Urban street greening is an important part of urban green infrastructure, and Green View Index (GVI) is widely used to assess urban street quality and ecosystem service value as an important indicator to quantify the perception of green street landscape from a pedestrian perspective. However, the distribution of street greenery is imbalanced. Therefore, to explore the differences in street greening levels within urban cities, we crawled streetscape data using the Internet to assess the spatial distribution patterns of urban street GVI using deep learning and spatial autocorrelation, and combined 11 surrounding environmental features with multi-source geographic data to further analyze the key factors influencing the spatial variation of block GVI using ordinary least squares, geographically weighted regression (GWR) models, and multi-scale geographically weighted regression (MGWR) models. The results show that the mean value of GVI in Fuzhou city is low (23.08%), with large differences among neighborhoods and a significant spatial autocorrelation. Among the regression models, MGWR has the best fit with an R2 of 0.702, where the variables of NDVI, house price, accessibility of water bodies and parks, and the proportion of built-up land have a greater impact on GVI, and the factors do not have the same spatial effect size. The results can provide a scientific basis for promoting green visual equity in different blocks.
Read full abstract