Governmental and environmental agencies have recognized that data Warehouses (DWs) oriented to the environmental domain can have a deep impact on the ability to preserve the environment. However, extracting environmental data from external sources in order to load them into the data organization area of a DW is much more complex when compared with conventional DWs. In environmental DWs, extraction is frequently based on autonomous, heterogeneous external sources distributed over the Web. In order to improve this process, we propose an architecture that combines a middleware system and DW technology. The use of this architecture has been achieved through the implementation of a tool called LSExtract, which has been built on top of a middleware system called Le Select. Its goal is to locate and extract relevant and accurate information from heterogeneous data sources over the Web, extract data published by Le Select, and load it, after some transformations, into the data organization area of an environmental DW.