Monitoring Spatiotemporal Changes of Impervious Surfaces in Beijing City Using Random Forest Algorithm and Textural Features

As the capital city of China, Beijing has experienced unprecedented economic and population growth and dramatic impervious surface changes during the last few decades. An application of the classification method combining the spectral and textural features based on Random Forest was conducted to monitor the spatial and temporal changes of Beijing’s impervious surfaces. This classification strategy achieved excellent performance in the impervious surface extraction in complex urban areas, as the Kappa coefficient reached 0.850. Based on this strategy, the impervious surfaces inside Beijing’s sixth ring road in 1997, 2002, 2007, 2013, and 2017 were extracted. As the development of Beijing has a special regional feature, the changes of impervious surfaces within the sixth ring road were assessed. The findings are as follows: (1) the textural features can significantly improve the classification accuracy of land cover in urban areas, especially for the impervious surface with high albedo. (2) Impervious surfaces within the sixth ring road expanded dramatically from 1997 to 2017, had three expanding periods: 1997–2002, 2002–2007, and 2013–2017, and only shrank in 2007–2013. There are different possible major driving factors for each period. (3) The region between the fifth and sixth ring roads in Beijing underwent the most significant changes in the two decades. (4) The inner three regions are relatively highly urbanized areas compared to the outer two regions. Urbanization processes in the interior regions tend to be completed compared to the exterior regions.