Timely and accurate information of large-scale urban land distributions is fundamental to the understanding of global environmental changes. However, research of the global-scale urban land expansion and its long-term environmental impacts has been restricted by the shortage of high-resolution multi-temporal global urban land data. Most of the contemporary global urban land products have the coarse resolution of 500 to 1000 m, and the pertinent data is available for one year or two years only. Inconsistency among these products further exacerbates issues faced by researchers. Therefore, it is still difficult to obtain a clear picture of how global urban land expands over a long historical period using solely contemporary global urban land products.
To overcome this issue, we developed a new multi-temporal global impervious surface product, which is derived from Landsat images pertaining to the 1990-2010 period with a five-year interval. This is the world’s first multi-temporal data set of global impervious surface at 30-m resolution. The production of this data requires sophisticated tools that provide functions for efficient image selection and extensive computation. The Google Earth Engine, which is an open access cloud-based computing platform with comprehensive image data (including the collection of Landsat images), can perfectly fulfill the technical needs for the extraction of global impervious surfaces from an extensive amount of Landsat images. Using this platform, we designed an approach for automatic impervious surface extraction by segmenting the calculated Normalized Urban Areas Composite Index (NUACI), a recently developed indicator for detecting impervious surfaces. We conducted the region-specific calibration and testing for this approach based on the stratification scheme of ‘urban ecoregions’ proposed in extant literature. In comparison with the existing global urban land products, our mapping results provide much more detailed information, while also yielding a significantly improved accuracy, as indicated by the Kappa values are 0.4280-0.4953 at the global level, and ~0.3306 (in China) and ~0.4163 (in the US) at the country level. These figures reveal that the produced multi-temporal global impervious surface data are of reasonably good quality and can substantially support ongoing and future research focusing on the dynamics of global urban land expansion.
Multi-temporal urban land products and reference datasets from 1990 - 2010 are available to download in: Baidu Drive (recommend) or Google Drive or FTP
If you have any comments and suggestions, please contact Prof. Xiaoping Liu: Liuxp3@mail.sysu.edu.cn
Last updated: 4 Aug., 2017