Table 2.
Overall MPA (%) and Brier scores for the 2016–2020 study period estimated between each DL data model and ERA5 and weather station derived soil FT observations.
| ERA | Weather stations | |||
|---|---|---|---|---|
| MPA | Brier score | MPA | Brier score | |
| ERA5 | 100 | N/A | 85.0 | N/A |
| SMAP | 84.5 | 0.1038 | 89.4 | 0.0883 |
| AMSR18 | 89.5 | 0.07661 | 89.3 | 0.0880 |
| AMSR36 | 87.9 | 0.0851 | 89.2 | 0.0892 |
| AMSR18 + 36 | 90.0 | 0.0773 | 90.0 | 0.0890 |
| AMSR + SMAP (a.m.) | 92.9 | 0.0508 | 91.0 | 0.0769 |
| AMSR + SMAP (p.m.) | 92.4 | 0.0581 | 91.1 | 0.0779 |
The ERA5 record is compared against itself and the weather station observations as a baseline for gaging model performance. The different DL data models were derived using SMAP 1.4 GHz (SMAP), AMSR2 18.7 GHz (AMSR18), AMSR2 36.5 GHz (AMSR36), AMSR2 18.5 and 36.5 GHz (AMSR18 + 36), and combined AMSR2 and SMAP (AMSR + SMAP) TB inputs for morning (6 a.m.) conditions. The AMSR + SMAP results include both morning and evening (6 a.m./p.m.) periods. The best performing metrics are highlighted (in bold).