SMAP enhanced Level-2 radiometer soil moisture data (SPL2SMP_E) are derived from SMAP Enhanced L1C Radiometer Half-Orbit 9 km EASE-Grid Brightness Temperatures, Version 2 (SPL1CTB_E).

SMAP enhanced Level-2 radiometer soil moisture data (SPL2SMP_E) are derived from SMAP Enhanced L1C Radiometer Half-Orbit 9 km EASE-Grid Brightness Temperatures, Version 2 (SPL1CTB_E). Utilizing a baseline plus four optional soil moisture algorithms discussed below, SMAP brightness temperatures are converted into an estimate of the 0-5 cm surface soil moisture in units of m³/m³.

For information regarding the Backus-Gilbert optimal interpolation algorithm used to enhance these data, refer to the SPL1CTB_E user guide.

The following information has been adapted from O'Neill et al. (2016).

Algorithm Inputs and Outputs

The main input to the processing algorithm for this product is the water-corrected brightness temperatures in the SMAP Enhanced L1C Radiometer Half-Orbit 9 km EASE-Grid Brightness Temperatures, Version 2 (SPL1CTB_E) product that contains the time-ordered, geolocated, calibrated SMAP enhanced Level-1B radiometer brightness temperatures (L1B_TB_E) that have been resampled to the fixed 9 km EASE-Grid 2.0. In addition to general geolocation and calibration, the enhanced Level-1B brightness temperature data have also been corrected for atmospheric effects, Faraday rotation, and low-level RFI effects prior to regridding. If the RFI encountered is too large to be corrected, the brightness temperature data are flagged accordingly and no soil moisture retrieval is attempted. Refer to the SPL1CTB_E ATBD for additional details.

In addition to brightness temperature observations, the SPL2SMP_E algorithm also requires ancillary data sets for the soil moisture retrieval. The specific parameters and sources of these and other externally provided ancillary data are listed in Section 6: Ancillary Datasets of the baseline SPL2SMP ATBD. In order for soil moisture to be retrieved accurately, a variety of global static and dynamic ancillary data are required. Static ancillary data are data which do not change during the mission, while dynamic ancillary data require periodic updates in time frames ranging from seasonally to daily. Static data include parameters such as permanent masks (land, water, forest, urban, mountain, etc.), the grid cell average elevation and slope derived from a Digital Elevation Model (DEM), permanent open water fraction, and soils information (primarily sand and clay fraction). Dynamic ancillary data include land cover, surface roughness, precipitation, vegetation parameters, and effective soil temperatures.

Ancillary data are also employed to set flags that help to determine either specific aspects of the processing, such as corrections for open water and frozen ground, or the quality of the retrievals, such as the precipitation flag. Refer to the Data Flags below. All input data to the SPL2SMP_E process are pre-mapped using the 9 km EASE-Grid 2.0 projection and are then aggregated at a spatial extent that is approximately the same as the native resolution (approximately 36 km) of the SMAP radiometer. Other parameters used by the SPL2SMP_E algorithm include a freeze/thaw flag, an open water fraction, and a vegetation index.

The primary contents of output SPL2SMP_E data are the retrieved soil moisture and associated Quality Assessment (QA) flags, as well as the values of the ancillary parameters needed to retrieve the output soil moisture for that grid cell.

Soil Moisture Algorithms

Decades of research by the passive microwave soil moisture community have resulted in a number of viable soil moisture retrieval algorithms that can be used with SMAP brightness temperature data. The European Space Agency (ESA) Soil Moisture and Ocean Salinity Mission (SMOS) mission currently flies an aperture synthesis L-band radiometer which produces brightness temperature data at multiple incidence angles over the same ground location. The baseline SMOS retrieval algorithm is based on the tau-omega model described in the SPL2SMP ATBD, Section 2.1: Physics of the Problem, but utilizes the SMOS multiple-incidence-angle capability to retrieve soil moisture. SMAP retrievals are also based on the tau-omega model, but use the constant-incidence-angle brightness temperature data produced by the SMAP conically-scanning radiometer. Other necessary parameters in the retrieval are obtained as ancillary data.

Prior to implementing the actual soil moisture retrieval, a preliminary step in the processing is to perform a Water/Land Contamination Correction to the input brightness temperature data for cases where a significant percentage of the grid cell contains a mix of land and open water. This procedure corrects for anomalous soil moisture values seen near coastlines in the previous version and should result in less rejected data due to waterbody contamination. In addition, a Soil Depth Correction is performed in order to reduce the dry bias seen when comparing SMAP data to in situ data from the core validation sites.

The Version 2 SPL2SMP_E product contains soil moisture retrieval fields produced by a baseline and several optional algorithms (Refer to Table 3). Inside an SPL2SMP_E file, the soil_moisture field is the one that links to the retrieval result produced by the currently-designated baseline algorithm, the Single Channel Algorithm V-pol (SCA-V). At present, the operational SPL2SMP_E Science Production Software (SPS) produces and stores soil moisture retrieval results from the following five algorithms:

Given results from SPL2SMP_E Calibration/Validation (Cal/Val) analyses, the SCA-V algorithm seems to deliver better performance than the SCA-H algorithm, which was designated as the pre-launch baseline retrieval algorithm. For this reason, the SCA-V is designated as the current baseline algorithm for the validated release of SPL2SMP_E. However, all five algorithms will be continuously assessed; the choice of the operational algorithm for the validated release of the product will be evaluated on a regular basis as analyses of new observations and Cal/Val data become available, and algorithm parameters are tuned based on a longer SMAP radiometer brightness temperature time series record.

All five algorithms operate on the same zeroth-order microwave emission model commonly known as the tau-omega model. In essence, this model relates brightness temperatures (SMAP L1 observations) to soil moisture (SMAP L2 retrievals) through ancillary information (e.g. soil texture, soil temperature, and vegetation water content) and a soil dielectric model. The algorithms differ in their approaches to solve for soil moisture from the model under different constraints and assumptions. The following sections provide concise descriptions of each algorithm. Further details are provided in O'Neill et al., 2016.

Baseline Single Channel Algorithm V-pol (SCA-V)

In the SCA-V, the vertically polarized brightness temperature (TB) observations are converted to emissivity using a surrogate for the physical temperature of the emitting layer. The derived emissivity is corrected for vegetation and surface roughness to obtain the soil emissivity. The Fresnel equation is then used to determine the dielectric constant from the soil emissivity. Finally, a dielectric mixing model is used to solve for the soil moisture given knowledge of the soil texture. Analytically, SCA-V attempts to solve for one unknown variable (soil moisture) from one equation that relates the vertically polarized TB to soil moisture. Vegetation information is provided by a 13-year climatological data base of global Normalized Difference Vegetation Index (NDVI) and a table of parameters based on land cover types.

Single Channel Algorithm H-pol (SCA-H)

The SCA-H is similar to SCA-V, in that the horizontally polarized TB observations are converted to emissivity using a surrogate for the physical temperature of the emitting layer. The derived emissivity is corrected for vegetation and surface roughness to obtain the soil emissivity. The Fresnel equation is then used to determine the dielectric constant. Finally, a dielectric mixing model is used to obtain the soil moisture given knowledge of the soil texture. Analytically, SCA-H attempts to solve for one unknown variable (soil moisture) from one equation that relates the horizontally polarized TB to soil moisture. Vegetation information is provided by a 13-year climatological data base of global NDVI and a table of parameters based on land cover.

Dual Channel Algorithm (DCA)

In the DCA, both the vertically and horizontally polarized TB observations are used to solve for soil moisture and vegetation optical depth. The algorithm iteratively minimizes a cost function (Φ2) that consists of the sum of squares of the differences between the observed TBs (TB^obs) and estimated TBs (TB^est):

   (Equation 1)

In each iteration step, the soil moisture and vegetation opacity are adjusted simultaneously until the cost function attains a minimum in a least square sense. Similar to SCA-V and SCA-H, ancillary information such as effective soil temperature, surface roughness, and vegetation single scattering albedo must be known a priori before the inversion process. DCA permits polarization dependence of coefficients in the forward modeling of TB observations. As currently implemented for SPL2SMP_E, the H and V coefficients are set the same.

Microwave Polarization Ratio Algorithm (MPRA)

The MPRA is based on the Land Parameter Retrieval Model (Owe 2015) and was first applied to multi-frequency satellites such as AMSR-E. Like DCA, MPRA attempts to solve for soil moisture and vegetation optical depth using the vertically and horizontally polarized TB observations. However, it does so under the assumptions that (1) the soil and canopy temperatures are considered equal, and (2) vegetation transmissivity and the single-scattering albedo are the same for both H and V polarizations. When these assumptions are satisfied, it can be shown that the soil moisture and vegetation optical depth can be solved analytically in closed form, resulting in the same solutions as would be obtained iteratively using DCA. Similarly to DCA, ancillary information such as effective soil temperature, surface roughness, and vegetation single scattering albedo must be known a priori before the inversion process.

Extended Dual Channel Algorithm (E-DCA)

The E-DCA is a variant of DCA. Like DCA, E-DCA uses both the vertically and horizontally polarized TB observations to solve for soil moisture and vegetation optical depth. In E-DCA, however, the cost function (Φ2) is formulated in a way different from that of DCA. Instead of minimizing the sum of squares of the differences between the observed and estimated TBs as in DCA (see Equation 1), the E-DCA attempts to minimize the sum of squares of the difference between the observed and estimated normalized polarization differences (expressed in natural logarithm) and the difference between the observed and estimated TBs (also expressed in natural logarithm) as follows:

   (Equation 2)

In each iteration step, soil moisture and vegetation opacity are adjusted simultaneously until the cost function attains a minimum in a least square sense. It is clear that when both Φ_DCA^2 and Φ_(E-DCA)^2 attain their theoretical minimum value (i.e. zero) in the absence of uncertainties of modeling, observations, and ancillary data, T_(B,v)^obs = T_(B,v)^est and T_(B,h)^obs = T_(B,h)^est, implying that DCA and E-DCA converge to the same solutions. The advantage of E-DCA over DCA, however, is apparent when in reality there is finite uncertainty (e.g., a dry bias associated with the ancillary soil temperature data); this uncertainty will be cancelled from the numerator and denominator in the calculation of the normalized polarization difference in Φ_(E-DCA)^2, leaving such uncertainty affecting only one component of the cost function instead of two components as in Φ_DCA^2. This reduction in the impact of soil temperature uncertainty on soil moisture retrieval should make E-DCA more tolerant of soil temperature uncertainty, resulting in fewer instances of retrieval failure than DCA. 

Anthropogenic RFI, principally from ground-based surveillance radars, can contaminate both radar and radiometer measurements at L-band. The SMAP radiometer electronics and algorithms include design features to mitigate the effects of RFI. The SMAP radiometer implements a combination of time and frequency diversity, kurtosis detection, and use of thresholds to detect and, where possible, mitigate RFI.

Level-2 radiometer data can also contain bit errors caused by noise in communication links and memory storage devices. Consultative Committee on Space Data Systems (CCSDS) packets include error-detecting Cyclic Redundancy Checks (CRCs), which are used to flag errors.

More information about error sources is provided in Section 4.6: Algorithm Error Performance of the ATBD. (O'Neill et al. 2018)