Identifying driving factors of urban land expansion using Google Earth Engine and machine-learning approaches in Mentougou District, China
Linlin Cheng, Chao Tian, Yin Ting-ting
Abstract
The research on driving mechanisms of urban land expansion is hot topic of land science. However, the relative importance of anthropogenic-natural factors and how they affect urban land expansion change are still unclear. Based on the Google Earth Engine platform, this study used the support vector machine classifier to extract land-use datasets of Mentougou district of Beijing, China from 1990 to 2016. Supported by machine-learning approaches, multiple linear regression (MLR) and random forests (RF) were applied and compared to identify the influential factors and their relative importance on urban land expansion. The results show: There was a continuous growth in urban land expansion from 1990 to 2016, the increased area reached 6097.42 ha with an average annual rate of 8.01% and average annual intensity rate of 2.57%, respectively. Factors such as elevation, risk of goaf collapse, accessibility, local fiscal expenditure, industrial restructuring, per capita income in rural area, GDP were important drivers of urban land expansion change. The model comparison indicated that RF had greater ability than MLR to identify the non-linear relationships between urban land expansion and explanatory variables. The influencing factors of urban land expansion should be comprehensively considered to regulate new land policy actions in Mentougou.