Quantifying landscape effects on ecological processes relevant to disease emergence is becoming increasingly recognized as important for preserving biodiversity and protecting human health. This is particularly true for diseases with little ecological information related to reservoirs or vectors. This multidisciplinary study employs remote sensing satellite technology to quantify land use/cover; standard ecological sampling and epidemiological survey procedures to integrate aquatic ecology with human disease incidence; and, microbiological and molecular techniques to understand the habitats and distribution of an environmental pathogen, Mycobacterium ulcerans. Our primary goal was to evaluate landscape relationships with water quality and aquatic community structure that influence the distribution of M. ulcerans among waterbodies in Ghana, West Africa. Specific objectives were to 1) use data mining techniques to identify potential environmental-biological interactions influencing pathogen distribution and human disease for focused hypothesis testing; 2) evaluate common environmental drivers of pathogen versus human disease distribution; and, 3) establish aquatic bioindicators of disease risk. In three regions of Ghana land use/cover was quantified using satellite imagery and analyzed with 23 water quality variables and metrics of community structure (algae and invertebrates) for 98 waterbodies. The pathogen was detected using PCR of loci specific for M. ulcerans. Principle component analysis was used to evaluate land use/cover and water quality variables to include in classification tree (CT) modeling.

Results/Conclusions

Environmental CT models included region, waterbody classification type (e.g., pond, river, wetland), waterbody flow type (i.e., lentic, lotic), waterbody history (i.e., natural, human made), dominant macrophyte taxa, latitude and 21 water quality variables (e.g., dissolved oxygen, nitrates, ions). Landscape CT models included a subset of 20 original land use/cover types in large-scale buffers surrounding a waterbody.There were very different classification trees constructed for human disease and pathogen distribution. Including macrophyte taxa and latitude into models affected site partitioning in the environmental analyses, while buffer width affected partitioning in landscape analyses. For all trees, region was the most important variable for partitioning sites, where no human disease or pathogen was detected in an entire region. Non-metric multidimensional scaling identified relationships of different algae and invertebrate communities associated with pathogen distribution relative to human disease incidence, and several integrated metrics of ecological conditions were related to human disease but not pathogen occurrence. We hypothesize that focal human disease emergence is a function of dynamic ecological processes within waterbodies that are mediated by complex human landscape associations.