GISD30: global 30 m impervious-surface dynamic dataset from 1985 to 2020 using time-series Landsat imagery on the Google Earth Engine platform

作者：Xiao Zhang, Liangyun Liu, Tingting Zhao, Yuan Gao, Xidong Chen, Mi Jun · 发表于：Earth system science data · 年份：2022 · DOI：10.5194/essd-14-1831-2022 · 被引用次数：211 · 研究领域：Urban Heat Island Mitigation、Land Use and Ecosystem Services、Urban Green Space and Health

Abstract. Accurately mapping impervious-surface dynamics has great scientific significance and application value for research on urban sustainable development, the assessment of anthropogenic carbon emissions and global ecological-environment modeling. In this study, a novel and automatic method of combining the advantages of spectral-generalization and automatic-sample-extraction strategies was proposed, and then an accurate global 30 m impervious-surface dynamic dataset (GISD30) for 1985 to 2020 was produced using time-series Landsat imagery on the Google Earth Engine cloud computing platform. Firstly, the global training samples and corresponding reflectance spectra were automatically derived from prior global 30 m land-cover products after employing the multitemporal compositing method and relative radiometric normalization. Then, spatiotemporal adaptive classification models, trained with the migrated reflectance spectra of impervious surfaces from 2020 and transferred pervious-surface samples in each epoch for every 5∘×5∘ geographical tile, were applied to map the impervious surface in each period. Furthermore, a spatiotemporal-consistency correction method was presented to minimize the effects of independent classification errors and improve the spatiotemporal consistency of impervious-surface dynamics. Our global 30 m impervious-surface dynamic model achieved an overall accuracy of 90.1 % and a kappa coefficient of 0.865 using 23 322 global time-series validation samp...