Error estimation and data fusion of root zone soil moisture products over China based on the three corned hat method

Abstract

Root zone soil moisture (RZSM) plays a critical role in numerous ecological and environmental processes and holds significant importance for agriculture, hydrology, and climate studies. Although it can be estimated by hydrology or land surface models, the accuracy of such estimations is often limited. Data fusion offers a promising approach to improving RZSM estimation accuracy, yet few studies have explored this avenue. In our study, we address this gap by providing error estimation and data fusion for five RZSM datasets (ERA5-Land, MERRA2, CFSR, SMAP, GLDAS_NOAH2.1 (NOAH)) using the Three Cornered Hat (TCH) method. We evaluated the performance of the TCH method in assessing RZSM data products and in RZSM merging. Our results demonstrate that the TCH method accurately assesses the performance of RZSM products as validated against in situ measurements. Both in situ-based RMSE and TCH-based uncertainties reveal that MERRA2 and NOAH exhibit the best performance, followed by SMAP, CFSR and ERA5, with uncertainty medians of 0.019, 0.0187, 0.023, 0.021 and 0.028 (m³/m³), respectively. Comparisons of the accuracy for the TCH merged result and the individual RZSM product indicate that the merged result outperforms each individual product. The percentages of RMSE differences between the TCH merged result and the individual products less than −0.005 are 60.8 %, 62.3 %, 36.8 %, 41.7 %, and 51.2 % for CFSR, ERA5-Land, MERRA2, NOAH, and SMAP, respectively. These are significantly higher than the percentages of RMSE differences greater than 0.005. Given the TCH method's independence from in situ measurements, it is a promising option for RZSM data fusion. Overall, our study underscores the potential of the TCH method in evaluating RZSM products and performing data fusion to enhance RZSM estimation accuracy.