AUTHOR=Yang Ting , Sun Zhigang , Wang Jundong , Li Sen TITLE=Daily Spatial Complete Soil Moisture Mapping Over Southeast China Using CYGNSS and MODIS Data JOURNAL=Frontiers in Big Data VOLUME=4 YEAR=2022 URL=https://www.frontiersin.org/journals/big-data/articles/10.3389/fdata.2021.777336 DOI=10.3389/fdata.2021.777336 ISSN=2624-909X ABSTRACT=

Daily spatial complete soil moisture (SM) mapping is important for climatic, hydrological, and agricultural applications. The Cyclone Global Navigation Satellite System (CYGNSS) is the first constellation that utilizes the L band signal transmitted by the Global Navigation Satellite System (GNSS) satellites to measure SM. Since the CYGNSS points are discontinuously distributed with a relativity low density, limiting it to map continuous SM distributions with high accuracy. The Moderate-Resolution Imaging Spectroradiometer (MODIS) product (i.e., vegetation index [VI] and land surface temperature [LST]) provides more surface SM information than other optical remote sensing data with a relatively high spatial resolution. This study proposes a point-surface fusion method to fuse the CYGNSS and MODIS data for daily spatial complete SM retrieval. First, for CYGNSS data, the surface reflectivity (SR) is proposed as a proxy to evaluate its ability to estimate daily SM. Second, the LST output from the China Meteorological Administration Land Data Assimilation System (CLDAS, 0.0625° × 0.0625°) and MODIS LST (1 × 1 km) are fused to generate spatial complete and temporally continuous LST maps. An Enhanced Normalized Vegetation Supply Water Index (E-NVSWI) model is proposed to estimate SM derived from MODIS data at high spatial resolution. Finally, the final SM estimation model is constructed from the back-propagation artificial neural network (BP-ANN) fusing the CYGNSS point, E-NWSVI data, and ancillary data, and applied to get the daily continuous SM result over southeast China. The results show that the estimation SM are comparable and promising (R = 0.723, root mean squared error [RMSE] = 0.062 m3 m−3, and MAE = 0.040 m3 m−3 vs. in situ, R = 0.714, RMSE = 0.057 m3 m−3, and MAE = 0.039 m3 m−3 vs. CLDAS). The proposed algorithm contributes from two aspects: (1) validates the CYGNSS derived SM by taking advantage of the dense in situ networks over Southeast China; (2) provides a point-surface fusion model to combine the usage of CYGNSS and MODIS to generate the temporal and spatial complete SM. The proposed approach reveals significant potential to map daily spatial complete SM using CYGNSS and MODIS data at a regional scale.