An approach to land zoning for fire planning purposes through classification of geographical units referred to as Basic Units - in a multivariate context is proposed. The method, developed for large areas, employs a few statistics computed from historical fire data, that were selected with the aim of depicting a ''fire history profile'' of each Basic Unit. The statistics were chosen in order to describe different aspects, such as fire frequency, fire continuity, average and maximum area burned and average area spread rate of fires. Using the computed statistics as variables, the Basic Units can be aggregated with cluster analysis techniques and classes can be defined, each class of Basic Units representing a land zone with a specific mean fire history profile. Fire management decisions can be tailored according to the properties of clusters. The method was successfully applied for supporting fire management plans in two Regions of Northern Italy, where the Basic Units were chosen with an administrative criterion.