Quantifying the non-linear response of bird diversity to landscape features in metropolitan areas: A machine learning-based analysis

Abstract

Understanding the relationship between urban bird diversity and landscape features is essential for effective biodiversity conservation and sustainable urban development. This study utilizes bird observation data collected from a citizen science platform in Nanjing, China, to explore the spatial scales most appropriate for assessing urban bird diversity (species richness and functional diversity) and their non-linear associations with various landscape features. Advanced machine learning models, such as Boosted Regression Trees (BRT), eXtreme Gradient Boosting (XGBoost), and SHapley Additive exPlanations (SHAP), were employed to analyze these relationships. Key findings include: (1) A sum of 275 bird species were recorded in Nanjing's main urban area, with Passeriformes (77.06 %) being the dominant order. Both indices increased in average value as the scale size expanded (300 m, 500 m, 700 m, 900 m, 1100 m, 1300 m, and 1500 m). Statistical analysis indicated a more random spatial distribution of bird richness at smaller scales (300–700 m, Moran's I = 0.24), while functional diversity remained largely stable across scales. (2) The BRT results indicated that the relationship between bird diversity indices and landscape features peaked at the 900 m scale, which was thus identified as the optimal spatial scale. (3) Excessive nighttime lighting (values >62), high building density (> 0.15), and road density (> 5) were found to suppress bird diversity, while favorable terrain conditions, such as moderate elevation (50–150 m), aspect angles (100°–200°), and gentle slopes (< 5°), along with enhanced ecological conditions (RSEI >0.5) promoted it. Interaction analyses further demonstrated that the effects of terrain variables, building density, and road density attenuated the negative impacts of urban nighttime lighting. These findings offer practical implications for urban planning and design by identifying scale-sensitive and threshold-based strategies to support bird biodiversity and promote harmonious coexistence between humans and nature in metropolitan landscapes.