2.3. Regional sorting
In order to investigate north–south variation, we isolated 192 dispersion curveswith ray paths approximately parallel to the orogen, and divided them into ten sets corresponding to separate regions (Fig. 1b).Within each regional set, we discarded any dispersion curves that deviated fromthemean of the set bymore thantwo standard deviations,thus removing potentially spurious data (the necessarily subjective nature of manually picking dispersion curves is likely to result in some inaccuracies). This process removed 52 outlying curves, leaving 140 consistent dispersion curves. The number of accepted curves per set varied between 11 and 20. The 140 dispersion curves are shown in Fig. 2a (each colored according to its region in Fig.1b).Within each setwe took the mean group velocity at each period for which there were three or more data points, yielding an average dispersion curve for the set (Fig. 2b), and hence for the corresponding region (Fig. 1b).