Evaluation of Surface Fractional Water Impacts on SMAP Soil Moisture Retrieval

Fractional water (FW) correction of satellite microwave brightness temperature (Tb) observations is a prerequisite for accurate soil moisture (SM) mapping over mixed land and water areas. Failure to adequately account for the influence of FW within a satellite footprint can bias the Tb signal over land, resulting in over- or under-estimates of SM retrieval. Here, we evaluated the FW impacts on SMAP L-band (1.4 GHz) SM retrievals using three global FW data records generated for SMAP 36-km resolution grid cells. The FW records were processed, analyzed, and compared on the Google Earth Engine (GEE) platform. Three global FW data records are available on the GEE: the NASA MOD44W v5 static FW dataset (circa 2000), the MOD44W v6 multi-year (2000-2015) dynamic FW dataset, and the JRC Landsat-based monthly FW dataset averaged for non-frozen months of year 2000 and the period between 2015 and 2020. Overall, MOD44W v6 showed higher global consistency with Landsat compared with MOD44W v5 in terms of bias (0.005 vs -0.006), Root Mean Square Difference (RMSD; 0.017 vs 0.026) and correlation (0.783 vs 0.705) for grid cells with 0.005 Landsat FW 0.1. MOD44W v5 showed a wet bias ( 0.005 over 18.34% of the global land area) relative to MOD44W v6, and attributed to a widespread FW decade-long drying trend not captured by the static MOD44W v5. The Landsat record showed higher FW (0.005) over 10.65% and 24.98% of the northern high latitudes ( 45) relative to the respective MOD44W v5 and v6 datasets, which was attributed to its enhanced capabilities in detecting small water bodies. Relatively large FW inter-annual variability (standard deviation 0.005 over 6.16% of the global land area) was identified in MOD44W v6 over major river/lake areas and regional change hot spots, which may reflect reservoir construction/operation and recent climate extremes. We evaluated the impacts of FW dynamics on potential wet bias on SMAP soil moisture retrievals over northern high latitude and Mississippi basin sub-regions. Our results revealed SM retrieval biases due to the underlying ancillary FW data, retrieval uncertainties associated with sensor resolution and water dynamics (flood and drought) not captured by MOD44W v5. Updated ancillary data accounting for FW dynamics is expected to further improve SMAP SM retrieval accuracy.