Can SMOS Data be Used Directly on the 15-km Discrete Global Grid?

Radiometric observations from the Soil Moisture and Ocean Salinity (SMOS) mission are processed to Level 1 brightness temperature (Tb) with ~ 42-km spatial resolution and reported on a 15-km hexagonal Discrete Global Grid (DGG). While these data should be used at the 42-km resolution which the oversampled DGG represents, this paper poses the question of whether they can be used directly at 15-km resolution without undertaking downscaling or implementing multiscale-type procedures when used in data assimilation. To assess the error associated with using the 42-km SMOS Tb data at 15-km resolution, this study employs 1-km Tb data from the Australian Airborne Cal/Val Experiment for SMOS (AACES). The study compares SMOS-like data derived from AACES at 42-km resolution with Tb values actually observed on the 15-km DGG. These 15-km DGG data are subsequently interpolated to a regular 12-km model grid and compared with actual observations at that resolution. The results show that the average root mean square differences in Tb between the 15- and 42-km footprints are 4.5 K and 3.9 K for horizontal (H) and vertical (V) polarizations, respectively, with a maximum difference of 12.9 K. The errors when interpolating the 42-km data onto the 12-km model grid were estimated to be 3.3 K for H polarization and 2.9 K for V polarization under the assumption of independence or 4.5 K and 3.9 K for H and V polarizations, respectively, with 4.0 K in H polarization and 3.6 K in V polarization from the 15- to 12-km interpolation process alone. An evaluation of the Tb differences for 42-km data assumed on the 15-km DGG found no correlation with vegetation based on leaf area index and only slight correlation with the spatial variance of SMOS data and topographic roughness. Given these differences and the noise that currently exists in SMOS Tb at 42 km, the 15-km DGG data can be used directly on the hexagonal grid or interpolated onto a regular grid of equivalent spatial resolution without further degrading the data quality.