Evaluating model-simulated and satellite-derived SM using in situ observations under different environment conditions in China

Soil moisture (SM) is a key variable in understanding the climate system through its controls on the land surface energy and water budget. Large scale SM products have become increasingly available thanks to development in microwave remote sensing and land surface modeling. Comprehensive assessments on the reliability of satellite-derived and model-simulated SM products are essential for their improvement and application. In this research, the active, passive and combined Climate Change Initiative (CCI V04.2) SM products and the China Land Data Assimilation System (CLDAS V2.0) SM products were evaluated by comparing with in situ observed data over three networks in China: Hebi, Naqu and Heihe. The three sites have different environmental conditions and sensor densities, providing observations covering more than 2 years. Four statistic scores were calculated: R (considering both original data and anomalies), Bias, RMSE, ubRMSE. TC (Triple Collocation) analysis was also carried out in which uncertainties in observations are taken into account. Results indicate that the performance of the two SM products varies between the monitoring networks. For Naqu site, both products show good performance, with CCI-SM showing slightly higher R and lower ubRMSE. For Hebi site, CLDAS-SM performs better than CCI-SM, whereas for Heihe site, CLDAS-SM performs worse than CCI-SM. The expected uncertainty (0.04 m³/m³) can be achieved in Naqu and Heihe site by CCI-SM, and in Hebi and Naqu site by CLDAS-SM, which is quite encouraging. The two products agree in terms of sign of the Bias value, which is positive in Hebi and negative in Naqu and Heihe. Among all the three networks, Heihe site exhibits the lowest accuracy due to its complicated terrain and heterogeneous land surface. R_anom of CLDAS-SM in Heihe is close to 0, indicating its inability to capture short term variability. TC results reveal that for Naqu site the observation data have quite good qualities, while for Hebi site CLDAS-SM is more approximate to ‘ground truth’ than in situ observations, suggesting a refinement of network maybe needed in the future. Overall, the two products are complementary. CLDAS-SM performs better in populated area (e.g., Hebi) where meteorological forcing is more accurate and CCI-SM performs better in remote areas (Naqu, Heihe) where RFI is usually low. More reliable validation networks are needed in the future to comprehensively understand the advantages and disadvantage of the two SM products in China.