CYGNSS Soil Moisture Retrieval and Intercomparison with SMAP

Surface soil moisture, linking the water, energy and carbon cycles over land surfaces, is crucial for weather and flooding forecasts and many other high impact societal applications, such as drought monitoring and crop yield predictions. Due to its significance, the Soil Moisture Active Passive (SMAP) mission was launched in January 2015 for global soil moisture observations every 2 to 3 days. Unfortunately, the failure of the SMAP radar in July 2015 is not allowing the production of soil moisture products at 9 km/3km spatial resolution although the SMAP radiometer continues to provide high quality soil moisture at 36 km resolution. The CYGNSS data has excellent potential to fill the data gap for high resolution soil moisture science.

To develop a physical-based algorithm for retrieving soil moisture from CYGNSS reflectivity, the development of a geophysical model function, which links the reflectivity to soil moisture, vegetation optical depth (VOD), and surface roughness is crucial. Towards this end, we have completed the matchup of CYGNSS data with SMAP Single Channel-V soil moisture, surface roughness from SMAP dual-channel algorithm, and Vegetation Water Content (VWC) derived from the MODIS-NDVI data. The matchup thresholds are 1 day in time difference and 15 km in spacing. The CYGNSS V2.1 Level 1 data were processed using the coherent reflection model to derive the surface reflectivity. The matchup data were binned as a function of SMAP soil moisture, surface roughness, VWC and CYGNSS incidence angle. Our binning analysis of the MODIS VWC and CYGNSS data has shown that the CYGNSS reflectivity in dB is approximately linearly related to the VWC, suggesting that we can relate the VOD to the VWC by a scaling parameter “b”. We have estimated the value of “b” parameters from the data. The average value is about 0.15, which confirms the “b” value used by SMAP for soil moisture retrieval.

Our data analysis further revealed the dependence of CYGNSS reflectivity on SMAP surface roughness, suggesting the significance of correcting surface roughness for soil moisture retrieval. To generate a self-consistent surface roughness value for CYGNSS retrieval, we use the SMAP soil moisture to estimate the flat surface reflectivity and NDVI VWC to estimate the VOD, which are then subtracted from the CYGNSS data to isolate the surface roughness contribution. The residuals are then used to estimate the surface roughness value at each CYGNSS footprint. Gridding and averaging were then performed to construct monthly surface roughness maps. Some temporal variability was observed and could be due to time varying surface roughness or calibration.

The retrieval of soil moisture from CYGNSS data was performed in two steps. The SMAP “b” value, NDVI VOD and CYGNSS surface roughness for May 2018 were used as ancillary data to remove surface roughness and vegetation impact. Subsequently, we retrieve the soil moisture from the corrected CYGNSS surface reflectivity. The spatial pattern of CYGNSS soil moisture is in general consistent with the SMAP soil moisture. However, there are regional biases between CYGNSS and SMAP soil moisture retrievals. We hypothesize that the differences were caused by time varying surface roughness. Validation of CYGNSS soil moisture with in-situ data from SMAP core cal/val sites is performed to provide quantitative assessment of CYGNSS soil moisture products and relative quality with respect to SMAP soil moisture.