Interannual Sea Level Variations in the Tropical Indian Ocean from Geosat and Shallow Water Simulations

Sea Level variations of the Indian Ocean north of 20°S are analyzed from Geosat satellite altimeter data over April 1985–September 1989. These variations are compared and interpreted with numerical simulations derived from a reduced gravity model forced by FSU observed winds over the same period. After decomposition into complex empirical orthogonal function the low-frequency anomalies are described by the first two modes for observations as well as for simulations. The sums of the two modes contain 34% and 40% of the observed and simulated variances respectively. Averaged over the basin, the observed and simulated sea level changes are correlated by 0.92 over 1985–1988. The strongest change happens during the El Niño 1986–1987, between winter 1986 and summer 1987 the basin-averaged sea level rises by ∼1 cm. These low-frequency variations can partly be explained by changes in the Sverdrup circulation. The southern tropical Indian Ocean between 10° and 20°S is the domain where those changes are strongest: the averaged sea level rises by ∼4.5 cm between winter 1986 and winter 1987. There, the signal propagates southwestward across the basin at a speed similar to free Rossby waves. Sensitivity of observed anomalies is examined over 1987–1988, with different orbit ephemeris, tropospheric corrections, and error reduction processes. The uncertainty of the basin-averaged sea level estimates is mostly due to the way the orbit error is reduced and reaches ∼1 cm. Nonetheless, spatial correlation is good between the various observations and better than between observations and simulations. Sensitivity of simulated anomalies to the wind uncertainty, examined with FSU and ECMWF forcings over 1985–1988, shows that the variance of the simulations driven by ECMWF is 52% smaller, as FSU winds are stronger than ECMWF. Results show that the wind strength also affects the dynamic response of the ocean: anomalies propagate westward across the basin more than twice as fast with FSU than with ECMWF. It is found that the discrepancy is larger between ECMWF and FSU simulations than between observations and FSU simulations.