Analysing the mechanisms of coordinated development in the YRD using panel data models and explainable machine learning: a perspective of spatial restructuring and industrial evolution

This study draws on a panel of 41 prefecture-level cities in the Yangtze River Delta (YRD) from 2010 to 2022. By integrating statistical yearbook data with nighttime light (NTL) remote-sensing information, it investigates – through the lens of functional spatial structure – the effect of functional polycentricity on regional coordinated development and its conditional mechanisms.The baseline empirical strategy relies on a two-way fixed-effects panel model, with System GMM adopted to mitigate potential endogeneity. To identify mechanisms, interpretable machine learning is introduced to decompose heterogeneous contributions at the provincial and prefecture-city levels; furthermore, K-means clustering is performed on each city’s sample-period mean SHAP vector to identify representative city types. The results indicate that: (1) For coordinated development in the Yangtze River Delta, Shanghai stays ahead, Jiangsu and Zhejiang advance steadily and Anhui speeds up catch-up. Provincial capitals exhibit growing polarizing effects, resulting in a spatial pattern of strength in the east, riverside and Coastal Zones versus weakness in the west and inland mountainous areas. (2) The main effects of both functional polycentricity and industrial upgrading are statistically nonsignificant; however, their interaction term is robustly and significantly negative. (3) The spatial impacts of functional polycentricity are highly differentiated: positive effects are exclusively observed in core cities along the Yangtze River urban belt and the northern Hangzhou Bay corridor, whereas peripheral cities exhibit negative effects. The interaction effect between functional polycentricity and industrial structure upgrading is positive in select cities of Zhejiang and Anhui provinces, but negative in cities within the hinterland of Jiangsu Province. (4) Based on SHAP-based clustering, the urban driving patterns in the Yangtze River Delta region are categorized into four types: dual-high synergy type (Shanghai and Nanjing), main-effect-dominant type (Suzhou, Wuxi and Hefei), upgrade-remedial type (eight cities) and dual-low constrained type (28 cities).