No. Not really. You can never know the true, underlying functional form for any list of scattered points, since there are infintiely many surfaces that will pass through any set of points. You can try using things like polynomial fits, which sadly, tend not to do very well. Polynomials are great tools in theory. and for low order approximations. But to generate a fully general surface model over a wide enough region, they tend to be poor. They are frequently used due to their simplcity. Everyone knows what a Taylor series, is, so they should be good, right? They are over-abused far more often than well used. (And I'm the person who posted polyfitn on the file exchange.)
You don't suppy the data, so I can go into too much depth here. It is scattered, but seemingly not very noisy, so an option is a tool like scatteredInterpolant or griddata.
Or, if you want a gridded surface, that can then be evaluated using interpolation like interp2, or plotted using contour or surf, use my gridfit, found on the file exchange.
It won't give you a clean functional form you can write down on paper. As I said, nothing will do that. But it will extrapolate moderately well into the corners, and you can use it to predict any point, using interp2.
As I said, lacking your data, I cannot do much more here. Post it in a comment if you want me to look at it.
