Modeling the Effects of Urban Park 3D Environmental Features on Bird Diversity with Interpretable Machine Learning: A Case Study of Harbin, Northeastern China

This study used bird observation records from the China Bird Report (CBR), covering major urban parks in Harbin between 2020 and 2024. We retained only records located within the parks or within a 1 km buffer surrounding them. Records of species with fewer than five individuals were excluded. In total, 765 valid records were compiled, representing 141 species, 18 orders, 47 families, and 25,874 individual birds. The three-dimensional (3D) environmental features of the parks were quantified using an H3 hexagonal grid system with a resolution level of 8, corresponding to a cell area of approximately 0.7 km². After feature selection with LASSO and correlation analysis to remove highly correlated variables, 17 3D urban environmental variables were retained for analysis.