|Di Luzio, Mauro|
Submitted to: Transactions of the ASABE
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 2/11/2011
Publication Date: 11/8/2011
Publication URL: http://handle.nal.usda.gov/10113/54074
Citation: Beeson, P.C., Doraiswamy, P.C., Sadeghi, A.M., Di Luzio, M., Tomer, M.D., Arnold, J.G., Daughtry, C.S. 2011. Treatments of Precipitation Inputs to Hydrologic Models: Guages and NEXRAD. Transactions of the ASABE. 54(6):2011-2020. Interpretive Summary: Hydrological models are used to assess many water resources problems from agricultural use and water quality to engineering issues such as flood or drought prediction potential, erosion rates estimate, contaminant transport, and available water quantity. But these models need good rainfall data, since it is one of the four main components of water balance, as input for them to accurately predict model simulation water budget values. These records can normally come from either land-based rain gauges or remotely sensed radar estimates. Land-based rain gauges only accurately measure a single point on the landscape and require a dense network to capture the spatial variability. In contrast, radar-base records cover wide areas at high resolution but do not always have accurate estimates. Using raw radar-based rainfall estimates as input have resulted in poor model predictions. This study elaborates on a new methodology that combines these two rainfall sources and provides a more reliable rainfall record to drive hydrologic and water quality/quantity models.
Technical Abstract: Hydrological models are used to assess many water resources problems from agricultural use and water quality to engineering issues. The success of these models are dependent on correct parameterization; the most sensitive being the rainfall input time series. These records can come from land-based rain gauges (accurate at a point) or remotely sensed radar estimates (covers a wide area). This study shows how the combination of these two sources can provide a reliable rainfall record to drive the Soil Water Assessment Tool (SWAT) model, which provides decision support for producers and policy makers alike. Results show significant improvement in model predictions by selecting and modifying the rainfall data from three sources- the National Weather Service Land Surface COOP stations and the Next Generation Radar (StageIV and Multisensor Precipitation Estimate). Specifically: 1) when using gauged data the treatment of trace events must be considered- here trace events were treated either as zero, 0.1mm, or as 0.1mm only days before measureable rain events. The latter was found to best describe reality as it fulfills soil moisture deficit the day before measureable rain events resulting in direct runoff and infiltration; 2) if the user does not have rain gauge data, simple correction of the radar estimates to match total rainfall amounts in the region results in adequate results. Here simply multiplying by 15% improved results from radar sources; and 3) if the user has both sources simple kriging with varying local means offered very good results.