Poster Presentation C Thursday, 5:00 – 6:00 pm C Edmund Fitzgerald Exhibit Hall
Is It, and Does It Matter? Predicting In-Stream Conditions from Land Use at
Varying Distances from the Channel
Natural Resources Research Institute
University of Minnesota Duluth
5013 Miller Trunk Highway
Duluth, MN 55811
Riparian land cover has many important functions, yet is among the most difficult landscape features to characterize and map. In non-forested landscapes the riparian zone is frequently smaller than the minimum resolution of traditional satellite imagery, with the result that some land cover classes are either missed altogether or misclassified. We mapped land cover in 12 watersheds in Southeastern Minnesota using high-resolution aerial photography (0.5 meter resolution) and satellite imagery (30 meter). Our objectives were to quantify the bias for each data source and to determine how land cover at varying distances from the stream influences in-stream properties. Satellite imagery underestimated the proportion of forest and wetland classes and overestimated the proportion of agricultural land within 30, 60m and 120m of the channel compared to the high resolution imagery. We developed regression models predicting TP and TN concentrations, chlorophyll a, percent fines, maximum temperature, number of debris dams, number of fish species, and number of scrapers and predators from land cover in buffer widths varying from 2 to 500 m. Percent predators was best predicted by land cover within 10 m of the channel, while % fines and TN were best predicted by land cover within 120m of the channel, illustrating how buffer size can affect the power to predict various in-stream processes.