368
G.A. Schultz
16.3.3Quantitative Precipitation Forecast (QPF)
As mentioned above there are several techniques available for producing a QPF.
Since atmospheric models for medium range weather forecasting have been improved considerably during the last decade it can be expected, that such information, if available in real-time, will represent a very useful QPF to be used as input
into a rainfall-runoff model in addition to the already observed rainfall at the time
of the forecast.
In the example presented here, however, a stochastic QPF procedure has been
applied to the Giinz river catchment. This stochastic approach computes at given
time intervals (e.g. every hour) the depths and duration of the future rainfall until
the end of the storm event. The computed values are given along with prespecified conditional probabilities ofnon-exceedence (e.g. 90 %). The condition is
the quantity of rainfall which occured from the beginning of the storm event until
the time of the forecast. More details of this technique are given in the literature
(Klatt and Schultz, 1985). In Sect. 16.3.4 a flood forecast will be shown along
with the causative rainfall consisting of observed radar rainfall data and QPF
based on the stochastic approach discussed here.
Besides such stochastic QPF approaches also deterministic techniques may be
applied. One such technique uses a time sequence of consecutive radar measurements from which the velocity vector of the moving rainstorm has to be identified.
This process can then be extrapolated thus giving a forecast of the expected rainfall field in the river catchment of interest towards which the rainstorm is moving
(Schilling, 1990). Such forecasts can be further extended if data from groundbased
weather radar are combined with satellite imagery. The British FRONTIERS
System is using this approach (Sargent, 1987).
16.3.4Rainfall-Runoff-Model Application for Flood Forecasting
A real-time forecast of a flood hydrographbased on rainfall information can be
made, if the following items are available:
a rainfall-runoff model with parameters calibrated for the catchment area, for
which the forecast shall be made,
observed precipitation from the beginning of a storm until the time of forecast
with an adequate resolution in time and space,
a quantitative precipitation forecast (QPF) of future rainfall (after the time of
forecast) until the end of the storm event.
The observed plus the forecast precipitation is used as input into a suitable rainfall runoff model in order to generate a flood forecast in real time with an adequate lead time. Figure 16.7 shows such a forecast for the Giinz river based on
radar rainfall measurements and a rainfall forecast computed with the aid of the
stochastic forecasting model mentioned in Sect. 16.3.3.
In this case it is necessary to choose a probability of non-exceedence for the
computed flood. Figure 16.7 shows three examples of flood forecasts computed in
real-time on the basis of probabilities of non-exceedence of 60, 90 and 99 % for
the forecast precipitation. Since such QPF's are usually not very accurate, it can be
G.A. Schultz
16.3.3Quantitative Precipitation Forecast (QPF)
As mentioned above there are several techniques available for producing a QPF.
Since atmospheric models for medium range weather forecasting have been improved considerably during the last decade it can be expected, that such information, if available in real-time, will represent a very useful QPF to be used as input
into a rainfall-runoff model in addition to the already observed rainfall at the time
of the forecast.
In the example presented here, however, a stochastic QPF procedure has been
applied to the Giinz river catchment. This stochastic approach computes at given
time intervals (e.g. every hour) the depths and duration of the future rainfall until
the end of the storm event. The computed values are given along with prespecified conditional probabilities ofnon-exceedence (e.g. 90 %). The condition is
the quantity of rainfall which occured from the beginning of the storm event until
the time of the forecast. More details of this technique are given in the literature
(Klatt and Schultz, 1985). In Sect. 16.3.4 a flood forecast will be shown along
with the causative rainfall consisting of observed radar rainfall data and QPF
based on the stochastic approach discussed here.
Besides such stochastic QPF approaches also deterministic techniques may be
applied. One such technique uses a time sequence of consecutive radar measurements from which the velocity vector of the moving rainstorm has to be identified.
This process can then be extrapolated thus giving a forecast of the expected rainfall field in the river catchment of interest towards which the rainstorm is moving
(Schilling, 1990). Such forecasts can be further extended if data from groundbased
weather radar are combined with satellite imagery. The British FRONTIERS
System is using this approach (Sargent, 1987).
16.3.4Rainfall-Runoff-Model Application for Flood Forecasting
A real-time forecast of a flood hydrographbased on rainfall information can be
made, if the following items are available:
a rainfall-runoff model with parameters calibrated for the catchment area, for
which the forecast shall be made,
observed precipitation from the beginning of a storm until the time of forecast
with an adequate resolution in time and space,
a quantitative precipitation forecast (QPF) of future rainfall (after the time of
forecast) until the end of the storm event.
The observed plus the forecast precipitation is used as input into a suitable rainfall runoff model in order to generate a flood forecast in real time with an adequate lead time. Figure 16.7 shows such a forecast for the Giinz river based on
radar rainfall measurements and a rainfall forecast computed with the aid of the
stochastic forecasting model mentioned in Sect. 16.3.3.
In this case it is necessary to choose a probability of non-exceedence for the
computed flood. Figure 16.7 shows three examples of flood forecasts computed in
real-time on the basis of probabilities of non-exceedence of 60, 90 and 99 % for
the forecast precipitation. Since such QPF's are usually not very accurate, it can be