Global long term daily 1km surface soil moisture dataset with physics informed machine learning

Although soil moisture is a key factor of hydrologic and climate applications, global continuous high resolution soil moisture datasets are still limited. Here we use physics-informed machine learning to generate a global, long-term, spatially continuous high resolution dataset of surface soil moisture, using International Soil Moisture Network (ISMN), remote sensing and meteorological data, guided with the knowledge of physical processes impacting soil moisture dynamics. Global Surface Soil Moisture (GSSM1km) provides surface soil moisture (0-5 cm) at 1 km spatial and daily temporal resolution over the period 2000-2020. The performance of the GSSM1km dataset is evaluated with testing and validation datasets, and via inter-comparisons with existing soil moisture products. The root mean square error of GSSM1km in testing set is 0.05 cm3/cm3, and correlation coefficient is 0.9. In terms of the feature importance, Antecedent Precipitation Evaporation Index (APEI) is the most important significant predictor among 18 predictors, followed by evaporation and longitude. GSSM1km product can support the investigation of large-scale climate extremes and long-term trend analyses. Due to the whole dataset for the global scale is too big (779GB) to deposit at once, we uploaded the data in the Netherlands to figshare. For other areas, the data is stored in Google Earth Engine (https://code.earthengine.google.com/?asset=users/qianrswaterr/GlobalSSM1km0509), and we provide codes to download our data (https://code.earthengine.google.com/4b577bb83981e1ac43fd77127cfbdb4a). Due to the dataset is exported from Google Earth Engine, the bandNames can’t display in ArcGIS, the band is displayed as band1, band2,…. Just in case other softwares also can't display, I put the bandNames in the csv file “bandNames2000-2020”. The full dataset is also available at: https://data.tpdc.ac.cn/en/data/a11479c5-b0a8-40b8-8092-f47719a6c882.

尽管土壤水分是水文与气候应用的关键因子之一，但全球连续高分辨率土壤水分数据集仍较为匮乏。本研究借助物理引导机器学习方法，以国际土壤水分网络（International Soil Moisture Network, ISMN）、遥感与气象数据为数据源，结合影响土壤水分动态的物理过程先验知识，构建了一套全球、长时序、空间连续的高分辨率地表土壤水分数据集。全球地表土壤水分数据集（Global Surface Soil Moisture, GSSM1km）的时空分辨率为1千米空间分辨率与逐日时间分辨率，覆盖2000年至2020年时段，反演土层深度为0~5厘米的地表土壤水分。
本研究通过测试集与验证集，以及与现有土壤水分产品的交叉对比，对GSSM1km数据集的性能进行了评估。GSSM1km数据集在测试集中的均方根误差为0.05立方厘米/立方厘米，相关系数为0.9。在特征重要性分析中，前期降水蒸发指数（Antecedent Precipitation Evaporation Index, APEI）是18个预测因子中贡献度最高的显著变量，其次为蒸发量与经度。GSSM1km数据集可支撑大尺度气候极端事件研究与长时序趋势分析。
由于全球完整数据集体量过大（779GB），无法一次性上传存储，本研究将其上传至荷兰地区的Figshare平台。其余区域的数据则存储于谷歌地球引擎（Google Earth Engine, GEE）平台，其访问地址为：https://code.earthengine.google.com/?asset=users/qianrswaterr/GlobalSSM1km0509，同时提供了数据集下载代码，代码地址为：https://code.earthengine.google.com/4b577bb83981e1ac43fd77127cfbdb4a。
由于该数据集由谷歌地球引擎导出，其波段名称无法在ArcGIS中正常显示，仅会以band1、band2……的形式展示。为避免其他软件亦出现此类问题，本研究将波段名称存储于"bandNames2000-2020.csv"文件中。完整数据集亦可通过以下地址获取：https://data.tpdc.ac.cn/en/data/a11479c5-b0a8-40b8-8092-f47719a6c882。

搜集汇总

背景与挑战

背景概述

该数据集是一个全球长期每日1公里分辨率的地表土壤湿度数据集（2000-2020年），采用物理信息机器学习方法生成，具有高精度（测试集RMSE=0.05 cm3/cm3，R=0.9），适用于大尺度气候研究和趋势分析。数据存储在Google Earth Engine等平台。