Assimilation of GNSS-R Soil Moisture Retrievals in a Hydrological Model for Root Zone Soil Moisture Estimation

The distribution of water in soil profile from surface to root zone can impact hydrological cycles, terrestrial processes and atmospheric dynamics. Soil moisture profile estimation is of critical importance in meteorological applications, as it directly influences weather patterns, climate modeling, and various atmospheric processes. The distribution of moisture within the soil affects the exchange of water and energy between the land surface and the atmosphere, significantly contributing to the formation of clouds, precipitation, and temperature regulation. Accurate soil moisture data is essential for enhancing weather forecasts, particularly in predicting the onset and severity of droughts and floods. Furthermore, it improves the precision of numerical weather prediction models by providing crucial information on surface fluxes, thereby aiding in the understanding and simulation of regional and global climate variations. Consequently, soil moisture profiling is a key tool in meteorology, enabling more reliable climate predictions and better-informed responses to weather-related disasters.

The Cyclone Global Navigation Satellite System (CYGNSS) is a constellation of eight satellites launched by NASA in 2016, which uses Global Navigation Satellite System Reflectometry (GNSS-R) for the purpose of sensing ocean surface winds in the tropics. GNSS-R has also demonstrated considerable utility in land applications, particularly in soil moisture estimation. A soil moisture product is now generated from CYGNSS observables. These measurements facilitate the retrieval of soil moisture data with high temporal frequency. GNSS-R observations, however, are in L-band, which has a penetration depth of only a few cm.

Assimilation of L-band radiometric observations (e.g. SMAP) into hydrological models have been used to estimate the evolution of the sub-surface soil moisture profile at larger depths. This study presents a first example of such a data assimilation approach applied to GNSS-R retrievals.

HYDRUS-1D is a widely used hydrological model for simulating water and heat movement in one-dimensional vadose zone with variable water saturation. The model employs finite element analysis to simulate soil moisture distribution over time, based on a set of input parameters. The root zone soil moisture profile can be estimated by assimilating satellite-derived surface observations, constrained by the hydrological model.

In this study, we initially implemented an ensemble Kalman filter method to assimilate the Level 2 CYGNSS soil moisture product with HYDRUS-1D simulations, using meteorological data from the North American Land Data Assimilation System (NLDAS). The model evaluation was conducted using multi-depth in-situ measurements collected from the Texas Soil Moisture Observation Network (TxSON) sites over the period of 2017 to 2022. The results indicate a slight improvement in soil moisture estimation at all depths when compared to either observation or simulation alone.