Distributed hydrological model can be divided into two parts, as runoff and the convergence. Runoff calculation is the basis of distributed hydrological model, its results will determine the accuracy of the simulation results directly. Rainfall is an important input of runoff calculation, thus its accuracy has special significance to runoff calculation. For most small watershed, now mainly rely on rainfall stations to obtain rainfall information. In this case, the most effective approach is base on space correlativity principle, use the interpolation algorithm to obtain distributed rainfall data. Take XueYe reservoir 2000-2008 year of reality measures data as an example, we compared the apply condition of several kinds algorithm such as co-kriging interpolation, kriging interpolation, reverse-weighted interpolation, and so on, prove that co-kriging interpolation is most fit XueYe reservoir .