Mapping of 30m global urban boundaries from 1972 to 2021 based on multi-source geographic information fusion

Urbans are the main places of human gathering and activities. With the acceleration of global urbanization, it is significative to quickly and accurately obtain the scope of global urban boundaries and analyse their dynamic changes. Urban development affects many fields, such as economy, ecology and politics. What’s more,  the human demand for urban boundary products is gradually increasing. In recent years, great progress has been made in products such as global impervious surface area (GISA), which makes it is possible to obtain high-precision, large-scale and long-term urban boundary products. However, the existing urban boundary products cannot meet the requirements of scientific research on the timeliness, accuracy and quality of data. It is of great significance to optimize urban boundary products. Therefore, based on multi-source geographic information data, this paper uses kernel density estimation (KDE) and convolution operations to obtain global urban boundary products from 1972 to 2021. This product was named global urban settlement boundary vector (GUSV). It includes the elimination of false alarms in non-urban areas, the acquisition of initial urban boundaries, the supplement of urban core areas and the refinement of urban edges. Through the verification of 100 urban boundary samples in 8 years (1985, 1990, 1995, 2000, 2005, 2010, 2015, 2020), the mean relative error (MRE) of GUSV is the lowest 0.160 and the mean intersection over union (mIoU) is the highest 0.729. The MRE of the existing global urban boundaries (GUB) is the lowest 0.428, and the mIoU is the highest 0.542. The minimum value of the correlation coefficient between GUSV and GUB is 0.881, and the maximum value is 0.932. So GUSV is in good consistency with GUB . Moreover, it supplements the omission of GUB. The GUSV's statistical results show that the total area of cities will increase year by year from 1972 to 2021.