144 Understanding invertebrate responses to urbanization: The role of multilevel regression models

Tuesday, May 19, 2009: 11:45 AM
Pantlind Ballroom
Thomas F. Cuffney , North Carolina Water Science Center, U.S. Geological Survey, Raleigh, NC
Roxolana Kashuba , Nicholas School of the Environment, Duke University, Durham, NC
Song Qian , Nicholas School of the Environment, Duke University, Durham, NC
Gerard McMahon , North Carolina Water Science Center, U.S. Geological Survey, Raleigh, NC
The effects of urbanization on macroinvertebrates were studied in nine metropolitan areas (Boston, MA; Raleigh, NC; Atlanta, GA; Birmingham, AL; Milwaukee-Green Bay, WI; Denver, CO; Dallas-Fort Worth, TX; Salt Lake City, UT; Portland, OR) as part of the U.S. Geological Survey’s National Water-Quality Assessment Program. Simple linear regression models relating invertebrate responses (e.g.,  Y = ordination site scores or metrics) to urbanization (e.g., X= multimetric urban intensity index, % developed land, or % imperviousness) showed statistically significant responses to urbanization in six of the nine metropolitan areas.  Multiple regression models incorporating additional basin-scale land cover (e.g., forest, agricultural land) and environmental variables (e.g., temperature, precipitation) as predictor variables could not explain differences in responses among metropolitan areas.  Multilevel models incorporating basin-scale (% developed land) and regional-scale predictors were able to explain differences in responses among metropolitan areas.  These models demonstrated that regional-scale climate (temperature and precipitation) and land cover (antecedent agriculture) strongly affect invertebrate responses to urbanization.  The non-significant regressions initially encountered in Milwaukee-Green Bay, Denver, and Dallas-Fort Worth resulted from the high levels of antecedent agriculture that obscured the effects of urbanization in these metropolitan areas and could not be detected using other regression models.