| Abstract: |
Soil moisture, a component of Earth's water and energy budgets, is a key driver of both agriculture and weather. NASA's Soil Moisture Active Passive (SMAP) satellite mission provides global observations of surface soil moisture approximately once per day in the Corn Belt of the United States. A core validation site has been estabilished in the South Fork Iowa River (SFIR) watershed in the Corn Belt. The SFIR network contains 20 in situ stations over a SMAP pixel where 85% of land is devoted corn/soybean rotation.
SMAP-retrieved soil moisture (Dual Channel Algorithm, DCA) is too dry (bias: -0.038 m^3 m^-3) when compared to the SFIR in situ soil moisture. The retrievals exceed the mission noisiness goal of 0.04 m^3 m^-3 as quantified by the unbiased RMSE (ubRMSE: 0.064 m^3 m^-3). The DCA retrievals were examined rather than the baseline SCA due to observed year-to-year variation of vegetation water content (VWC) in agricultural regions; the SCA relies on climatology. The ESA's Soil Moisture Ocean Salinity (SMOS) mission, which operates at a similar temporal and spatial resolution to SMAP, is also too dry and too noisy in the SFIR (bias: -0.054 m^3 m^-3; ubRMSE: 0.064 m^3 m^-3). The following have been identified as potential sources of the dry bias due to their role in the soil moisture retrieval algorithm: cold bias in modeled surface temperature, low clay content in ancillary maps, and the assumption of a too smooth soil surface. These factors have been evaluated as potential sources of the SMOS dry bias in the SFIR by Walker et al., 2017; however, SMAP uses different ancillary data than SMOS. There is little L-band radio frequency interference (RFI) in the Midwest United States and as such it is highly unlikely to be a source of the SMAP dry bias in the SFIR.
SMAP defines the surface temperature as the arithmetic mean of the skin temperature and 0 - 10 cm soil temperature as given by the Goddard Earth Observing System Model, v5 (GEOS-5). The GEOS-5 temperature would have to consistently be 1.5 K colder than in situ to produce a dry bias in retrieved soil moisture over a full maize canopy (colder for bare soil). Most forecasting models are calibrated to have a zero bias; the European Center for Medium-Range Weather Forecasting (ECMWF) soil temperature used by SMOS has a 0.25 K warm bias compared to SFIR. The GEOS-5 temperature is not expected to be a source of the SMAP dry bias when evaluated against the SFIR in situ 5 cm soil temperature and airborne thermal infrared (TIR) obtained during SMAPVEX16-IA.
SMAP obtains clay content from the STATSGO 1 km dataset, re-gridded to the larger pixel scale, for soil moisture retrievals over the United States. STATSGO itself was compiled from detailed soil surveys and likely represents the SFIR well. The Mironov model, where SMAP accounts for the effects of clay content on the dielectric properties of soil, produces wetter retrievals than would be obtained using the Dobson model.
The SMAP retrieval algorithm currently parameterizes soil as semi-smooth (HR = 0.108) for croplands. Pinboard, gridboard, and lidar observations of soil surface roughness during SMAPVEX16-IA indicate that this value is too smooth for the SFIR. Using a rougher soil during retrievals alleviates the dry bias at the cost of decreasing retrieval sensitivity to actual soil moisture. Furthermore, soil surface roughness is not static throughout the year in agricultural regions due to farm management activities such as harvest and tillage disturbing the soil and subsequent smoothing by rainfall.
Future areas of investigation include: dynamic soil surface roughness, the croplands b-parameter, and spatial heterogeneity of soil moisture within the pixel. |
