Environmental Engineering Reference
In-Depth Information
sent the rainfall pattern associated with strongly
convective systems (see Chapter 7).
The prediction of quantitative rainfall amounts
is even more problematic, essentially due to these
same non-linear processes. For the time domain of
a few hours ahead, the extrapolation of radar or
satellite imagery yields results that can be opera-
tionally useful. This process is generally known as
'nowcasting'. For longer time horizons, forecasts
generated by solving the underlying dynamical
equations need to be used. These models may be
classified into 'global' and 'meso', depending on
the scale at which they operate. All NWP models
suffer from the difficulty that they too must at-
tempt to deal with the chaotic behaviour of the
real atmosphere, with rainfall probably being the
most difficult parameter of the atmosphere to
predict. The mesoscale models probably have a
more difficult time since they are attempting to
model very small-scale systems, which could well
be convective in nature.
It is very important to be realistic about the
difficulties NWP is going to have in representing
small-scale precipitation events, such as that il-
lustrated inFigure 10.3.Weather radar imagery of a
small band of thunderstorms in New Zealand is
advantages of both approaches ismaximized, offers
the hydrologist an improved approach to develop-
ing reliable short- and long-range forecasts within
the same model input dataset. Developing an en-
semble of precipitation forecasts enables the de-
velopment of a probability distribution of the
magnitude of the predicted flood, which in turn is
a useful tool for thosewho respond to emergencies.
Meteorological Discussion
The Earth's atmosphere is characterized by fields
of temperature and pressure that are relatively
smooth in space and time when compared with
the distribution of water, particularly water stored
as cloud and rain. This is primarily because clouds
and rain are produced by very non-linear processes
involving condensation. The measurement of ar-
eal rainfall is therefore a difficult problem in view
of the very high spatial heterogeneity of the rain-
fall pattern. The density of raingauges required to
adequately represent the rainfall pattern is thus
very high and depends on the meteorological pro-
cesses that are giving rise to the rain. Radar remote
sensing may be the only way to adequately repre-
Fig. 10.3
Radar imagery of shower passing
over a small catchment (outline solid
black). The raingauge network was unable
to sample spatial variability of the
hydrometeor in this event. The radar is
located at 0,0 (
D
). Terrain contours and
roads have also been shown. (See the
colour version of this figure in Colour
Plate section.)
