2013-2022年蒙古高原逐年生长季地表水分布数据集

A dataset of annual surface water distribution in the growing season on the Mongolia Plateau from 2013 to 2022

蒙古高原地处干旱半干旱地区,水文水资源是其资源环境发展的重要制约条件。掌握蒙古高原的水体时空分布情况对于提示水资源和水环境时空特征及其在区域气候变化和防灾减灾方面的影响和响应具有重要意义。然而,由于该区域辽阔且跨越中蒙两个国家,精确自动化获取流域尺度的大范围长时序水体面临很大挑战。本研究采用本地深度学习训练和谷歌地球引擎(Google Earth Engine,GEE)分布式计算相结合的方法,对GEE赋予深度学习计算能力,使GEE可以快速自动化部署深度学习模型。基于此,完成蒙古高原2013-2022年逐年生长季地表水分布的获取,空间分辨率为30 m。人工选择验证点5000个,总体验证为88.0%。数据集为TIFF栅格形式,以5°×5°×10年的形式存储为28个瓦片影像,数据量为339 MB,压缩后为88.1 MB,在原始数据格式下为189 GB。本数据集采用的模型方法可以自动化、高效地在云端进行水体制图,为干旱半干旱地区大范围、长时序、高效率的水体的自动化处理提供了可能,具有应用和推广价值。

Mongolia Plateau is located in arid and semi-arid areas,and hydrology and water resources are important constraints for the development of its resources and environment.Grasping the temporal and spatial distribution of water bodies on the Mongolian Plateau is of great significance for indicating the temporal and spatial characteristics of water resources and the water environment and their impacts on and responses to regional climate change as well as disaster prevention and reduction.However,as the vast Plateau spans both China and Mongolia,it is a great challenge to accurately and automatically obtain large-scale and long time series water bodies at the basin scale.In this research,we adopted the method of combining local deep learning training and Google Earth Engine(GEE)distributed computing to endow GEE with deep learning computing capabilities so that GEE could rapidly and automatically deploy deep learning models.Based on this,we obtained the distribution of surface water in the growing season of the Mongolia Plateau from 2013 to 2022 with a spatial resolution of 30 meters.5,000 verification points were manually selected,and the overall verification rate was 88.0%.The dataset is in the form of TIFF grid,containing 28 tile images of with 5°×5°×10 years,with a data volume of 339 MB(88.1 MB compressed,189 GB in RAW).The data volume in the raw format is 189 GB.With the method used in this dataset,users can automatically and efficiently map water bodies in the cloud platform,which makes it possible to automatically and efficiently process large-scale and long-time series water bodies in arid and semi-arid regions.This is a valuable dataset for application and promotion.