High temporal resolution quasi-global landscape soil freeze–thaw map from spaceborne GNSS-R technology and SMAP radiometer measurements

Abstract

Landscape freeze–thaw (F/T) state parameters are an integral part of research on terrestrial hydrological processes, vegetation growth dynamics, and terrestrial–atmospheric trace gas exchange. Therefore, the development of large-scale, continuous, and rapid F/T observation records is essential. This study proposes a scheme to effectively retrieve quasi-global daily soil F/T states from spaceborne Global Navigation Satellite System Reflectometry (GNSS-R) observations. This method takes into account the GNSS-R surface reflectivity and its influence factor. Specifically, the method employed surface reflectivity derived from Cyclone Global Navigation Satellite System (CYGNSS) data from 2018 to 2020, along with vegetation opacity data from Soil Moisture Active Passive (SMAP) as input layers in random forests for F/T retrieval. This approach provides F/T results on a quasi-global scale, does not rely on continuous F/T observations, and employs the least ancillary data. To validate the retrieval results, SMAP F/T data over the same period and temperature data from more than 100 ground stations were also used as reference. Results showed that the F/T retrieval accuracy of CYGNSS was 95.5% and 82.0% with respect to the SMAP data and data from independent in situ stations, respectively. Given the high temporal coverage of the CYGNSS F/T results, the temporal coverage improvement is 23.3% compared with the actual SMAP observations on a daily scale. Furthermore, as gap-filling data complemented SMAP, CYGNSS data can effectively improve the SMAP F/T sample interval (accuracy improvement of 26.0%). More importantly, the F/T accuracy of the CYGNSS and SMAP combination based on monthly performance is 85.1%, i.e., 20.5% better than that of the existing SMAP daily F/T product. Overall, the quasi-global daily F/T results obtained from CYGNSS and SMAP measurements can complement existing databases with well-defined accuracy.