Many rainfall-run-off distributed models need rainfall data as input on a pixel by pixel basis, for each time interval. Due to the large amount of pixels that can make up a basin (proportional to the map scale), a fast and efficient method must be devised in order to obtain the rainfall field for each time interval (e.g. 20 minutes). Most models use interpolation methods such as the Inverse Distance Weighted. However, we propose the use of a Delaunay Triangulation using the incremental algorithm developed by Watson. The rainfall stations are used as the vertices of the triangles. They represent a three dimensional plane of the rainfall. Once the equation of the plane is known, a rainfall value for each pixel is calculated. We compare both methods and evaluate the sensitivity of the latter.