The stratification of rainfall fields to improve specific rainfall models is a subject that has received relatively little attention in the literature. It is shown here that objective classification techniques, based on the intensities and spatial distribution of the rainfall fields, can produce meaningful results in terms of the area threshold method (ATM) model and climatology. Four approaches for rainfall classification, using rain gauge data, are proposed in order to improve the average areal rainfall estimation in the Sahel by the ATM model. Two of them are based on the structural behavior of the rainy area (area where it rains above a given threshold) function against a threshold. Based on this function, a new parameter, called the under profile area (UPA), has been proposed for the classification of rainfall fields. The groups obtained from the method based on this parameter are characterized by different average spatial structures. A significant improvement on the ATM model is observed by considering classification based on the UPA parameter. An average reduction of 34% of the root-mean-square error is observed in a validation term. This improvement is a direct consequence of the fact that the optimal thresholds are different from one group to another, which is an important point when considering the impact of classification on the ATM model. ¿ American Geophysical Union 1996