1-km soil moisture predictions in the United States

Monthly and weekly soil moisture predictions in 2010 at 1-km spatial resolution using four different Machine Learning Methods integrated in the Satellite Soil Moisture based on a modular SOil Moisture SPatial Inference Engine (SOMOSPIE- Rorabaugh et al. 2019) (kernel-weighted k-nearest neighbors <KKNN>, Random Forests <RF>, Surrogate-Based Model <SBM> and a Hybrid Piecewise Polynomial Modeling Technique <HYPPO>). Data were acquired from the European Space Agency Climate Change Initiative (ESA CCI) soil moisture product version 6.1, 0.25-degrees spatial resolution. Modeled soil moisture layers are delivered for two regions in the conterminous United States. Each region encompasses a polygon of 7.5° x 3.75° (n = 450 pixels with 30 columns and 15 rows in the native resolution of the ESA CCI Soil moisture product). Region 1 <so called West Region> comprises an area of 275,516 km2. Region 2 <so called Midwest region> comprises an area of 283,499 km2. Predicted soil moisture values were validated by means two approaches, cross-validation using the ESA CCI estimates and independent ground-truth records from the North American Soil Moisture Database (currently known as the National Soil Moisture Network). Detailed methods and results of this dataset are described in: Llamas, R.M; Valera, Leobardo; Olaya, Paula; Taufer, Michela; Vargas, Rodrigo “Downscaling Satellite Soil Moisture based on a modular SOil Moisture SPatial Inference Engine (SOMOSPIE)”, Remote Sensing (submitted).

本数据集包含2010年空间分辨率为1千米的逐月、逐周土壤湿度预测结果，依托模块化卫星土壤湿度空间推断引擎（modular SOil Moisture SPatial Inference Engine, SOMOSPIE，Rorabaugh等，2019）集成的四种不同机器学习方法生成，分别为核加权k近邻法（kernel-weighted k-nearest neighbors, KKNN）、随机森林（Random Forests, RF）、基于代理模型（Surrogate-Based Model, SBM）以及混合分段多项式建模技术（Hybrid Piecewise Polynomial Modeling Technique, HYPPO）。数据集基础数据取自欧洲空间局气候变化倡议（European Space Agency Climate Change Initiative, ESA CCI）发布的6.1版土壤湿度产品，该产品原始空间分辨率为0.25度。本数据集提供美国本土两个区域的模拟土壤湿度图层，每个区域对应7.5°×3.75°的多边形范围（在ESA CCI土壤湿度产品的原生分辨率下，共包含450个像素点，即30列×15行）。其中区域1（又称西部区域）面积为275516平方千米，区域2（又称中西部区域）面积为283499平方千米。预测得到的土壤湿度值通过两种方式开展验证：一是采用ESA CCI估算结果进行交叉验证，二是使用来自北美土壤湿度数据库（North American Soil Moisture Database，现更名为国家土壤湿度网络（National Soil Moisture Network））的独立地面实测记录进行验证。本数据集的详细方法与结果已发表于：Llamas, R.M.; Valera, Leobardo; Olaya, Paula; Taufer, Michela; Vargas, Rodrigo，《基于模块化卫星土壤湿度空间推断引擎（SOMOSPIE）的卫星土壤湿度降尺度研究》，《遥感》（Remote Sensing，已投稿）。