Spatially varying determinants of farmland conversion across Qiantang watershed, China.

This paper employed geographically weighted regression (GWR) to characterize the determinants of farmland conversion at administrative scale between 1994 and 2003 across Qiantang watershed, China. Six determinants were identified: total area of forest, distance to highway, distance to second road, distance to river, population, and gross domestic product. Relationships between these identified determinants and farmland conversion showed great spatial non-stationarity, since their character, nature, and strength varied significantly across space. Typically, for cities whose development was heavily relied on road infrastructure development, the impacts of "distance to second road" and "distance to river" was negative. However, in mountainous areas, the restriction of terrain factors led to positive impacts from these two variables. For areas undergoing rapid socio-economic development, farmland conversion was accelerated by population growth and economic development. However, for more urbanized regions, a slow-down rate of farmland conversion would be expected. Our study highlighted that the problem of spatial non-stationarity should be addressed when qualifying the determinants of farmland conversion. Linking our results within the context of farmland protection, we argue that implementing local-specific land management practices, instead of the current one-size-fits-all framework, is the key for the success of farmland protection in China.