Evaluation of satellite and reanalysis estimates of surface and root-zone soil moisture in croplands of Jiangsu Province, China


High-quality and long-term surface soil moisture (SSM) and root-zone soil moisture (RZSM) data are critical for agricultural water management of Jiangsu province, which is a major agricultural province in China. However, few studies assessed the accuracy of SSM and RZSM datasets in croplands of Jiangsu province. The study addressed this gap by firstly using observations from ninety-one sites to assess thirteen satellite and model-based SSM products (Advanced Scatterometer (ASCAT), European Space Agency Climate Change Initiative (ESA CCI) Combined/Passive/Active, Soil Moisture and Ocean Salinity in version IC (SMOS-IC), Land Parameter Retrieval Model (LPRM) Advanced Microwave Scanning Radiometer 2 (AMSR2), Soil Moisture Active Passive (SMAP)-Multi-Temporal Dual-channel Algorithm (MTDCA)/Level 3 (L3)/Level 4 (L4)/SMAP-INRAE-BORDEAUX (IB)/Multi-channel Collaborative Algorithm (MCCA), the fifth generation of the land component of the European Centre for Medium-Range Weather Forecasts atmospheric reanalysis (ERA5-Land), and the Noah land surface model driven by Global Land Data Assimilation System (GLDAS-Noah)), and four RZSM products (ERA5-Land, GLDAS-Noah, SMAP-L4 and ESA CCI (retrieved using ESA CCI Combined SSM coupled with an exponential filter)). We also inter-compared time-invariant and time-variant Triple Collocation Analysis (TCA)-based R with in situ-based R calculated using SSM anomalies. Various evaluation strategies were compared using different groups of available sites and temporal samplings. Our results showed that the model-based and combined SSM products (i.e., ERA5-Land, SMAP-L4, ESA CCI Combined/Passive/Active, GLDAS-Noah, ASCAT) performed better than the other SSM products and ERA5-Land, SMAP-L4 and ESA CCI RZSM generally performed better than the GLDAS-Noah RZSM product with higher R. Similar performance rankings were observed among time-invariant and time-variant TCA-R and in situ-based R, in which the TCA-R values for all SSM datasets were higher than the in situ-based R as the representativeness errors of the in situ measurements may bias in situ-based R. The accuracy of the ESA CCI, GLDAS-Noah and ERA5-Land SSM products was expected to be enhanced by considering the water effect and high uncertainties were observed for MTDCA and SMAP-MCCA SSM over dense vegetation periods and regions. Also, it is important to select appropriate evaluation strategies to conduct the SSM and RZSM evaluations according to the situation as the available sites and temporal samplings may bias the evaluation results.

Remote Sensing of Environment