A test of an expert-based bird-habitat relationship model in South Carolina
Wildlife-habitat relationships models are used widely by land managers to provide information on which species are likely to occur in an area of interest and may be impacted by a proposed management activity. Few such models have been tested. We used recent avian census data from the Savannah River Site, South Carolina to validate BIRDHAB, a geographic information system (GE) model developed by United States Forest Service resource managers to predict relative habitat quality for birds at the stand level on national forests in the southeastern United States. BIRDHAB is based on the species-habitat matrices presented by Hamel (1992). Species-specific accuracy rates for BIRDHAB predictions (the percentage of all stands in which a species was predicted correctly as present or absent) ranged from 33.6-93.0X, with a mean of 67.4±17.3% (SD, n=46 species). Accuracy was >90% for 5 species, but < 50% for 9 species. BIRDHAB performed well (P < 0.05) in predicting presence-absence of 32 species. Generally, the model was more accurate in predicting presence-absence for habitat specialists than for generalists. Habitat-specific accuracy rates (the percentage of species for which a habitat's prediction was correct) ranged from 52.7-92.7%, with a mean of 71.81 ± (SD) 9.8% (n-26 habitat types). BIRDHAB was a useful tool for many of the species that we tested, but it had no predictive ability for many others. Such species-specific variation in accuracy probably is common among wildlife-habitat relationships models, reinforcing the need for thorough testing before these models are used in land-use planning.