Modeling ESV losses caused by urban expansion using cellular automata and geographically weighted regression.

Driven by increasing urban demand, spatially-varying urban expansion has led to significant ecosystem degradation in China and elsewhere. Spatial nonstationarity affects the relationship between urban expansion and ecosystem service value (ESV) loss, but its significance has been under-emphasized. To study the spatially-heterogeneous ESV loss, we integrated cellular automata (CA) with geographically weighted regression (GWR) in a model that considers the relationships between urban expansion and its driving factors. We used ten GWR bandwidths to construct the CA models for reproducing rapid urban expansion at Chongqing from 2005 to 2010. We then used the CA model with the best bandwidth to predict future urban scenarios out to 2030. Our modeling shows that CA is strongly sensitive to bandwidth, and that the overall accuracy and Figure-of-Merit are maximized with a ~2 km bandwidth (about 150 samples). We examined ESV losses in eleven ecosystem classes and found that climate regulation and water flow regulation are the dominant drivers of ESV loss. From 2010 to 2030, Chongqing's urban area will increase by about 87%, resulting in substantial encroachment on agricultural land, dryland and shrubs, causing significant ESV losses of about 38%. Our results constitute an early warning of ecosystem degradation caused by massive urban development. This study improves our understanding of spatially-varying urban expansion and related ESV losses in rapidly developing areas and should help improve urban planning regulation and regional policy for sustainable development to maintain environmentally-friendly cities.