This paper reports on a study which combines corpus-based and genre based approaches to the analysis of a 225,000-word corpus of 60 environmental recommendation-based reports. It first describes the discourse-based coding system, which draws on the concept of genre move structure analysis, accounting for three different, but inter-related levels of text: (i) macrostructure; (ii) genre structure, and (iii) textual patterning, i.e. elements of the Problem-Solution pattern. It then describes the keyword analysis for the corpus as a whole and the key-key word analysis for each individual report. These keyword analyses provide internal linguistic evidence for classifying the reports as Problem-Solution based. An analysis of selected words (problem / problems and impact / impacts) reveals that their collocational behavior and involvement in certain causative phrases are related to specific discourse-based move structures.