Many streams in Florida, and in the US in general, are impaired due to excessive sediment loading. Erosion of streambanks is a major source of the sediment loading. The sediments have many negative economic and environmental impacts and restoration of streams to mitigate these impacts is a multi-million-dollar industry in the US. Because of the large number of eroding streambanks, insufficient funds are available to restore all impaired streams. Consequently, funding agencies need a method to prioritize streambanks for restoration. Prioritization of streambanks has to be based on observed and expected future erosion rates but future rates are hard to predict because prediction models must be based on multi-year observations and easily applied. Several methods have been developed to predict streambank erosion but the most popular one is the BANCS method. The BANCS method predicts streambank erosion with regression models for various categories of banks based on the near-bank-stress of the streamflow and the erosion potential of the streambank. Because the BANCS method is an empirical model it has to be calibrated for every physiographic region. The present study collected data at 75 sites in the Florida Panhandle and adjacent areas of southern Alabama and Southwest Georgia over a three-year period to calibrate the BANCS method for the coastal plain of the northern Gulf of Mexico. Data were collected in accordance with the standard BANCS method but additional ancillary information to enhance the model were also gathered. Annual streambank erosion rates ranged from 2 mm to 1.97 m over the two-year study period. None of the sites had a bank erosion potential in the very low category, indicating that most streams in the northern Gulf coastal plain are prone to bank erosion. The standard BANCS model is not a good predictor of streambank erosion in the study area: R2 values of regression models for various BEHI and NBS categories of banks were very low, relationships were sometimes the inverse of what was expected based on an understanding of physical processes involved, and models were statistically not significant. Dimensionalizing NBS method 5 resulted in the best but still moderately effective model. We developed a more robust statistical approach using a nonlinear model, data for additional hydrological and geomorphological parameters, and assessment of the effectiveness of all possible subsets of predictor variables. This approach resulted in a much better predictive model (R2 ≈ 0.6 with five predictor variables).
Files and links (1)
pdf
Development of regional bank erosion relationships for...1,003.20 kBDownloadView