Predicted Number of Grassland Nesting Bird Pairs in the Prairie Pothole Region of Minnesota and Iowa
Iowa and Minnesota PPR Grassland Bird Models 2007
Landscape models for grassland birds were based on point count data collected in June 2003-2005 by and under the direction of Frank Quamen (PhD student), Dave Naugle (major professor, Montana State University), Shane Patterson (MS student) and Rolf Koford (Iowa State University). A stratified random sample was drawn each year based on the amount of grass in the landscape, cover type (grass (70% of points), hay (15%), crop (15%)), and by seven wetland management districts to distribute the sample across the prairie pothole region of Minnesota and Iowa.
The number of detections was large enough to produce models for Sedge Wren, Clay-colored Sparrow, Grasshopper Sparrow, LeConte’s Sparrow, Savannah Sparrow, Bobolink, Dicksissel, Western Meadowlark, and Horned Lark. Horned Lark were most often found in cropland and had a negative relationship with grass; therefore they are not considered in the following analyses and products. A variety of regression methods were used to model bird density (bird detections/ha) including negative binomial (NB), zero-inflated negative binomial (ZINB), and logistic regression. NB and ZINB models predict number of birds per ha, while logistic regression predicts presence/absence. Logistic regression was used for Western Meadowlark because there was usually only one bird detected when this species was found. For each species, the best model was selected using information theoretic selection methods (i.e., AIC, BIC). Note: incorporating local habitat factors improved models significantly (Table 4a-c, Quamen 2007), but such factors cannot be mapped on broad geographic scales.
Models were mapped by applying regression equations to grids of percent cover and BBS relative abundance (used as a spatial covariate to account for species range), and limited to grass areas >1 ha. For models that predicted the number of pairs (i.e., all but western meadowlark), the total number of pairs per 40 acres was predicted by calculating the pairs predicted for the grass portion of each 40 acre cell. Maps were ‘smoothed’ by taking the average predicted pairs in a 40 ac cell for a 1-mile radius. Although species models differed by landscape scale and specific landscape variables, the resulting maps were similar. To summarize and simplify, the total predicted number of pairs was summed for Sedge Wren, Grasshopper Sparrow, LeConte’s Sparrow, Savannah Sparrow, Bobolink, and Dicksissel. Clay-colored Sparrows were not included because of numerical issues in estimating the model (see Quamen 2007); results for this species should be interpreted with caution. Western Meadowlark were not included because the model predicts probability of presence, not number of pairs. Because all species models improved with the inclusion of local variables, landscape models depict the potential for breeding pairs where suitable local habitat exists.
Quamen, F. R. 2007. A landscape approach to grassland bird conservation in the prairie pothole region of the northern great plains. Dissertation, University of Montana, Missoula.