CORRELATED ENVIRONMENTAL CORRECTIONS IN TOPEX POSEIDON, WITH A NOTE ON IONOSPHERIC ACCURACY/

Estimates of the effectiveness of an altimetric correction, and interp retation of sea level variability as a response to atmospheric forcing , both depend upon assuming that residual errors in altimetric correct ions are uncorrelated among themselves and with residual sea level, or knowing the correlations. Not surprisingly, many corrections are high ly correlated since they involve atmospheric properties and the ocean surface's response to them. The full corrections (including their geog raphically varying time mean values), show correlations between electr omagnetic bias (mostly the height of wind waves) and either atmospheri c pressure or water vapor of -40%, and between atmospheric pressure an d water vapor of 28%. In the more commonly used collinear differences (after removal of the geographically varying time mean), atmospheric p ressure and wave height show a -30% correlation, atmospheric pressure and water vapor a -10% correlation, both pressure and water vapor a 7% correlation with residual sea level, and a bit surprisingly, ionosphe ric electron content and wave height a 15% correlation. Only the ocean tide is totally uncorrelated with other corrections or residual sea l evel. The effectiveness of three ionospheric corrections (TOPEX dual-f requency, a smoothed version of the TOPEX dual-frequency, and Doppler orbitography and radiopositioning integrated by satellite (DORIS) is a lso evaluated in terms of their reduction in variance of residual sea level. Smooth (90-200 km along-track) versions of the dual-frequency a ltimeter ionosphere perform best both globally and within 20 degrees i n latitude from the equator. The noise variance in the 1/s TOPEX ionos pheric samples is similar to(11 mm)(2), about the same as noise in the DORIS-based correction; however, the latter has its error over scales of order 10(3) km. Within 20 degrees of the equator, the DORIS-based correction adds (14 mm)(2) to the residual sea level variance.