This thesis investigates the variability in management practices on coffee and banana farms in an attempt to identify practices that reduce the environmental impact of export crop production. Different banana production systems are studied to determine their level of environmental impact. Insect sampling and bird surveys are used to assess the level of ecological quality on banana farms and their surrounding environments. The first two studies are based on interview methods and focus more on the social aspects of the production system. Paper I identifies how farmers utilize labor and herbicides in weed control practices, and found that small-scale coffee farmers overuse herbicides when their relative use of herbicides to labor to control weed densities is compared to their large-scale counterparts and small-scale organic producers. Paper II attempts to identify variability in management practices for the production of export bananas, but instead finds that there is only one type of export banana production system. However, there are lessons to be learned from organic and banana-coffee intercropping systems of production.

Papers III-V use the information gathered in the interview studies of Paper II to give context to the results from analysis of ecological indicators collected from banana farms. Paper III is a comparison of insect community composition on high-input, low-input and organic banana farms. Paper IV is an analysis of aquatic macroinvertebrate in surface water sites upstream and downstream of banana farm canal entry points. Finally, Paper V is a comparison of ecological effects of management practices between Rainforest Alliance certified farms and non-RainforestAlliancecertified farms. Results showed that low-input banana production is not as good as organic production with regards to ecological impact, but it can still make a difference when compared to high-input banana production. Rainforest Alliance certified farms, however, are not low-input systems and the changes that they make in production practices are not enough to influence the quality of the ecological system. These results are encouraging for low-input production systems, but show that standards for Rainforest Alliance certification need to be tougher in order to make an impact on ecological indicators.