Optimal methods for gridding 3D velocity-data?

Hi guys

I am looking for an efficient way to grid a number of 3D velocity data.

Specifically I have on the order of 1500 points for which coordinates (X,Y,Z) and velocities (U,V,W) are known. Moreover I know that all coordinates exist within a matrix, A, of size size(A)=(800,800,180) and that the top and bottom surface have velocities equal to zero. That is U(Z=1)=0 and U(Z=180)=0 (likewise for V and W).

I would like to grid these data such that the entire matrix A is filled with extrapolated/interpolated data. I have been medling around with griddata and TriScatteredInterp but have found that they require quite substantial memory and calculating time. That is if they even finish the task.

Any suggestions?