Evaluation of nine major satellite soil moisture products in a typical subtropical monsoon region with complex land surface characteristics

Abstract

Assessing accuracies of satellite soil moisture (SM) products in areas with strong anthropogenic activities, abundant precipitation, and dense vegetation is important but limited. In this study, performances of nine satellite SM products, including the European Space Agency Climate Change Initiative (ESA CCI), and the ascending and descending products of the Advanced Microwave Scanning Radiometer 2 (AMSR2), Chinese Fengyun-3B (FY3B), Chinese Fengyun-3C (FY3C), and Soil Moisture Active Passive (SMAP), were evaluated against in-situ SM in the Yangtze River Delta (YRD) from 2015 to 2018. Results showed that the ESA CCI outperformed other products, with averaged correlation coefficient (R) of 0.522, correlation coefficient for SM anomalies (Rano) of 0.419, and unbiased root mean square error (ubRMSE) of 0.040 m3 m−3, followed by the SMAP_A and SMAP_D. Overestimation was observed in the AMSR2 products (Bias >0.081 m3 m−3), while underestimation was found in the FY3B and FY3C products (Bias < −0.104 m3 m−3). Moreover, R values decreased with the increase in water area percentage (WAP), land cover heterogeneity (GSI), fractional vegetation cover (FVC), and precipitation (PRE). Threshold values were observed at 3% for the WAP, 0.05 for the FVC, and 900 mm yr−1 for the PRE, below which sharp declines of accuracy were observed. The WAP, GSI, and FVC were divided into five sub-classes with their values increasing with the sub-class level (I to V). Correlations coefficients between these perturbing factors and R (correlation between satellite and in-situ SM) increased as sub-class levels increased. Under sub-classes I, II, and III, these three factors had limited influences on accuracies of satellite SM products. However, under sub-classes IV and V, significant impacts of these three perturbing factors were observed. Moreover, when sub-classes of the PRE increased from I to V, correlations between Bias and PRE decreased first and then increased. These findings can help to determine suitable satellite SM products in the YRD and similar areas.