People use data warehouses to help them make decisions. For example, public policy decision-makers can improve their decisions by using this technology to analyze the environmental effects of human activity. In production systems, data warehouses provide structures for extracting the knowledge required to optimize systems. Designing data warehouses is a complex task; designers need flexible and precise methods to help them create data warehouses and adapt their analysis criteria to developments in the decision-making process. In this paper, we introduce a flexible method based on UML (Unified Modeling Language). We introduce a UML profile for building multi-dimensional models and for choosing different criteria according to analysis requirements. This profile makes it possible to specify integrity constraints in OCL (Object Constraint Language). We apply our method to the construction of an environmental system for analyzing the use of certain agricultural fertilizers. We integrate various data sources into a multi-dimensional model showing several categories of analysis, and the consistency of data can be checked with OCL constraints.