Skip to main content
ARS Home » Northeast Area » University Park, Pennsylvania » Pasture Systems & Watershed Management Research » Research » Publications at this Location » Publication #205825
Title: Modeling a Small, Northeastern Watershed with Detailed Field-Level Data

Author
item Veith, Tameria - Tamie
item SHARPLEY, ANDREW - UNIV OF ARKANSAS
item Arnold, Jeffrey

Submitted to: Transactions of the ASABE
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/10/2008
Publication Date: 4/1/2008
Citation: Veith, T.L., Sharpley, A.N., Arnold, J.G. 2008. Modeling a Small, Northeastern Watershed with Detailed Field-Level Data. Transactions of the ASABE. 51(2):471-483.

Interpretive Summary: Agricultural watersheds are hard to measure and model because of the many environmental processes which singly and jointly impact downstream water quality. This study tested how well the Soil and Water Assessment Tool water quality model predicted runoff, erosion, and phosphorus loss from a northeastern Pennsylvania watershed using very detailed weather, slope, soil, and land management data collected by workers at the watershed location. Even with the careful use of detailed data, the model results differed by measured results by 41% in 1997-2000 and 15% in 2001-2004. Testing how well the model results matched the measured results helped identify how well the model simulates the watershed processes and helps to better explain the abilities of different management practices to improve water quality of the watershed.

Technical Abstract: As watershed models are increasingly used to forecast effects of changed management practices on downstream water quality, knowledge about sources and levels of uncertainty within models have become more important. One source of uncertainty in water quality modeling comes from input data. Watershed FD-36, a 39.5-ha agricultural area in northeastern PA, was modeled from 1997-2004, using detailed land management data at the field-level, daily weather data, and 5-meter soil and topographic data. Monthly simulated runoff and dissolved phosphorus loads were generally within one standard deviation of corresponding measured values. Monthly dissolved and particulate phosphorus losses followed the same trends as monthly sediment losses. Monthly biases in simulated and observed stream depth data correlated positively with monthly biases in simulated and observed sediment loss data. Nash-Sutcliffe and R2 statistics were about 0.40 and 0.55, respectively, for flow volume irrespective of calibration or validation period. However, percent bias for flow volume was 41% during the calibration period and 15% during the validation period. All three statistics reported higher values for sediment loss and concentration for the calibration period than the validation period, except for the Nash-Sutcliffe which changed from -1.79 for the calibration period to -.063 for the validation period. These results indicate the impact of single extreme storms on statistical calculations. Additionally, while detailed spatial and temporal data improve modeling of a natural system, they do not remove all sources of uncertainty. This emphasizes the importance of using models as a relative, as opposed to absolute, comparison tool when forecasting water quality effects of differences in watershed, farm, or field management.