Understanding SMOS Satellite L-Band Data over Antartica Using In Situ Measurements And Radiative Transfer Modeling
Leduc-Leballeur, Marion¹; Picard, Ghislain¹; Mialon, Arnaud²; Lefebvre, Eric¹; Rudiger, Christoph³; Kerr, Yann²; Dupont, Florent⁴; Arnaud, Laurent¹; Fily, Michel^1
1LGGE, FRANCE; ²CESBIO, FRANCE; ³Department of Civil Engineering, Monash University, AUSTRALIA; ⁴LGGE / CARTEL, FRANCE

In Antarctica, remote sensing data are particularly important for the study of regional climate due to the sparse coverage of the in situ measurements (Turner et al., 2006). The observations of space-borne sensors are a unique mean to provide a detailed picture of the spatial and temporal variations of important climate variables. In particular, a dataset of well-calibrated monthly surface temperature, at the typical resolution of tens of kilometers and spanning several decades, would be highly valuable to science community.
At present, the microwave radiometers on-board SMMR, SSM/I and AMSR-E satellites (high-frequency radiometers) offer a very long record of observations sensitive to surface temperature irrespective of the cloudiness conditions. However, successful retrieval of the temperature over the snow fields from microwave data is still difficult because this requires an accurate prior of the snow emissivity. For example, retrieving temperatures with accuracy better than 2 K requires to know the emissivity with uncertainty lower than 0.01. This is a challenge, as the emissivity varies spatially from 0.65 to 0.9 across the continent and these variations are not well understood. A possible option to solve this issue is to obtain an accurate reference temperature, then infer precise emissivities at high frequencies, and produce temperature data from brightness temperature.
In this context, the MIRAS radiometer on board ESA's SMOS satellite (Kerr et al., 2001), with its low observation frequency (1.4 GHz), offers a great opportunity to estimate this accurate reference temperature. Indeed, as a result of the high volume scattering of the snow field produced by the snow grains at the higher frequencies, the snow emissivity is much less than unity. On the other hand, the scattering is reduced at lower frequencies such as L-band and the emissivity is close to unity when the surface reflection effect vanishes at V-polarization near the Brewster angle of ~50°. As a consequence, the spatial variation of the emissivity at L-band should be smaller and more predictable than at higher frequencies.
In this study, to explore those characteristics of the L-band emissivity, we jointly use the SMOS L3 daily global polarised brightness temperature product, a snow passive microwave radiative transfer model (DMRT-ML ; Picard et al., 2012) and the in-situ measurements from two ice cores up to a depth of 80m recently performed at Dome C during the austral summer campaign of 2012-13. Moreover, the estimation of the emissivity from brightness temperature requires highly accurate surface temperature, which is not available as stated in the introduction. However, for the purpose of the present analysis, we have chosen to use the surface temperature predicted by the ECMWF ERA-Interim reanalysis. The interpretation of the results must account for this point. The approach to derive emissivity is similar to the one presented in Picard et al. (2009). No atmospheric corrections were considered, as the Antarctic atmosphere is cold, dry and shallow due to the high elevation of the continent and the low altitude of the tropopause. Similarly, Faraday rotation is also negligible at Dome C.
The first comparisons show that the emissivity at L-band is mainly between 0.95 and 0.97. However, the analysis data are not perfect, in particular, the well-known warm-bias of a few Kelvin in the Antarctic region may affect the ERA-Interim reanalysis. Thus, the emissivity may be slightly higher. This is confirmed at Dome C (75°S-123°E) where the snow temperature is measured continually. However, it becomes evident from those data that the SMOS emissivity at V-polarization and near the Brewster angle is not exactly at 1. In order to explore the SMOS TB at V-polarization and near the Brewster angle, it is proposed to use a radiative transfer model. Simulations are realized with snow density, grain size and temperature profiles measured at Dome C up to a depth of 80 meters. Results show that the brightness temperature is over-estimated at both polarizations and any incidence angle. Moreover, the difference between V and H -polarizations at high incidence angle is underestimated. The sensitivity tests suggest the importance of the density profile and the stratification to a better understanding of the V and H-polarized brightness temperature at L-band.

Kerr Y. H., Waldteufel P., Wigneron J. -P., Martinuzzi J., Font J., Berger M., "Soil moisture retrieval from space: the Soil Moisture and Ocean Salinity (SMOS) mission," IEEE Transactions on Geoscience and Remote Sensing, vol. 39, no. 8, pp. 1729-1735, 2001.
Picard G., L. Brucker, A. Roy, F. Dupont, M. Fily, and A. Royer, "Simulation of the microwave emission of multi-layered snowpacks using the dense media radiative transfer theory: the DMRT-ML model", Geoscientific Model Development Discussion, 5, 3647-3694, 2012.
Picard, G. Brucker, L., Fily, M. , Gallée, H., Krinner, G., "Modeling timeseries of microwave brightness temperature in Antarctica", Journal of Glaciology, 55 (191), pp 537-551, 2009.
Turner J, Lachlan-Cope TA, Colwell S, Marshall GJ, Connolley WM, "Significant warming of the Antarctic winter troposphere". Science, 311, pp 1914-1917, 2006.