks. The GCN branch uses an adaptive multi-scale superpixel segmentation method to build fusion adjacency matrices at various scales, which improves the graph convolution efficiency and node representations. Additionally, a spectral feature enhancement module (SFEM) enhances the transmission of crucial channel information between the two graph convolutions. Meanwhile, the CNN branch uses a convolutional network with an attention mechanism to focus on detailed features of local areas. By combining the multi-scale superpixel features from the GCN branch and the local pixel features from the CNN branch, this method leverages complementary features to fully learn rich spatial–spectral information. Our experimental results demonstrate that the proposed method outperforms existing advanced approaches in terms of classification efficiency and accuracy across three benchmark data sets.