Facilitating urban climate forecasts in rapidly urbanizing regions with land-use change modeling

The local climate zone (LCZ) mapping scheme classifies urban lands into multiple types according to their climate-relevant surface properties, enabling forecasts of changes in urban climate. However, stationary LCZ maps are insufficient for forecasts in the rapidly urbanizing regions, where there are frequent transitions among multiple urban lands and thus changing surface properties. To facilitate climate forecasts with these changing properties, we propose a new methodological framework to predict future LCZ maps using the Cellular Automata (CA) landcover/land-use change (LCLUC) model. Different from most existing LCLUC studies that treat the urban area as homogeneous, our work is the first attempt to simulate the complex conversions among low-, mid-and high-rise urban lands defined in LCZ. To validate our method, we apply it in the Pearl River Delta (PRD) metropolitan area, China, a rapidly urbanizing region with more than 50 million residents. First, we use the World Urban Database and Portal Tool (WUDAPT) method to generate LCZ maps of the PRD region in 2009 and 2014, with satellite images. Then, we apply the CA model on the 2009 LCZ map to forecast 2014 one based on the LCLUC rules discovered by the data mining technique. The comparison between the forecasted and observed 2014 LCZ maps yields a kappa coefficient of 0.77 and an overall accuracy of 82%. Our method achieves satisfactory accuracies on the high-(84%) and low-rise (82%) urban lands while performing relatively poorly on the mid-rise (40%) lands. Our results demonstrate that the combination of the LCZ scheme and LCLUC modeling has the potential of capturing the structural changes within cities and providing the necessary input datasets for urban climate forecasts.