You are here:
Data Supplement to Development and Evaluation of the Beta Streamflow Duration Assessment Methods for the Northeast and Southeast
Citation:
Gross, S., M. Eddy, K. Fritz, B. Topping, Tracie-Lynn Nadeau, R. Edgerton, R. Mazor, AND K. Nicholas. Data Supplement to Development and Evaluation of the Beta Streamflow Duration Assessment Methods for the Northeast and Southeast. U.S. Environmental Protection Agency, Washington, DC, EPA-843-B-23003, 2023.
Impact/Purpose:
Flow duration classification is used to implement several federal, state and local stream management programs. Because the flow duration of streams via existing maps, remote sensing, and gauging is constrained, field-based tools are often needed by practitioners. The data supplement report describes the study design, data analysis and interpretation of results that was used to produce the beta methods to rapidly classify stream reaches as ephemeral, intermittent, perennial, and at-least intermittent in the Northeast and Southeast regions of the United States.
Description:
The data supplement is a companion document to the user manual for the beta Streamflow Duration Assessment Methods for the Northeast and Southeast (beta SDAMs NE and SE) of the United States. The beta SDAMs NE and SE are field-based methods for rapidly classifying stream reaches as perennial, intermittent, at-least intermittent, or ephemeral. This data supplement details the data collection, data analysis, and evaluation steps that were used to develop the beta SDAMs NE and SE. Three biological, three geomorphological, and two geospatial indicators form the basis of the beta SDAM NE, whereas three biological, two geomorphological, and 2 geospatial indicators form the basis of the beta SDAM SE. The accuracy for classifying ephemeral, intermittent and perennial reaches in a testing dataset was 72.2% using the best NE random forest model and was 70.0% for the best SE random forest model. The accuracy for differentiating ephemeral from at least intermittent (perennial and intermittent) reaches increased to 92.2% and 91.4% for the NE and SE random forest models, respectively.