Tuesday, August 7, 2007 - 8:00 AM

COS 34-1: Predictive modeling for state endangered bird populations in Michigan

Katherine J. Kahl and Kelly F. Millenbah. Michigan State University

Our objective was to develop a predictive management tool based on land cover composition and structure variables selected for by State-endangered Prairie Warbler (Dendroica discolor) populations. Currently, specific land cover composition and structure characteristics of breeding territories are not adequately understood. To accurately document variables associated with species presence, point count surveys were used in 2004 and 2005 to locate Prairie Warblers along Michigan's west coast. Each male's territory was mapped using GPS (n = 41). Vegetation composition and structure were measured at two nested scales (30m2 and 1m2) within each territory and in areas where Prairie Warblers were absent. To gain insight into the appropriate management scale for the species, we compared vegetation measurements at specific geographic locations with corresponding Landsat (30m2 scale) and 1m2 aerial imagery pixels to determine at which scale land cover characteristics, identified in our analysis as key for species presence, can be “seen”. To determine where other territories could exist (based on identified breeding sites), we are developing a predictive model using the R statistical program, ArcView and ArcGIS. Preliminary correspondence analysis shows spatial separation between species presence and absence locations. Analysis suggests presence is associated with certain vegetation species and that a combination of structural variables may be driving habitat selection. Model predictions will be validated using other 2004-2006 Prairie Warbler location records in Michigan (not used to develop the model). Current analyses suggest that a large amount of suitable, yet under-utilized, habitat exists in Michigan supporting the need for edge of range management considerations.