16 Flood Forecasting and Control
361
tation data into a forecast runoff hydro graph. Distributed models are desirable,
since the spatial distribution of precipitation varies very much in space. Since this
variability has a major influence on the shape of a flood hydro graph it is necessary
to model the rainfall-runoff process in such a way, that it can handle the high
spatial variability of pre.cipitation, i.e. the model structure must be such, that it can
represent spatial variability. Such models usually require knowledge of the area
distribution of soil water storage capacity within the catchment area, within subcatchments or within so-called Hydrologically Similar Units - depending on the
specific model structure. Colour Plate 16.A shows the area distribution of soil
water storage capacity in the catchment of the Priim river in Gennany. The map
on the left side of Colour Plate 16.A shows soil porosity derived from a soil map.
The map in the center shows root depths derived from Landuse classification,
which is based on the infonnation of Landsat-TM imagery and the map on the
right side shows the soil water storage capacity of the Nims catchment distributed
in space which is generated as product of root depth times soil porosity. This way
a combination of remote sensing data with data from a digitized map can be
merged in order to produce the new infonnation in fonn of a soil water storage
capacity map, which in turn fonns the basis for a special set of model parameters
of the model discussed in Sect. 16.2.4.
Model Input. Input into rainfall-runoff models is either precipitation or snowmelt
or both. Rainfall data with high resolution in time and space can be acquired in
real-time by groundbased weather radar. Colour Plate 16.B shows rainfall obtained with the aid of weather radar over a city in Gennany which demonstrates
the high variability of precipitation intensity in space. The transfonnation of the
echo of the radar signal into rainfall intensity is discussed in Chap. 6 of this book.
Another way of using remote sensing data for rainfall estimation uses satellite
data, usually from geostationary satellites. Also this technique is briefly discussed
in Chap. 6 of this book. If snowmelt is relevant, remote sensing data can be used
in order to determine snowmelt runoff as discussed in Chap. 11 of this book.
Precipitation Forecasting. As discussed in Sect. 16.2.1 and shown in Fig. 16.2 it
is necessary for using flood forecasts effectively to have a long lead time of the
forecast, which requires not only observed precipitation, but also a precipitation
forecast in real-time at the time of a flood forecast. Such quantitative precipitation
forecasts (QPF) can be generated in several different ways. In Sect. 16.3.2 a
method of flood forecasting is presented, which uses stochastic QPF's.
In many countries, e.g. allover Western and Central Europe the national Met
offices produce routinely QPF's distributed in area over a time span up to 72
hours. These QPF's may be used together with observed precipitation for a catchment area as input into a rainfall-runoff model for the purpose of real-time flood
forecasting.
A third technique of QPF is based on real-time precipitation measurements by
weather radar. From a sequence of observed radar images prior to the time of the
forecast it is usually possible to derive a vector indicating direction and speed of
the movement of a raincell with observed precipitation intensity. The forecast of
361
tation data into a forecast runoff hydro graph. Distributed models are desirable,
since the spatial distribution of precipitation varies very much in space. Since this
variability has a major influence on the shape of a flood hydro graph it is necessary
to model the rainfall-runoff process in such a way, that it can handle the high
spatial variability of pre.cipitation, i.e. the model structure must be such, that it can
represent spatial variability. Such models usually require knowledge of the area
distribution of soil water storage capacity within the catchment area, within subcatchments or within so-called Hydrologically Similar Units - depending on the
specific model structure. Colour Plate 16.A shows the area distribution of soil
water storage capacity in the catchment of the Priim river in Gennany. The map
on the left side of Colour Plate 16.A shows soil porosity derived from a soil map.
The map in the center shows root depths derived from Landuse classification,
which is based on the infonnation of Landsat-TM imagery and the map on the
right side shows the soil water storage capacity of the Nims catchment distributed
in space which is generated as product of root depth times soil porosity. This way
a combination of remote sensing data with data from a digitized map can be
merged in order to produce the new infonnation in fonn of a soil water storage
capacity map, which in turn fonns the basis for a special set of model parameters
of the model discussed in Sect. 16.2.4.
Model Input. Input into rainfall-runoff models is either precipitation or snowmelt
or both. Rainfall data with high resolution in time and space can be acquired in
real-time by groundbased weather radar. Colour Plate 16.B shows rainfall obtained with the aid of weather radar over a city in Gennany which demonstrates
the high variability of precipitation intensity in space. The transfonnation of the
echo of the radar signal into rainfall intensity is discussed in Chap. 6 of this book.
Another way of using remote sensing data for rainfall estimation uses satellite
data, usually from geostationary satellites. Also this technique is briefly discussed
in Chap. 6 of this book. If snowmelt is relevant, remote sensing data can be used
in order to determine snowmelt runoff as discussed in Chap. 11 of this book.
Precipitation Forecasting. As discussed in Sect. 16.2.1 and shown in Fig. 16.2 it
is necessary for using flood forecasts effectively to have a long lead time of the
forecast, which requires not only observed precipitation, but also a precipitation
forecast in real-time at the time of a flood forecast. Such quantitative precipitation
forecasts (QPF) can be generated in several different ways. In Sect. 16.3.2 a
method of flood forecasting is presented, which uses stochastic QPF's.
In many countries, e.g. allover Western and Central Europe the national Met
offices produce routinely QPF's distributed in area over a time span up to 72
hours. These QPF's may be used together with observed precipitation for a catchment area as input into a rainfall-runoff model for the purpose of real-time flood
forecasting.
A third technique of QPF is based on real-time precipitation measurements by
weather radar. From a sequence of observed radar images prior to the time of the
forecast it is usually possible to derive a vector indicating direction and speed of
the movement of a raincell with observed precipitation intensity. The forecast of