Improving the robustness of uncertainty algorithms in quantification of uncertainty in water balance forecasting

2014 
Recent studies have highlighted the potential challenges in US southeastern (SE) watersheds from climate variability. There may be shifts in water balance due to complexity of the flow generation processes that determine how water is partitioned in these landscapes. The main objective of this study was to capture the feedback relationships among the water balance components using the Soil & Water Assessment Tool (SWAT) watershed-scale streamflow model linked with the Sequential Uncertainty Fitting (SUFI-2) and Particle Swarm Optimization (PSO) parameter uncertainty algorithms in the Waccamaw River watershed, a low-gradient forested watershed on the coastal plain of the southeastern United States. Water balance uncertainty analysis suggested close correspondence of the model with the physical behavior and system dynamics during different hydroclimatological periods in the 2003-2007 calibration interval. SUFI-2 water balance analysis revealed that surface runoff, ground water, and lateral flow contributed 22.2%, 3.9% and 0.4% of the total water yield during simulation period while PSO analysis indicated 16.7%, 13.2% and 0.3% of their contributions respectively. Both uncertainty methods found that 71.1% of the total rainfall was lost by evapotranspiration during the simulation interval. The total water yields using both algorithms were over predicted by up to 14.0% of the annual rainfall inputs during the dry period (2007); this was related to the extra contribution of shallow aquifer flow to the river system. Both algorithms also specified that surface flow and ground water runoff dominated the water balance during October and December respectively. The distribution of predictive uncertainty was least in the
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    7
    References
    1
    Citations
    NaN
    KQI
    []