Marsh bird habitat relationships at multiple spatial scales: A case study of the swamp sparrow (Melospiza georgiana) in the Great Lakes Region

Although Aves is one of the best-studied taxa, much remains to be learned about marsh bird populations. Understanding species-environment relationships for marsh birds is complicated by the low detection probabilities of many species and the challenges associated with sampling wetland habitats. I evaluated habitat relationships at multiple spatial scales for a marsh passerine, the Swamp Sparrow (M. g. georgiana), in the Great Lakes Region. At the local scale, I monitored breeding Swamp Sparrows at restored and natural marshes and assessed the degree to which habitat attributes influenced nesting success. I hypothesized that nesting success would be lower in restored marshes because of shallower water depths and differences in vegetation composition that may increase nest site accessibility to predators or alter food resources. The daily nest survival rate varied as a quadratic function of date, with survival being highest in the middle of the nesting season. Pooled daily nest survival rates were 0.917 (SE = 0.010) for the restored marshes and 0.937 (SE = 0.009) for the natural marshes. The nest success rate of 0.11 for restored marshes was significantly less than the nest success rate of 0.18 in natural marshes. All four marshes appeared to function as demographic sinks, although there were large uncertainties in estimated adult and juvenile survival rates.;At the patch or wetland scale, I surveyed 21 wetlands using the North American Marsh Bird Monitoring Protocol to compare detection probabilities, species densities, and marsh bird community composition at restored and natural marshes. I expected that detection probabilities and community composition would be generally similar, although I expected slightly fewer wetland species at restored marshes because of lower water levels. I recorded 35 wetland-associated species, comprised of 30 species at natural marshes and 31 species at restored marshes. I estimated detection probabilities for the Swamp Sparrow, Common Yellowthroat (Geothlypis trichas), and Song Sparrow ( Melospiza melodia). Detection probabilities differed among species but did not differ between restored and natural marshes. Densities of these three species were similar in natural marshes; however, Swamp Sparrows trended toward lower densities in restored marshes. Species richness and similarity and diversity indices suggested that the breeding bird communities were similar at restored and natural marshes.;At the population scale, I modeled Swamp Sparrow abundance and evaluated how the choice of spatial and temporal scales used to aggregate Breeding Bird Survey (BBS) data influenced model fit and performance. I hypothesized that Swamp Sparrow abundance would be negatively influenced by the amount of cropland and medium- and high-density development, road density, and temperature, and be positively influenced by the amount of wetland area, stream density, and winter precipitation. I expected that models based on intermediate spatial data aggregation scales (e.g., 5-stop or 10-stop segments) would exhibit the highest concordance with environmental variables but that route-scale models would exhibit the best performance. All top-ranked models of Swamp Sparrow abundance contained a measure of proximal wetland land cover. The highest concordance between wetland land cover and Swamp Sparrow abundance occurred at small rather than intermediate spatial extents, ranging from 200 m to 400 m from BBS stops, segments, or routes. Contrary to my expectations, road density, the amount of cropland, and the amount of high- and medium-density development did not exhibit a strong negative effect on Swamp Sparrow abundance. The average temperature recorded during each breeding bird survey influenced Swamp Sparrow abundance within a single year, but this effect disappeared when counts were summed over multiple years. Winter precipitation positively influenced Swamp Sparrow abundance aggregated for 5-stop and 10-stop segments, but was not a significant predictor in route-scale models. Route-scale models exhibited the best performance, with Spearman's rank coefficients between observed and predicted mean values averaging 0.47. Correlation coefficients for models based on 5-stop and 10-stop segments ranged from 0.26 to 0.38. My findings confirm that Breeding Bird Survey data may be useful for monitoring and modeling the populations of abundant marsh birds; however, greater uncertainty may be present in marsh bird models based on fine spatial and temporal data aggregation scales. (Abstract shortened by UMI.).