Modeling nest-site occurrence for the Northern Spotted Owl at its southern range limit in central California

At the southern end of its range, the Northern Spotted Owl (Strix occidentalis caurina) occurs in high densities and nests in a wide range of forest types and ages, exhibiting different foraging and nesting habits than in the northern part of its range. The intensive monitoring of this subspecies on public lands in Marin County, California, combined with the availability of fine-scale geographic information system (GIS) data, provided a unique opportunity to apply and evaluate a habitat-based species occurrence modeling approach at the scale most relevant to local land managers and planning agencies. We used 4 years of breeding owl survey data (1998-2001) and GIS layers representing topographic, anthropogenic, and vegetation-based landscape characteristics to build logistic regression models of owl nest-site occurrence. Models were used to develop spatial predictions of occurrence within the study area and in adjacent ecoregions, which were validated with an independent dataset. We also compared the predictive performance of two vegetation layers differing in their floristic detail and spatial accuracy. The model based on a local vegetation layer generally exhibited better model performance than the model based on the more generic regional layer. Model results indicated that forest connectivity and topographic conditions, rather than forest type or age, were the strongest predictors of nesting owl presence. Predicting outside the original study area was somewhat successful for a coastal ecoregion similar in vegetation and climate, but not better than random for a nearby inland ecoregion, suggesting that locally derived models are necessary to adequately predict nest-site occurrence.