SMOS L2 OS ATBD - ARGANS

47
ICM-CSIC
LOCEAN/SA/CETP
IFREMER
SMOS L2 OS
Algorithm Theoretical
Baseline Document
Doc: SO-TN-ARG-GS-0007
Issue: 3 Rev: 9
Date: 25 January 2013
Page: 47
when compared to exact numerical simulation using the Method of Moment with a general
agreement between SSA-1 and MoM within 3dB in VV and within 1.5 dB in HH
polarizations [12]. In average, SSA-1 overestimate HH and underestimates VV so that SSA-1
systematically overestimates the H/V ratio with a mean of order +20%. The errors on the sea
surface roughness statistics are difficult to estimate but will clearly have an important impact
as well.
A complete error budget estimate can not be provided without any estimate of the error on
the auxiliary sun brightness temperature data at 1.4 GHz. If it comes out of L1 processor, we
need an error budget on the estimate of that parameter from L1.
3.6.4. Practical considerations
3.6.4.1. Calibration and validation
Dedicated CAL/VAL activities should be envisaged for the SMOS sunglint model with two
main components:
-an earth-based campaign aiming at measuring precisely the sunglint scattering at L-band
(e.g., experiment similar to [4]), with high-quality concomitant auxiliary solar fluxes
measurements at 21 cm as well as surface roughness information to calibrate and validate the
bistatic-scattering coefficient models.
-a SMOS-data based analysis. Re-analysis of all flagged pixels and brightnesses for which
good quality (close in time and space) co-localized auxiliary wind and solar flux data at 21
cm are available shall be performed to assess the efficiency of the model.
3.6.4.2. Quality control and diagnostics
Assuming the major source of error in the model shall be the estimation of the sun brightness
temperature at 1.4 GHz, quality control and diagnostics will strongly depend on the accuracy
for that auxiliary data.
If it comes out of L1 processor (without a priori geophysical input), a complementary quality
check shall be performed for that auxiliary data using earth-based solar flux measurements
available at 1.4 GHz. These are available from sun-tracker radiometers by the US Air Force,
at Sagamore Hill (Massachusetts), since 1966. They can be obtained through the National
Geophysical Data Center at Boulder, Colorado. These data sets also include other solar fluxes
measurements conducted at 1415 MHz since 1988 from radiometers in Palehua (Hawaii), San
Vito (Italy) and Learmonth (Australia), and 1GHz data are also collected daily at Nobeyama
Radio Observatory (Japan). If high temporal resolution solar fluxes can be obtained, the
closest data in time from SMOS acquisitions shall be used to monitor quality controls, as sun
brightness temperature values might evolve very significantly over short time scales. The socalled
R-components of the sun brightness temperature indeed consist of the second and
minute-duration bursts produced by the active sun components: sunspots (manifestations of
magnetically disturbed conditions at the sun's visible surface), flares (huge explosions on the
surface of the sun) and other transient activity. This high-temporal varaibility of the sun
signals might strongly affect the quality of the forward model estimates.