The equatorial sea level analysis of the National Centers for Environmental
Predictions deviates by as much as 8 cm from independent TOPEX/Poseidon (T
/P) observations. This may be due to the model's underestimation of salinit
y variability. Therefore, methods are developed to improve the model's sali
nity field through T/P data assimilation and use of sea surface salinity (S
SS) observations.
In regions where temperature is well known, salinity estimates are made wit
h the use of climatological temperature-salinity (T-S) correlations. These
estimates are improved by combining T-S with SSS observations and corrected
with dynamic height, which provides information on salinity variability. T
ests with independent conductivity temperature depth data show that the com
bination of TS with SSS significantly improves salinity estimates. In the w
estern Pacific, the maximum root-mean-square (rms) estimation error of 0.55
psu is reduced to 0.42 psu by the use of SSS in the salinity estimate. Cor
rection with dynamic height reduces this rms to 0.22 psu. Also in other par
ts of the tropical Pacific Ocean the salinity estimation errors are reduced
by a factor of 2 by combination of the T-S estimate with SSS and dynamic h
eight. This study provides the first step toward an assimilation scheme in
which salinity is corrected with the use of T/P sea level observations.