Land-based monitoring of large areas requires optimized reduction of the surface under investigation and the parameters measured. In this study, the sampling universe consists of patches of vegetation scattered over a wide area. A step-wise procedure is proposed to monitor pattern. At first, the patches are recognized, mapped and monitored by remote sensing. Then, the detailed vegetation composition must be investigated in the field by stratified sampling who's design represents yet another step. The monitoring objectives are reduced to (1) the recognition of overall changes within individual patches and (2) the observation of changes within the types, considering the entire area. In order to reduce the number of plots while maximizing the overall variation in the yearly investigations, an adapted version of ranking is proposed for the set of sampling units of each patch, based on orthogonal components (Orlóci 1973). An example is presented using published data of the area.