Fig. S3. Comparison of modelled against observed sea surface elevation anomaly for Models 1 to 5. The T/P altimeter data are shown as crosses. The T/P track follows that of Jason-1 about 6.5 min later and 1.3° to the west. It should be noted that T/P has acquisition problems making it harder to lock onto the sea level reflection. Hence it misses a large amount of data from this track, as well as for earlier and later cycles. In addition, if one wants to calculate the difference between this track and one 10 days earlier, as carried out for Jason-1, then too much of the signal is lost. Therefore here we simply present the data from Track 452. Gower (in press) provides a detailed analysis of T/P. In order to calculate a reference signal, he uses an average of data from tracks both 10 days before and 10 days after. Importantly, the signal in the south-west in Track 452 (with an amplitude of  0.6 m) is similar to that found by subtracting the smoothed average of Tracks 451 and 453 (Gower, in press). However, the signal in the north-east is influenced more by the reference signal. Therefore we only discuss the leading wave in the south-east. It is a robust feature that shows the leading wave as one broad peak instead of the 2 peaks captured by Jason-1. A notable feature of the comparison against the T/P data, appears to be that Models 2 and 5 have the correct phase, while Model 3 has a phase lag.