Linking the diversity and abundance of stream biota to landscapes in the mid-Atlantic USA

The amount of human altered land surface within a watershed has long been known to influence the biota of receiving streams and waterways. In contrast, vegetation in riparian zones can filter pollutants and reduce flow velocities that incise stream channels. We examined the relationship between the built environment and water quality for streams (1st through 4th order) across three physiographic provinces of the State of Maryland, USA, including 59 watersheds with some 865 stream sampling locations. We used image data products capable of discriminating fine-scale information of the land surface, including proportional impervious, tree, grass and crop cover. Stepwise multiple linear regression and decision tree statistical approaches were used to assess the relationship between the land cover predictors and benthic indices of biological integrity (BIBI) and number of sensitive invertebrate taxa (NEPT), response variables derived from the Maryland Biological Stream Survey (MBSS). Impervious and tree cover were found to be the best predictors of stream biota, although this varied with physiographic province and the response variable of interest. The best multivariate models predicted 65% of variability in BIBI and 62% of NEPT (p < 0.001). The configuration of tree cover across the landscape and the distance of land cover to the stream channel was found to be important in many, but not all cases for improving the predictive quality of statistical models estimating stream biota metrics. This work advances the estimation of stream health characteristics in areas where MBSS measurements do not exist but can be estimated using comparable land cover information, and informs guidelines for management and restoration.