A PRACTICAL STREAMBANK EROSION MODEL FOR THE COASTAL PLAIN OF THE NORTHERN GULF OF MEXICO

Predictive field-based models have been developed to rapidly identify streams that can be expected to experience accelerated erosion and should be prioritized for restoration. One of the most widely applied models, the Bank Assessment for Non-point Source Consequences of Sediment (BANCS) model, correlates observed rates of streambank erosion with near-bank stress (NBS) and bank erodibility (BEHI). The BANCS model is region specific and has to be calibrated in every hydrophysiographic region. Our study tried to calibrate the BANCS model for the northern Gulf of Mexico coastal plain by collecting field data over a two year period at 75 sites in the Florida Panhandle, South Alabama, and South Georgia. The average annual streambank erosion rate ranged from 2 mm to 2 m and showed considerable variability between the two years. The spatial and temporal variability of the erosion may be due to variability in the extreme precipitation events that commonly occur in the region. Prediction of measured bank erosion rates with the standard BANCS model yielded low, statistically non-significant R² values. Modifying the standard BANCS shear stress factor (NBS) by multiplying it by bankfull width, which dimensionalizes NBS, improved the predictive power of the BANCS model (R²=0.31, p<0.05), especially for banks with low to moderate BEHI erodibility. However, to develop a more predictive streambank erosion model for the region we identified 19 variables related to channel geometry, precipitation, and vegetation. We then applied two automated statistical model selection methods using Akaike’s Information Criterion and repeated cross-validation. The most parsimonious models identified by the two methods were similar. A model incorporating bank slope, biomass density, channel curvature, vegetation cover, and BEHI had the strongest correlation to measured bank erosion rates (R² = 0.54, p<0.05). If erosion rate is multiplied by bank height, R² of this model increases to 0.65. Our study indicates that the standard BANCS approach is not optimal for the Gulf of Mexico coastal plain and that a well selected statistical model may be a more promising alternative tool for the stream restoration community in the region.