The spatial and temporal patterns of impervious surface and their socioeconomic impacts in a Chinese rapid urbanized area—A hybrid of remote sensing and machine learning approaches.

The expansion of impervious surface in fast urbanized areas dominates the dramatic changes in physical environment, and impacts soci o economic system . The relationship between urbanization and citizenization will exist and evolve for a long time in urban and rural areas of the central and western China. The quantification of the spatial and temporal patterns of impervious surface and their socioeconomic impacts not only elucidates the urbanization effect on resident s' l ife change , but also informs the formulation of smart and people-oriented urban and rural planning strategies.

The study site was the Liangjiang New Area in Chongqing, China , with an area of 1200 km² . In this study, a three-step method was applied to examine how the variation in the patterns of impervious surface affects the socioeconomic indicators of the 28 towns in the site. Firstly, t he mixed-pixel decomposition method was used to obtain the impervious surface maps in 2000, 2006 and 2012, with Landsat Imagery as the input . Landscape metrics such as aggregation index , connectivity index and cohesion were calculated to analyze the spatial patterns of the impervious surface. Next, the 28 towns were clustered with Self-Organizing Map (SOM), according to the socioeconomic indicators (population growth rate, energy consumption, annual education rate, GDP, per capita income, and migrant workers number) together with the impervious surface pattern metrics . Lastly, Generalized Additive Models (GAMs), which provides a good fit for nonlinear relationships with noisy data, was adopted to show how the characteristics of the impervious surface pattern influences the socioeconomic indicators of towns.

The R² and RMSE of the classified impervious surface map were 0.986 and 0.024. The impervious percentage of the Liangjiang New Area increased from 0.062 to 0.278 from 2000 to 2012, with a total increase of 348.4% and an annual growth rate of 13.3%. We found that natural population growth rate and migrant workers number were significantly affected by the impervious surface expansion. Impervious surface expansion led to the increase of rural population and the migration of labors to large cities. The hybrid of remote sensing and machine learning approaches enlightens a better comprehension of the socio-environmental system in the Chinese rapid urbanized area.