Assessment of soil moisture products in response to a heavy rainfall event
High resolution and accurate soil moisture maps have substantial implications in understanding and revealing the impact of surface fields on atmospheric processes. In this study, we have used the High-Resolution Land Data Assimilation System (HRLDAS) to generate surface fields (4 km spacing) during 2018 over the U.S. east coast. This study specifically focuses on the soil moisture response to Hurricane Florence (September 2018) to evaluate the performance of the simulated soil moisture. We first evaluated the assimilated soil moisture at different soil depths at point locations using the in-situ observations from SCAN, ECONET, and USCRN stations. Next, we spatially compared the assimilated surface soil moisture with other satellite and land model products including SMAP, CYGNSS, and NLDAS.
The point-based comparisons showed a better performance of the assimilated soil moisture at the surface (R2=0.74, unRMSE=0.06) compared to the deeper layers (R2=0.26 to 0.50, unRMSE=0.08). The spatial intercomparisons showed the spatial dependency of the model’s performance and their inability to accurately capture the jump in soil moisture resulting from the Hurricane-induced rainfall during post-landfall. SMAP showed superior performance in the coastal region while its accuracy was significantly reduced in the mountainous region primarily covered with deciduous forests. CYGNSS data could not accurately capture the moisture dynamic pre and post Hurricane landfall. Overall, HRLDAS showed the best performance (ubRMSE=5.13, R2=0.89), followed by SMAP, NLDAS, and CYGNSS. The study provides useful insights for selecting soil moisture datasets in land-atmospheric evaluations and delivers publicly available high-resolution surface fields that could be served as a pre-processor to mesoscale weather prediction models.
