In *this note*, we show that the percentage of area when the relative vorticity ζ changes sign increases from 30% at midlatitudes to 50% in the tropics and subtropics. I suggested that this could suggest that the model of Hoskins and Bretherton (1972; hereafter HB72) could be relevant in the tropics and subtropics. But Dongxiao pointed out another interpretation: in the tropics and subtropics, ζ is not specifically controlled by either eSQG or HB72 and there is at much chance to have ζ changing sign between the surface and depth than not.

It was suggested to calculate the spatial correlation between ζ from the model and from eSQG for the points where ζ changes or not sign between the surface and depth. The correlation has been calculated for a series of 5° by 5° domain located near 140°W. The 2004 mean correlation is plotted in Fig. 1. In (a), all the domain is considered, in (b) only the points where ζ has the same sign at the surface (0-55 m) than at depth (150-320 m) are taken and in (c) only the points where ζ changes sign between the surface and depth are taken.

In (b), the correlation is improved compared to (a) between 10° and 20° from the equator while, not surprisingly, the correlation is almost everywhere negative between 150 and 320 m in (c).

Does this help us or not to understand if indeed either model is more important or not for the tropics and subtropics?