Spatial patterns of the bird-related sentiment index and its nonlinear environmental associations: Evidence from social media and explainable machine learning

As cities become denser and more artificial, encounters with birds are closely tied to how residents perceive urban nature. However, evidence on the spatial patterns and environmental correlates of bird-related emotions remains limited. Taking Shanghai as a case study, this study cleaned 4,142 bird-related Weibo check-in comments from 2022 to 2024. A deep-learning sentiment model was used to construct a bird-related sentiment index, which was linked to 500 m × 500 m grids and examined with climatic, ecological, landscape, and human-activity indicators. Spatial analysis and interpretable machine learning were applied to identify emotional clusters, key environmental predictors, and spatial differences in model contributions. Results show that: (1) bird-related sentiment is generally positive but spatially uneven, with more continuous positive areas in suburbs and waterfronts and more fragmented responses in the urban core; (2) higher heat exposure and air pollution are associated with lower predicted sentiment scores, whereas vegetation, green coverage, and proximity to green and blue spaces correspond to higher predicted scores; (3) central districts show stronger negative model contributions from heat and pollution, while suburbs and waterfront corridors are associated with more positive predicted sentiment patterns and may provide settings where bird habitats and positive human experiences overlap.