Evaluation of model-derived root-zone soil moisture over the Huai river basin
Abstract. Root-zone soil moisture (RZSM) is crucial for water resource management, drought monitoring and sub-seasonal flood climate forecast. RZSM is not directly observable from space but various model-derived RZSM products are available at the global scale and are widely used. In this paper, a comprehensive quantitative evaluation of eight RZSM products is made over the Huai river basin (HRB) in China. A direct validation is performed using observations from 58 in situ soil moisture stations from 1 April 2015 to 31 March 2020. Attention is drawn to the potential factors increasing uncertainties of model-generated RZSM, such as errors on atmospheric forcings (precipitation, air temperature), soil properties, and model parameterizations. Results indicate that the Global Land Data Assimilation System Catchment Land Surface Model (GLDAS_CLSM) performs best among all RZSM products with the highest correlation coefficient (R) and lowest unbiased root-mean square error (ubRMSE): 0.503 and 0.031 m3 m−3, respectively. All RZSM products tend to overestimate the in situ soil moisture values, except for the Soil Moisture and Ocean Salinity (SMOS) L4 product, which underestimates RZSM. The underestimated SMOS L3 SSM associated with low physical surface temperature triggers the underestimation of RZSM in SMOS L4. The RZSM overestimation by other products can be explained by the overestimation of precipitation amount, precipitation event frequency (drizzle effects) and by the underestimation of air temperature. Besides, the overestimation of the soil clay content and the underestimation of the soil sand content in different LSMs leads to larger soil moisture values. The intercomparison of the eight RZSM products shows that MERRA-2 and SMAP L4 RZSM are the most correlated with one another. These products are based on the same LSM and on the same surface meteorological forcing generated from the National Aeronautics and Space Administration (NASA) GEOS-5. In addition, model parameterizations in different LSMs vary considerably, affecting the transfer and exchange of water and heat in the vadose zone.
En Liu et al.
Status: open (until 01 May 2023)
- RC1: 'Comment on hess-2023-33', Anonymous Referee #1, 06 Mar 2023 reply
En Liu et al.
En Liu et al.
Viewed (geographical distribution)
I thought it was an interesting study. I have one main comment: SMOS L4 is based on SMOS L3. From what I understand (Wigneron et al., 2021) ECMWF soil moisture data is used in the SMOS L3 retrieval algorithm. So should SMOS L4 be considered a remote sensing product or a modeled product? Please adapt the discussion according to my comment
Wigneron 2021, https://doi.org/10.1016/j.rse.2020.112238
Please discuss results considering a similar study made in China by Fan et al., RSE 2022 DOI: 10.1016/j.rse.2022.113283
Sentence: "Previous studies have illustrated that the VOD retrievals from SMOS may be noisy" is not objective. All VOD products can be considered as noisy, not only SMOS ones. It depends on location and product version (L2, L3 or SMOS-IC). Some SMAP, ASCAT, AMSR2 versions of VOD can be considered as much more noisy than SMOS VOD. Usually a sliding window smoothing technique (T = 7 -30 days) should be used for all VOD products.
Li et al., 2020 https://doi.org/10.1016/j.rse.2020.112208