Quantifying the Variability of Stream Health Indicators for TMDL Assessment

Bruce Wilson, Professor
John Nieber, Professor
Gary Sands, Associate Professor
James Perry, Professor and Head, Fisheries and Wildlife
Bruce Vondracek, Professor, Fisheries and Wildlife

Funding Source

Minnesota Pollution Control Agency

Objective

The overall goal of the proposed study is to improve TMDL assessments by obtaining a better understanding of the variability of stream health indicators. Objectives are to:

1. Assess watershed conditions along different reaches of the channel,
2. Evaluate the natural variability within reaches of uniform watershed conditions, and
3. Develop a composite index of several indicators of stream health that can be used in TMDL assessments.

Need or Impact

Stream health indicators typically vary with longitudinal position. This variability contributes to the uncertainty of TMDL assessment. Variability is caused by changes in watershed factors, such as land use, geologic conditions, and soil characteristics, as well as natural variability of fluvial systems. This natural variability results in different indicator values for the same (apparently) watershed conditions. One useful outcome of the project will be to investigate the impact of TMDL implementations. If natural variability is large (i.e., stream health indicators vary with longitudinal position), then the impact of TMDL implementation will need to be substantial to detect its benefits to stream health.

Project Status

GIS data layers have been obtained for many locations in Minnesota and were used to identify six basins for analysis of stream health. Fish IBI was selected as a good indicator of stream health. Reasonably good regression relationships for Fish IBI scores were obtained for basins with relatively uniform characteristics. Poorer fits were obtained as the land use and other conditions within the basin became less uniform. The impact of these factors could not be captured by predictor variables of the study. Additional work is continuing using artificial neural networks where an inherent linear relationship among predictors and IBI is unnecessary.