Environmental indicators of macroinvertebrate and fish assemblage integrity in urbanizing watersheds

Abstract

Urbanization compromises the biotic integrity and health of streams, and indicators of integrity loss are needed to improve assessment programs and identify mechanisms of urban effects. We investigated linkages between landscapes and assemblages in 31 wadeable Piedmont streams in the Etowah River basin in northern Georgia (USA). Our objectives were to identify the indicators of macroinvertebrate and fish integrity from a large set of best land cover ( n = 45), geomorphology ( n = 115), and water quality ( n = 12) variables, and to evaluate the potential for variables measured with minimal cost and effort to effectively predict biotic integrity. Macroinvertebrate descriptors were better predicted by land cover whereas fish descriptors were better predicted by geomorphology. Water quality variables demonstrated moderate levels of predictive power for biotic descriptors. Macroinvertebrate descriptors were best predicted by urban cover (−), conductivity (−), fines in riffles (−), and local relief (+). Fish descriptors were best predicted by embeddedness (−), turbidity (−), slope (+), and forest cover (+). We used multiple linear regression modeling to predict descriptors using three independent variable sets that varied in difficulty of data collection. “Full” models included a full range of geomorphic, water quality and landscape variables regardless of the intensity of data collection efforts. “Reduced” models included GIS-derived variables describing catchment morphometry and land use as well as variables easily collected in the field with minimal cost and effort. “Simple” models only included GIS-derived variables. Full models explained 63–81% of the variation among descriptors, indicating strong relationships between landscape properties and biotic assemblages across our sites. Reduced and simple models were weaker, explaining 48–79% and 42–79%, respectively, of the variance among descriptors. Considering the difference in predictive power among these model sets, we recommend a tiered approach to variable selection and model development depending upon management goals. GIS variables are simple and inexpensive to collect, and a GIS-based modeling approach would be appropriate for goals such as site screening (e.g., identification of reference streams). As management goals become more complex (e.g., long-term monitoring programs), additional, easily collected field variables (e.g., embeddedness) should be included. Finally, labor-intensive variables (e.g., nutrients and fines in sediments) could be added to meet complex management goals such as restoration of impaired streams or mechanistic studies of land use effects on stream ecosystems.