Deep learning for daily spatiotemporally continuity of satellite surface soil Moisture over eastern China in summer
Yang Zhou,Yan Zhang,Ruliang Wang,Haishan Chen,Qifan Zhao,Binshuo Liu,Qing Shao,Lu Cao,Shanlei Sun
DOI: https://doi.org/10.1016/j.jhydrol.2023.129308
IF: 6.4
2023-03-05
Journal of Hydrology
Abstract:Except assimilated data of soil moisture (SM), there is always a circumstance that not every spatial point has valid value over a specific region in a day, and there are invalid values or gaps between days in a time series at a spatial point, especially the satellite and ground-based data. Thus, those data are not spatiotemporally seamless on daily time scale, and the seamless data of surface SM is fundamental for research on the land–atmosphere interactions on various time scales. Deep learning methods of multi-perceptron (MLP) and convolution (Conv) neural networks are applied for obtaining spatiotemporally seamless SM over eastern China during May-September of 2015–2020 to combine the advantages of ground-based and remote-sensed data. Firstly, taking atmospheric and land variables as inputs, MLP is trained against the satellite products, including the level-2 product of Chinese FengYun 3C (FY3C), level-3 product of the US Soil Moisture Active Passive (SMAP), and level-2 neural network product of the European Soil Moisture and Ocean Salinity (SMOS). After that, spatiotemporally seamless SM is predicted by MLP, which is stable during training and validation. Compared with the in-situ SM, MLP SM inherits the characteristics of the satellite products, though the quality is slightly improved. Secondly, Conv uses the MLP SM as inputs and is trained against in-situ SM. The unbiased errors are reduced notably that unbiased correlation increases and root mean square errors decreases. At last, from the pattern correlation between precipitation and SM anomalies, the error-reduced Conv SM can well reflect the relationships forced out by the major patterns of rainfall, but the satellite products cannot. In addition, during the evaluation, SMAP has the best quality over eastern China, and SMOS is generally better than FY3C. However, after processed by Conv, the differences of the quality among different satellites are negligible, except some differences in the mean patterns of Conv SM, which is mainly associated with the properties of the satellite products.
geosciences, multidisciplinary,water resources,engineering, civil