Craighead Environmental Research Institute

Home Conservation Area Design Conservation Biology & Genetics Conservation GIS Conservation Education About Us Links Join Us The mission of the Institute is to increase humankind's understanding, appreciation, and protection of our natural environment; particularly wildlife populations and wild landscapes.  Our goal is to enable human beings to live in harmony with other species.    Madison Valley - Taylor Fork Grizzly Bear Model Comparison Project Grizzly Bear Habitat Effectiveness Model Comparison Models of both habitat quality (or capability) and effectiveness (or suitability) for grizzly bears (Ursus arctos) are tools that can help understand the relative effects of human developments on bear distribution and movement, and can identify areas where grizzlies are likely to survive. To determine whether current habitat models accurately predicted grizzly bear locations, and to compare relative accuracy of various models, we evaluated habitat models from 4 sources; Carroll et al. (2001), Merrill and Mattson (2003), the Yellowstone Grizzly Bear Cumulative Effects Model (2002), and Walker and Craighead (1999) by comparing model results with known locations from GPS-satellite collars. The Walker-Craighead Model was a modified version that had been refined by Lance Craighead and Mike Rock: the CERI model. The Taylor Fork drainage. We used a large and very accurate dataset from 3 years of data collection by the Interagency Grizzly Bear Study Team (IGBST) using GPS collars on grizzly bears in the Yellowstone Ecosystem.  To maintain the security of grizzly location information, no location data was released by the IGBST.  CERI scientists provided standardized model results to Mark Haroldson of the IGBST.  Mark evaluated each of the models at each GPS location (each animal location was overlaid with landcover data to determine the landcover class occupied by the grizzly bear at that time and the ranking that each model gave to that location).  The Upper Madison Study Area (UMSA) comprised 8,425 km2 of habitat utilized by grizzly bears, located at the northwest corner of Yellowstone National Park.  Grizzly bears used habitat around the Taylor Fork drainage and the upper Madison River valley. About half (~44%) of the study area was within the Grizzly Bear Recovery Zone. In collaboration with CERI, the IGBST analyzed GPS locations from 6 bears during 2000 to 2002.  These comprised 4,855 total locations, accurate to within about 30m.  The composite home ranges of the six bears totaled 4,300 km2 considered "available habitat".  Of this total, 2,768 km2 of available habitat was within the Grizzly Bear Recovery Zone.   We conducted a statistical analysis considering all locations as independent samples: at 2 different spatial scales of selection employing receiver operators characteristic (ROC) curves.  A simple expert opinion-based habitat suitability model with two components (CERI model) performed as well, or better, than the other, resource selection function-based, models evaluated.  The basic CERI habitat effectiveness model was then further refined, and put into an ESRI ArcGIS Modelbuilder format. The CERI model rates habitat quality and then degrades it in relation to human influence to produce a final Habitat Effectiveness Model at each scale.  ROC curves for all models using all locations: Results of the ROC curves demonstrate that the CERI models, the Carroll model, and the CEM HV model fit the observable data (GPS point locations) better than other models.             Although all models used reasonable logic based upon scientific knowledge of grizzly bears, they resulted in very different outputs.  There was not a clear “best” model, but some were better than others at describing habitat which grizzly bears used.  The addition of human disturbance components to each model did not greatly improve the predictive capabilities, and in most cases reduced the performance of the model.  This indicates that the weighting of disturbance factors (roads, buildings, remoteness) can be improved and perhaps the overall performance of the models can be improved.  In addition, individual variation amongst bears reduced the overall model results.  Bears were very individualistic in terms of habitat use, and some bears ‘fit’ some models better than others.             These results are very encouraging because they lend credence to the use of the simple CERI model, which can be extended to map grizzly habitat throughout the GYE and (with appropriate regional habitat rankings) throughout the Rocky Mountains.  The Wildlife Conservation Society has asked to use this model as one of their landscape species habitat models for a Madison Valley Conservation Design.  We are currently planning to further evaluate the CERI model over the entire range of GPS data points in the GYE, and are working with state and federal agencies, and conservation NGOs, to use the model results, and linkage model results, to inform conservation and land management decisions in the GYE.    COPYRIGHT 2002, ALL RIGHTS RESERVED For problems or questions regarding this website contact ursus@grizzlybear.org www.grizzlybear.org