Modeling Eurasian watermilfoil (Myriophyllum spicatum) habitat with geographic information systems

Eurasian watermilfoil (Myriophyllum spicatum) habitat was predicted at multiple scales, including a lake, regional, and national level. This dissertation illustrates how habitat can be predicted for M. spicatum using publically-available data for both presence and environmental variables. Models were generated using statistical procedures and quantative methods to determine where the greatest likelihood of presence was located. For the single lake, presence and absence data were available, but the larger-scale models used presence-only methods of prediction. These models were paired with a Geographic Information System so that data could be visualized on a map. For the selected lake, Pend Oreille (Idaho), spatial analysis using general linear mixed models was used to show that depth and fetch could be used to predict habitat, although differences were seen in their importance between the littoral and pelagic zones. For the states of Minnesota and Wisconsin, Mahalanobis distance and maximum entropy methods were used to demonstrate that available habitat will not always mean presence of M. spicatum. The differing approaches to management in these states illustrated how an aggressive public education campaign can limit spread of M. spicatum, even when habitat is available. Bass habitat appeared to be the largest predictor of M. spicatum in Minnesota, although this was due to the similar environmental preferences by these species. Using maximum entropy, on a national level, presence of M. spicatum appeared to be best predicted by annual precipitation. Again, results showed that habitat is colonized as time permits, and not necessarily as conditions permit