One major task of the Michigan Department of Natural Resources (MDNR) is to provide accurate information to policy makers on the impacts of various land use decisions on the state’s natural resources. However, monitoring the impacts of resource use and landscape change on wildlife habitat throughout the state is a daunting assignment. MDNR personnel are seeking to implement a system of habitat accounting for all species, not just the important game species or rare species that have been monitored in the past.
To this end, we assessed the utility of Michigan’s Integrated Forest Monitoring, Assessment, and Prescription (IFMAP) database as a tool for tracking statewide quantities of wildlife habitat for over 200 species. We sampled vegetation, birds, and small mammals at over 300 field sites across the lower peninsula of Michigan. We used recursive partitioning (RPART) trees to build wildlife habitat models for birds and small mammals with vegetation measurements from both plot-level field surveys and the IFMAP stand-level field surveys.
Results/Conclusions
The plot-level field vegetation measurements led to only slightly better accuracy (when averaged for a large number of species) than the stand-level IFMAP data (Kappa = 0.5 vs. 0.45). Errors of omission (where a species is predicted to be present, but is actually absent in the field survey) are much higher for rare species vs. common species, while commission errors increase only slightly. Habitat models that were built solely with land-cover data were much less accurate overall than models that included detailed vegetation composition and structure information.
These results establish that the IFMAP database is appropriate for use in modeling wildlife habitat. The next step for this work is to employ wildlife habitat models in the IFMAP decision support tool. Since appropriate statewide survey data from which to fit statistical models do not exist, we created expert-based models for each species in a manner similar to GAP methods. These models were based on published habitat accounts and local habitat associations, and while the accuracy is lower than models fit with RPART, they are a dramatic improvement over the original Michigan GAP models. There are considerable technological hurdles to implementing a wildlife habitat module into the existing IFMAP decision support tool, but we believe that it will be very useful future efforts to simultaneously manage resource use and wildlife habitat in real time.
