| Summary: | Soil moisture (SM) plays a crucial role in the water and energy flux exchange between the atmosphere and the land surface. Remote sensing and modeling are two main approaches to obtain SM over a large-scale area. However, there is a big difference between them due to algorithm, spatial-temporal resolution, observation depth and measurement uncertainties. In this study, an assessment of the comparison of two state-of-the-art remotely sensed SM products, Soil Moisture Active Passive (SMAP) and European Space Agency Climate Change Initiative (ESACCI), and one land surface modeled dataset from the North American Land Data Assimilation System project phase 2 (NLDAS-2), were conducted using 17 permanent SM observation sites located in the Southern Great Plains (SGP) in the U.S. We first compared the daily mean SM of three products with in-situ measurements; then, we decompose the raw time series into a short-term seasonal part and anomaly by using a moving smooth window (35 days). In addition, we calculate the daily spatial difference between three products based on in-situ data and assess their temporal evolution. The results demonstrate that (1) in terms of temporal correlation R, the SMAP (R = 0.78) outperforms ESACCI (R = 0.62) and NLDAS-2 (R = 0.72) overall; (2) for the seasonal component, the correlation R of SMAP still outperforms the other two products, and the correlation R of ESACCI and NLDAS-2 have not improved like the SMAP; as for anomaly, there is no difference between the remotely sensed and modeling data, which implies the potential for the satellite products to capture the variations of short-term rainfall events; (3) the distribution pattern of spatial bias is different between the three products. For NLDAS-2, it is strongly dependent on precipitation; meanwhile, the spatial distribution of bias represents less correlation with the precipitation for two remotely sensed products, especially for the SMAP. Overall, the SMAP was superior to the other two products, especially when the SM was of low value. The difference between the remotely sensed and modeling products with respect to the vegetation type might be an important reason for the errors.
|
|---|
