Submitted to: Transactions of the ASABE
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/15/2009
Publication Date: 11/15/2009
Citation: Parajuli, P.B., Douglas-Mankin, K.R., Barnes, P.L., Rossi, C.G. 2009. Fecal bacteria source characterization and sensitivity analysis of SWAT 2005. Transactions of the ASABE. 52(6):1847-1858. Interpretive Summary: The presence of fecal bacteria in surface waters is a major health issue in the U.S. Fecal bacteria sources include the land application of animal manure, failing septic systems, and wildlife. The microbial sub-model within the USDA Soil and Water Assessment Tool (SWAT) version 2005 was used to characterize bacteria sources and determine how much each of the main five bacteria model parameters affected the model bacteria results. The bacteria parameter that impacted the model bacteria output results greatest was the parameter that reflects the number of bacteria in a defined soil volume that is moved in overland flow. Changes in two of the SWAT model default parameter values were suggested to reflect realistic natural conditions. Results of this study can help researchers in watershed management and modeling decisions choose model parameters respective of model parameter sensitivity.
Technical Abstract: The Soil and Water Assessment Tool (SWAT) version 2005 includes a microbial sub-model to simulate fecal bacteria transport at the watershed scale. The objectives of this study were to demonstrate methods to characterize fecal coliform bacteria (FCB) source loads and to assess the model sensitivity to five user-defined model parameters (BACTKDQ: bacteria soil partition coefficient in surface runoff; TBACT: temperature adjustment factor; WDLPQ: less-persistent bacteria die-off in solution phase; WDLPS: less-persistent bacteria die-off in sorbed phase; and BACTKKDB: bacteria partition coefficient in manure) and one input parameter (BACTLPDB: FCB concentration in manure). Fecal bacterial source loads were described and applied spatially for confined livestock, seasonal grazing livestock, failing human septic systems, and indigenous large mammal, small mammal, and avian wildlife. The relative sensitivity index (S) was tested using the independent parameter perturbation (IPP) method. Validation results for an uncalibrated SWAT model using nine runoff events from Rock Creek watershed (77 km**2) were considered adequate to proceed with sensitivity analyses. Flow simulation resulted in good coefficient of determination (R**2) of 0.67 and Nash-Sutcliffe Efficiency Index (E) of 0.55, and FCB source load characterization methods were sufficiently precise to result in fair correlation (R**2 = 0.29) and reasonable measured vs. predicted response slope (0.69). Within the ranges recommended for use in SWAT, BACTKDQ had moderate sensitivity (S < 2.67) within -99.5% from 175 (baseline value), BACTLPDB had low sensitivity (S < 0.25) within -90% from 3.29 × 10**7 cfu 100 mL-1, BACTKKDB had low sensitivity (S < 0.12) within -89% from 0.9, TBACT had low sensitivity (S < 0.36) +/- 20% from 1.07, WDLPQ had low sensitivity (S < 0.25) +/- 50% from 0.23, and WDLPS had no sensitivity (S less than 0.06) plus/minus 50 percent from 0.023 when compared with all surface runoff events. This study recommends that SWAT could adopt default values of 0.23 for WDLPQ and 0.023 for WDLPS without adversely affecting results. Moderate sensitivity for BACTKDQ indicates that users should select these with caution considering locally relevant data. The sensitivity of BACTKDQ was found high when compared with nine measured surface runoff events.