Spatially non-stationary relationships between urbanization and the characteristics and storage-regulation capacities of river systems in the Tai Lake Plain, China

• Published in: The Science of the total environment Vol. 824; p. 153684
• Main Authors: Lei, Chaogui, Wang, Qiang, Wang, Yuefeng, Han, Longfei, Yuan, Jia, Yang, Liu, Xu, Youpeng
• Format: Journal Article
• Language: English
• Published: Netherlands Elsevier B.V 10-06-2022
• Subjects: Bivariate Moran's I; China; Geographically weighted regression; Lakes; Non-stationary relationships; Rapid urbanization; Rivers; Storage-regulation capacities; Urbanization; China; Storage-regulation capacities; Rapid urbanization; Geographically weighted regression; Non-stationary relationships; Bivariate Moran's I
• ISSN: 0048-9697; 1879-1026
• DOI: 10.1016/j.scitotenv.2022.153684

Given environmental or hydrological functions influenced by changing river networks in the development of rapid urbanization, a clear understanding of the relationships between comprehensive urbanization (CUB) and river network characteristics (RNC), storage capacity (RSC), and regulation capacity (RRC) is urgently needed. In the rapidly urbanized Tai Lake Plain (TLP), China, various methods and multisource data were integrated to estimate the dynamics of RNC, RSC, and RRC as well as their interactions with urbanization. The bivariate Moran's I methods were applied to detect and visualize the spatial dependency of RNC, RSC, or RRC on urbanization. Geographically weighted regression (GWR) model was set up to characterize spatial heterogeneity of urbanization influences on RNC, RSC and RRC. Our results indicated that RNC, RSC and RRC variables each showed an overall decreasing trend across space from 1960s to 2010s, particularly in those of tributary rivers. RNC, RSC, or RRC had globally negative correlations with CUB, respectively, but looking at local scale the spatial correlations between each pair were categorized as four types: high-high, high-low, low-low, and low-high. GWR was identified to accurately predict the response of most RNC, RSC, or RRC variables to CUB (R2: 0.6–0.8). The predictive ability of GWR was spatially non-stationary. The obtained relationships presented different directions and strength in space. All variables except for the water surface ratio (Wp) were more positively affected by CUB in the middle eastern parts of TLP. Drainage density, RSC and RRC variables were more negatively influenced by CUB in the northeast compared to other parts. The quantitative results of spatial relationships between urbanization and RNC, RSC or RRC can provide location-specific guidance for river environment protection and regional flood risk management.
[Display omitted]
•Multisource data were used to depict river indicators (RNCs) and urbanization (CUB).•Bivariate Moran's I visualized spatial correlations between RNCs and CUB.•RNCs have markedly degraded due to urbanization, particularly for tributary rivers.•Spatial spillover effect occurred in the local relationship between RNCs and CUB.•CUB affected RNCs in a spatially non-stationary way.