LINKING CONTINENTAL CLIMATE, LAND USE, AND LAND PATTERNS WITH GRASSLAND BIRD DISTRIBUTION ACROSS THE CONTERMINOUS UNITED STATES

Abstract

Associations of the abundance and temporal incidence of 17 grassland bird species with climate. weather, farm crops, and landscape metrics were determined for the conterminous United States using hierarchical models. We developed statistical models using two versions of classification and regression tree analysis in which the variation of each species' response variable (both as number of individuals (1973-1989) and as temporal incidence (1981-1990) per Breeding Bird Survey route) was recursively partitioned into statistically distinct chains of environmental determinants or associ­ ations. The predictive power of these models was bimodal, yielding high R2 values (above 38 percent) for one group of 12 species and low values (below 20 percent) for a second group of 5 (gener'.llly scarce or restricted-range) species. The fit of the models was strongly correlated with the size of each species' range. Climate variables-long-term annual precipitation, January temperature, and July tem­ perature-appeared in many of the species models, often with strong effects (large R2 values). January weather (annual deviation from long-term mean temperature) was also a consistent, though weaker, correlate. Sorghum (Sorghum vulgare) was the only strong crop correlate of most species abundances, but grain corn (Zea mays) and enrollment in the Conservation Reserve Program were consistent smaller contributors to most models. Wheat (Triticum aestivum) and durum wheat (T. durum) were other noteworthy variables, occurring in about half of the species models. The presence of soybeans (Glycine max) was a local modifier of abundance for almost all species. Considering only the leading variables for individual species, precipitation occurred in five species, grain corn in three, and durum wheat and sunflower (Helianthus sp.) in two each. The Conservation Reserve Program variable pre­ empted grain corn for two species in the two years Conservation Reserve Program data were available. Other leading variables each appeared in only one species. A parallel analysis using remotely sensed land-use data to assess the relative roles of land-cover proportions and habitat patch attributes showed that grassland species were more strongly influenced by habitat patch variables, but less strongly influenced by land-cover proportions, than were nongrassland species. Grassland species' sensitivity to habitat patch variables appeared to be greater in wooded and cropland habitats than in habitats dominated by grass.