Abstract: Spatial distribution of pan evaporation(E_E) depending on multiple factors such as radiation,（vapor） pressure and wind is a crucial parameter for geographic models and climatic models,and also a key to（environmental） management.However the weather observation stations normally distribute sparsely and irregularly on the ground.The techniques to interpolate available point data to estimate the value of any location is demanded.This paper introduces the process to interpolate monthly E_E data of 53 stations for the period of 1980 to 2000 to create 252 E_E surfaces in Coarse Sandy Hilly Catchments of Loess Plateau.Quinat-variate partial thin plate spline models were developed in which vapor pressure deficit（P_d）,net radiation（P_n） and wind speed（wind） were used as three covariate sub-models.GCV statistics indicate that the resultant evaporation surfaces hold high fidelity and smoothness.The lapse rates（evaporation changing rate with each covariate） showed that: the dependence of E_E on Rn has no strong seasonal trend although at the equinoxes it is slightly big,whereas the dependence of E_E on pd has an obvious trend with strongest influence in summer,and wind speed has more relative influence in winter.