Urban sustainable development evenness and nonlinear drivers: evidence from three urban agglomerations in China

This study develops a localized SDG assessment framework using multi-source data integration and introduces an evenness index to characterize SDG progress among cities in three major urban agglomerations in China from 2015 to 2022.

Furthermore, explainable machine learning integrating Tabular Prior-data Fitted Network (TabPFN) and Shapley Additive Explanations (SHAP) is employed to identify key drivers and their nonlinear relationships. The results indicate that: (1) although urban sustainable development has improved, SDG progress exhibits persistent structural imbalance, with the average comprehensive score of the 30 cities increasing by 7.19 points from 2015 to 2022; (2) development evenness shows notable temporal fluctuations and stage-wise reversals, with the evenness index (EI) primarily ranging from 0.40 to 0.65, alongside dynamic spatial reconfiguration and heterogeneous evolution; (3) the explanatory contributions of driving factors are highly heterogeneous across cities, with magnitude and direction varying significantly; and (4) complex nonlinear relationships are observed between drivers and evenness, primarily characterized by threshold effects and non-monotonic responses.