The measurement of Class A pan evaporation is one of the more widely used methods to estimate evapotranspiration. Unfortunately pan evaporation measurement is often difficult due to instrumental limits and practical problems, therefore it would be helpful to be able to estimate this quantity. In this paper a method to estimate evaporation, based on artificial neural network model, is proposed with the purpose of replacing the evaporimeter. Neural networks are elaboration systems composed of many processing units, connected by communication channels, which carry numeric data and store knowledge. In a learning process knowledge is acquired giving data matrixes composed of both input parameters and relative, measured, output. The neural network, selected for this investigation, has a three layer feed-forward architecture with Quickprop learning algorithm, a variation of Backpropagation algorithm. Preliminarily, as inputs were tested the meteorological parameters affecting evaporation process; the input parameters which perform the best results were: wind speed, dry bulb temperature, wet bulb temperature, solar radiation. Output of the neural network is the estimated pan evaporation. Experimental results show the capabilities of neuronal techniques in evaporation studies: the neural network calculates daily pan evaporation with a good correlation with measured values (r=0.82) and a low standard error of estimate (SEE=0.46 mm/d). The proposed method can be applied to verification of historical evaporation data and can allow the class A pan evaporimeter to be eliminated and thus eliminate maintenance problems as well.