HYDROLOGIC ENSEMBLE PREDICTIONS USING ENSEMBLE METEOROLOGICAL FORECASTS

Abstract

<p>p.p1 {margin: 0.0px 0.0px 0.0px 0.0px; font: 11.5px Times}</p> <p>The objective of this thesis is to investigate the potential of ensemble</p> <p>meteorological forecasts (15 members for each day) in improving ensemble flow</p> <p>prediction up to 14 days ahead. Large scale ensemble meteorological forecasts generated</p> <p>by the National Centers for Environmental Prediction's (NCEP) Global Forecast System</p> <p>(GFS) are used. The hydrologic model used in watershed analysis of the study area is</p> <p>Hydrologiska Byrâns Vattenbalan-avdelning (HBV). The study area is located in the</p> <p>Saguenay-Lac-Saint Jean watershed in northeastern Canada and comprises the Serpent</p> <p>River and Chute-du-Diable basins and a reservoir in Chute-du-Diable.</p> <p>The NCEP ensemble meteorological forecast data is initially used as input in the p.p1 {margin: 0.0px 0.0px 0.0px 0.0px; font: 11.5px Times} span.s1 {font: 7.5px Helvetica}</p> <p>hydrological model HBV to simulate ensemble reservoir inflows and the Serpent River</p> <p>flows for 5 to 14 days ahead. The ensemble inflow and flow forecasts are compared with</p> <p>the case where only observed historical data are used. The study results show that there is</p> <p>a significant improvement in the model forecast performance when NCEP forecast data</p> <p>are used. The improvement for 5 to 14 day forecasts is revealed by an approximately 20%</p> <p>decrease in root mean square error (RMSE) for both reservoir inflow and river flow. A</p> <p>decrease in the Brier score (BS) and rank probability score (RPS) indicates considerable</p> <p>improvement and an increase in the correlation coefficient (r) and the Nash and Sutcliffe</p> <p>coefficient (R²) is shown for reservoir inflow and the Serpent River flow respectively,</p> <p>indicating the advantage of using NCEP data. This improvement is also revealed by the p.p1 {margin: 0.0px 0.0px 0.0px 0.0px; font: 11.5px Times}</p> <p>visual inspection of scatter plots, hydrographs of ensemble mean and ensemble members.</p> <p>The hydrologic forecasts are also assessed on a seasonal basis indicating an improvement</p> <p>in forecasting indicated by a 30% decrease in RMSE during the spring season, and a</p> <p>decrease in BS and RPS values. For other seasons, specifically autumn and summer, the</p> <p>use of the ensemble meteorological forecasts do not provide significant improvement</p> <p>because of the poor skill of predicted precipitation. More accurate predictions of reservoir</p> <p>inflow and river flow with adequate lead time will assist in improving relevant issues in</p> <p>water resources management and planning.</p>