| Abstract: |
SMOS and Aquarius demonstrated remote sensing of Sea Surface Salinity (SSS) from space by means of L-band radiometry, and the approach developed for Aquarius is now being applied to the L-band radiometer on SMAP, whose primary objective is retrieval of soil moisture. We report on the algorithm that we have developed as a research tool to improve the retrieval technique and the radiative transfer model. An important element of the algorithms developed for SMOS, Aquarius and SMAP is empirical adjustment in order to deliver optimum accuracy on the retrieved SSS. An example of empirical approaches is the Geophysical Model Function (GMF) that quantifies the increase in brightness temperature due to the surface roughness. The GMF is a polynomial fit of TB as a function of wind speed and direction, derived from a large sample of TB observed by the sensor. As such, implementations of the GMF tend to be sensor dependent, and can include effects not necessarily related to roughness, but correlated with it to some extent (e.g., foam). Other examples of adjustments apply to the sea water dielectric constant, atmospheric absorption and galactic emission reflected by the ocean surface. However, some of these corrections are not completely independent from each other. In particular, if either the rough surface emission or the reflected galaxy are adjusted, they can be inconsistent with each other.
While empirical adjustments are a desirable step to provide the best geophysical product to the science community, they do not always help improve our understanding of the processes at work, and can lead to repeated re-adjustments because the parameters are not as independent as they should be. We have developed an algorithm for a SSS product for SMAP using theoretical models with minimal number of empirical corrections. While it is not realistic to avoid any form of empirical parameterization, our objective is to focus on an assessment of the performances and limitations of the current semi-theoretical parameterizations. For example, instead of using a GMF for the roughness effects, we employ a two-scale emissivity model with a sea spectrum model and a foam layer. This model had limited success in the past (hence the frequent use of GMF) but has been recently refined using SMOS observations. While this constitutes an empirical adjustment, the model still uses physical principles to describe the surface (e.g. waves and foam layer) and should be applicable to another L-band sensor such as SMAP. Discrepancies between the model results and SMAP observations open the door to the investigation of the impact of the beam width and reflected atmospheric signal on the observed TB dependence to roughness (such effects are expected to be incorporated in a GMF). Another example concerns the sea water dielectric constant. Aquarius and SMAP currently use a model derived from radiometric measurements. We use and the algorithm to assess the performance of the models that are derived from laboratory experiments. Finally, our reflected galaxy model uses radio-astronomy surveys for the emission, and theoretical surface reflection models that are consistent with the ocean surface emission component.
We will present the details of our model and examples of retrieved salinity, and discuss its performance for the retrieval of SSS.
|
