Predicting in-stream pathogen levels has long been known to be a challenging problem due to complex interactions between microorganisms and the natural stream environment, and the spatial heterogeneity involved in stream networks of a watershed. Here we have developed models for predicting E. coli (a pathogen indicator) in streams. In E. coli estimation, the first modeling approach uses Geographic Information Systems (GIS) based watershed indexes considering the undisturbed land cover, which encompasses the natural land cover area, wetlands, and vegetated stream corridors, and the disturbed land cover extent which includes areas receiving manure from confined animal feeding operations, tile-drained areas, and areas under cropped and urban land cover. The second approach involves developing mathematical models for calculating E. coli resuspension, deposition, in-stream routing, and growth in the streams. A hydrological model capable of predicting in-stream E. coli concentrations in the streambed sediment as well as in the water column was developed. In order to develop the hydrological model for predicting in-stream E. coli concentrations, firstly a model capable of predicting E. coli resuspension was formulated. Secondly, formulations for calculating in-stream E. coli routing, water temperature depended E. coli growth, and the streambed sediment and water column E. coli concentrations were developed. Finally, these formulations were programmed in FORTRAN language, and were integrated into the Soil and Water Assessment Tool (SWAT), a watershed scale hydrological model, written in FORTRAN. In addition to the model development, this study also involves monitoring E. coli concentrations in the streambed sediment and the water column extensively starting from May 2009 to December 2011 in the Squaw Creek Watershed, Iowa, USA. The observations were used to verify the model predictions, and results indicated that the models performed well.

The GIS based approach developed here for estimating E. coli concentrations in streams can be potentially useful in predicting in-stream waterborne E. coli levels using watershed indexes. Approximately 95- 98% of the predictions were within 1 order magnitude of the observed values, when we used hydrologically corrected watershed indexes for E. coli estimation. The model skills varied from 0.39 to 0.55.

In E. coli resuspension model, approximately 81% of the predicted E. coli resuspension rates were within a factor of 2 of the inferred values (i.e., measured E. coli). All of the predicted resuspension rates were within a factor of 5 of the inferred values. The model skill value of 0.85 indicated that the model predicts E. coli resuspension rates successfully.

The application of the modified SWAT model in the Squaw Creek Watershed, which was developed here, performed well. For example, approximately 62% of the predicted streambed sediment E. coli, and 82% of the predicted water column E. coli concentrations were within 1 order magnitude of the measured concentrations. The R2 for monthly average daily flow was 0.99, while for daily flow predictions R2 was 0.42. The Nash-Sutcliffe's efficiency (NSE) for monthly average daily and daily flow predictions were 0.75 and 0.39, respectively.

We also developed a model for calculating in-stream total E. coli loads (i.e., contributions from the streambed as well as from free floating) in order to improve Total Maximum Daily Loads (TMDLs) estimation, and understand the potential impacts of streambed sediment E. coli on total in-stream E. coli loads. While comparing the total predicted E. coli loads with the measured E. coli loads, coefficient of determination (R2) was 0.82, and model skill was 0.78; these results indicates that the model for calculating total in-stream E. coli loads performed well, and should help in developing Total Maximum Daily Loads (TMDLs) for stream bacteria.

In addition to in-stream processes and overland flow, weather pattern can potentially impacts in-stream E. coli concentrations. To understand the impacts of weather pattern on in-stream E. coli concentrations, E. coli observations in the streambed sediment and the water column (from two locations) were related with climate data (i.e., air temperature, soil temperature, solar radiation, and rainfall). The results show that increase in temperature increases E. coli concentrations not only in the water column but also in the streambed sediment. Moreover, E. coli in the streambed sediment remained elevated even at relatively lower temperature. These results signify that increase in ambient temperature can potentially increase E. coli levels in the water bodies.

The results from monitoring and modeling of in-stream E. coli presented here will have significant importance in developing Total Maximum Daily Loads (TMDL) for in-stream pathogens as well as predicting E. coli concentrations in the streams at the watershed scale.