Abstract
AbstractThis paper presents a new way of modeling in-stream bacteria concentrations by examining a watershed in terms of wet and dry hydrologic states. Flow-duration curves were developed for four catchments of the Little River Experimental Watershed in Tifton, Georgia, and HSPF and SWAT bacteria models were built for the catchments. The flow-duration curves were used to designate wet and dry states of the catchments based on flow conditions instead of calendar day, the bacteria data sets were divided into wet and dry groups accordingly, and the models were calibrated to the wet and dry states. Water-quality parameter sensitivities revealed that each model placed varying emphasis on the parameters in each state according to the model’s structure, and certain parameters were insensitive in both wet and dry states. A custom parameter added to the models to represent background in-stream and/or distributed loads was consistently sensitive across hydrologic states and improved model predictions in both computer models. Seven of the eight scenarios considered attained better model predictions in the wet state than the dry state as evaluated by a log-transformed Nash-Sutcliffe efficiency. The results show that fundamental aspects of the models’ performance are revealed in light of the analysis by hydrologic state and can provide future guidance for the collection of better datasets for use in bacteria modeling.
