3 SMOS and Aquarius/SAC-D Missions
45
more successive snapshots and under more different angles. A maximum of above
60 horizontal and 60 vertical T B measurements is obtained in the centre, decreasing
to half of it at 300 km to both sides. The retrieval algorithm performs a minimization
loop using a cost function where the recorded T B is compared to a T B modeled value
for each one of the available angular measurements, until an optimal fit is reached.
The forward model, or geophysical model function (GMF), that provides the T B
values corresponding to specific seawater characteristics and viewing geometry is a
key component of the retrieval algorithm. It has to simulate the emission from the
top ocean layer, plus any other radiation at the same frequency coming from external sources (e.g. the cosmic background) and scattered on the ocean surface to the
concerned direction, and finally the transformation the overall radiation leaving the
surface suffers until reaching the SMOS antenna plane (from atmosphere attenuation and upward emission until Faraday polarization rotation in the ionosphere). The
emissivity of a flat sea as function of temperature, salinity, viewing angle, frequency
and polarization is quite well modeled using the geometric optics theory (Klein and
Swift, 1977), but the different processes that impact on the L-band emission of a
roughened surface were not fully described in the several theoretical formulations
available at the moment of starting the development of SMOS algorithms. It was
necessary to design several new components of the GMF for the SMOS Level 2
Ocean Salinity Processor (L2OP, Zine et al., 2008).
The effect of surface roughness on the T B is the main geophysical source of
error. Unlike Aquarius, SMOS does not have any means to acquire simultaneous
independent information of this roughness to be used in the GMF. In addition to
this, the available data reporting rough sea surface emissivity dependencies with
wind speed does not allow to discriminate the best adapted formulation among the
several theoretical models proposed (Font et al., 2006). The SMOS L2OP implements the approach of the polarized ocean T B being the addition of two terms,
one corresponding to the flat sea emission and the other one a correction to it due
to the impact of the surface roughness. For this correction three different options
were considered, to be tested, improved or even discarded once SMOS data are
available. Two of them are theoretical formulations (statistical description of the
sea surface plus electromagnetic scattering model) based on the two-scale model
(Dinnat et al., 2002) and the small slope approximation (Johnson and Zhang, 1999),
and the third one (Gabarró et al., 2004) is an experimental fit, using different roughness descriptors, from data acquired during the WISE trials carried out as part of the
SMOS science definition studies (Camps et al., 2004). All these roughness models
require the use of external information on wind speed, significant wave height, wave
age, etc. to describe the sea state. They are provided to the L2OP by global operational forecasts (atmospheric and ocean wave models) from the European Centre
for Medium range Weather Forecasts (ECMWF) that also deliver other parameters,
as sea surface temperature, needed by different modules of the retrieval algorithm.
These ECMWF variables are introduced as first guess values in the cost function,
and during the minimization process they are also tuned like SSS, initially obtained
from climatology, until reaching the optimum fit between modeled and measured
T B . This multi-parameter convergence is possible thanks to the over-determination
45
more successive snapshots and under more different angles. A maximum of above
60 horizontal and 60 vertical T B measurements is obtained in the centre, decreasing
to half of it at 300 km to both sides. The retrieval algorithm performs a minimization
loop using a cost function where the recorded T B is compared to a T B modeled value
for each one of the available angular measurements, until an optimal fit is reached.
The forward model, or geophysical model function (GMF), that provides the T B
values corresponding to specific seawater characteristics and viewing geometry is a
key component of the retrieval algorithm. It has to simulate the emission from the
top ocean layer, plus any other radiation at the same frequency coming from external sources (e.g. the cosmic background) and scattered on the ocean surface to the
concerned direction, and finally the transformation the overall radiation leaving the
surface suffers until reaching the SMOS antenna plane (from atmosphere attenuation and upward emission until Faraday polarization rotation in the ionosphere). The
emissivity of a flat sea as function of temperature, salinity, viewing angle, frequency
and polarization is quite well modeled using the geometric optics theory (Klein and
Swift, 1977), but the different processes that impact on the L-band emission of a
roughened surface were not fully described in the several theoretical formulations
available at the moment of starting the development of SMOS algorithms. It was
necessary to design several new components of the GMF for the SMOS Level 2
Ocean Salinity Processor (L2OP, Zine et al., 2008).
The effect of surface roughness on the T B is the main geophysical source of
error. Unlike Aquarius, SMOS does not have any means to acquire simultaneous
independent information of this roughness to be used in the GMF. In addition to
this, the available data reporting rough sea surface emissivity dependencies with
wind speed does not allow to discriminate the best adapted formulation among the
several theoretical models proposed (Font et al., 2006). The SMOS L2OP implements the approach of the polarized ocean T B being the addition of two terms,
one corresponding to the flat sea emission and the other one a correction to it due
to the impact of the surface roughness. For this correction three different options
were considered, to be tested, improved or even discarded once SMOS data are
available. Two of them are theoretical formulations (statistical description of the
sea surface plus electromagnetic scattering model) based on the two-scale model
(Dinnat et al., 2002) and the small slope approximation (Johnson and Zhang, 1999),
and the third one (Gabarró et al., 2004) is an experimental fit, using different roughness descriptors, from data acquired during the WISE trials carried out as part of the
SMOS science definition studies (Camps et al., 2004). All these roughness models
require the use of external information on wind speed, significant wave height, wave
age, etc. to describe the sea state. They are provided to the L2OP by global operational forecasts (atmospheric and ocean wave models) from the European Centre
for Medium range Weather Forecasts (ECMWF) that also deliver other parameters,
as sea surface temperature, needed by different modules of the retrieval algorithm.
These ECMWF variables are introduced as first guess values in the cost function,
and during the minimization process they are also tuned like SSS, initially obtained
from climatology, until reaching the optimum fit between modeled and measured
T B . This multi-parameter convergence is possible thanks to the over-determination