The only builtin non-linear scale is log, but you can roll your own. Here's a simple example:
%%Start with peaks
[x,y,z] = peaks;
%%Two linear equations meeting at y=cutoff
cutoff = 2;
scale = 5;
%%Mask of all Y values above cutoff
above = y>cutoff;
%%Compute scaled Y values
y2 = y;
y2(above) = cutoff + (y(above)-cutoff) * scale;
%%Create contour with y2
contourf(x,y2,z)
%%Fix up tick labels
ax = gca;
yt = ax.YTick;
above = yt>cutoff;
yt(above) = cutoff+(yt(above)-cutoff)/scale;
ax.YTickLabel = yt;
This uses two scales and abruptly switches between them at the cutoff. I've never been very fond of this approach. I think that it's better to use a function with a smooth transition. The basic idea is the same though. You create a scaled copy of your Y coordinates and create the contour with that. Then you fix-up the YTickLabel to be the inverse.
Does that make sense?
