1Methods for Estimating Distributions• Static Distributions– Polygon– Grid– Habitat Mapping• Polygon Method– Relies on empirical knowledge of specialists– Likelihood of occurrence unspecified• “Grid” Method– Delineated by all subunits where presence in confirmed– Likelihood of occurrence unspecifiedBlackpoll Warbler Distribution in New York State2Habitat Mapping• 2 Phases– Model occurrence-habitat relationship– Model distribution based on map of habitat • Example bull trout in Nevada and southern Idaho (Dunham et al. 2002)Dunham, J. B., B. E. Rieman, and J. T. Peterson. 2002. Patch-based models to predict species occurrence: lessons from salmonid fishes in streams. In Predicting Species Occurrences.Goal: Predict occurrence of fish in patches of habitat suitable for local breeding populationsPossible Factors Affecting Bull Trout Distributions• Natural and artificial dispersal barriers• Water temperature• Interactions with non-native salmonids and other fishes (brook trout)• Human disturbance (road density)• Geographic influences (‘patch size’, stream gradient and width)3Occurrence-Habitat Model• Used logistic regression to model probability of occurrence based on various combinations of several factors• Likely limiting factor for Nevada and southern Idaho was warm summer temps– Used elevation as surrogate for water temp. to delineate downstream distribution limit• “Patch” size– Delineated upstream patch area as size of watershed upstream from lower limitDistribution Evaluation• Patches with >.5 probability-of-occurrence were predicted to be occupied• Evaluation: Cross-validation19.7 (overall)15.4Unoccupied27.6OccupiedError (percent misclassified)Actual Patch Status4GAP Analysis• GAP seeks to identify “gaps” that may be filled through establishment of new reserves or changes in land management• Maps species distributions by combining habitat mapping method with known occurrence dataRequired Information for GAP • Digital map of vegetation, cover types, or habitat types• Digital map of study are divided into geographic units (e.g., counties, grid)• Database indicating presence/absence in each geographic unit• Database predicting presence/absence in each vegetation or habitat typeExample: 100 Breeding birds in California(Garrison and Lupo 2002)• Included habitats rated as Low, Medium or High by the California Wildlife Habitat Relationships (CWHR) system• Map further refined by retaining habitat polygons in counties where species was known to breed5Distribution Evaluation• Tested map predictions against Breeding Bird Survey records from 1977-1996 33.3 (range 5.1 – 71.7)Unoccupied1 (range 0 – 12.1)OccupiedMean Error (% misclassified)Patch StatusAccuracy Dependent On…• Maps most accurate for species that were– Relatively abundant– Relatively large breeding ranges– Territorial– Associated with terrestrial
View Full Document