Hydrology and flow forecasting

Hydrology and flow forecasting

Pang, H.
Rao, S.Q.
Wang, G.Q.
Huang, X.Q.

Vrijling, J.K.
Kwadijk, J.
Van Duivendijk, J.
Van Gelder, P.

Civil Engineering and Geosciences

Hydraulic Engineering

2002-06-05

We have studied and applied the statistic model (i.e. MMC) and hydrological models to Upper Yellow River. This report introduces the results and some conclusions from the model. The three models, MMC, MWBM and NAM, have be applied in the research area. The forecasted discharge by the three models are closed to the recorded in most low flow months, but in some flood months, the forecasted value is much different with recorded one. And forecasted accuracy by MWBM nearly matches with that by NAM model. Statistical results indicate that the three models have big forecasted errors in 1989, average relative errors in the year are all excess 60%. In the other four years, MWBM and NAM models have similar forecasted errors; average relative forecasting errors fall in the range from 15% to 55%. For MMC method, average relative forecasted error in 1987 is in smallest value of 23.5% and that in 1990 is in highest value of 102.8%). Forecasted errors in the first two years are relative lower than that in the rest three years, it has increasing trendy. If acceptable forecasted result is that, relative error is less than 20% or absolute error is less than 4mm, then acceptable percentage of forecasted discharge was also calculated. Result shows that most forecasted value by MWBM and NAM is qualified. For MMC method, only in the first two years and low flow period of other years, most forecasted value is acceptable. Forecasted results by MWBM and NAM are all based on given historical meteorological data. But in real time flow forecasting, forecasting accuracy is mainly up to two factors, one is the hydrological model accuracy and the other is the meteorological forecasting accuracy. And conclusions obtained mainly reflect the first factor. In addition, in flow forecasting with the statistic method, (i.e. the MMC method), long series are mostly needed. However, 40 years is a reasonable long period of data. The forecasted error in this case is mainly due to not using the rainfall data. All of results denoted that the three models can be used to forecast discharge in the next one or two years. If meteorological data in the future can be forecasted for a longer period and are of high quality, MWBM and NAM should be adopted first.

http://resolver.tudelft.nl/uuid:36c9770b-3131-46c4-8d68-913314dfef2f

Delft Cluster

Institutional Repository

report

© 2002 Delft Cluster
Hohai University
YRCC