High resolution mapping of global urban extents from 1870 to 2100 by integrating data and model driven approaches

Dynamics of urban extent, especially with a long temporal span and global coverage, are important to evaluate the impacts of anthropogenic activities on natural and human environments. Although satellite-based observations can provide the historical transformation of urban extent, the temporal span of the remotely sensed data is limited, resulting in an incomplete understanding of the urban land dynamics. A global urban extent dynamic dataset of hindcasting and forecasting with satellite-derived observations could be useful to investigate the long-term impact of global urbanization for decision making. In this study, we modelled global urban extent dynamics from 1870 to 2100, using a urban cellular automata (CA) model with consideration of historical trends of urban growth over a long temporal span and the satellite-derived urban extent time series data (1992-2013). We conducted historical hindcasting (1870-1990) using the urban shrink model developed in this study. We also projected future (2020-2100) global urban land extent dynamics under the five Shared Socioeconomic Pathways (SSPs). We found that global urban area was largest under SSP5 (i.e., fossil-fuelled development), having a global urban area more than 40 times that of 1870. The resulting dataset shows a relatively complete cycle of urban development at a grid cell level for more than two hundred years, which can be used to assess the long-term effect of global urbanization on urban land environmental changes. In addition, our results could be useful for exploring the human-environment interactions in Earth systems models.