Geoscience Reference
In-Depth Information
Carpenter and Georgakakos (2006) investigated the combined effects of radar rainfall errors
and catchment size on the uncertainty in predicted river flow with the aid of a distributed
hydrological model. Using a parsimonious model to represent the spatial structure and
variance of radar errors, they demonstrated that ensemble spread in predicted flow was log-
linearly related to catchment scale.
A handful of ensemble-based probabilistic QPN schemes have been developed during the
past decade or so. These were described earlier in this Chapter and include the String of
Beads Model (Pegram & Clothier, 2001; Berenguer et al., 2011), the Short-Term Ensemble
Prediction System (Bowler et al., 2006) and a method recently described by Kober et al.
(2011). Here we draw a distinction between ensemble QPN schemes and others such as
those described by Andersson and Iverson (1991) and Germann and Zawadzki (2004) which
produce forecasts of the probability distribution of precipitation at a point. The latter cannot
be used for ensemble-based probabilistic hydrological forecasting because they do not
provide a complete description of the joint probability distribution of precipitation, which
plays a key role in the hydrological response of a catchment.
In the UK, the Department of the Environment, Fisheries and Rural Affairs and the
Environment Agency recently funded an R&D project to explore the benefits of high
resolution precipitation forecasts to fluvial flood prediction and warning (Schellekens et al.,
2010). The potential for operational use of ensemble rainfall nowcasts from STEPS (Bowler et
al., 2006) was investigated in conjunction with lumped and distributed rainfall-run-off
models. The evaluation included hydrological configuration issues, data volumes, run times
and options for displaying probabilistic forecasts. No quantitative verification of the
precipitation ensemble-driven hydrological forecasts was undertaken.
Recently, the Environment Agency has implemented a nationally configured, distributed
hydrological model, known as Grid-to-Grid (Bell et al., 2007). In the near future (2012), this
model will be driven by ensemble precipitation forecasts integrating STEPS ensemble nowcasts.
7. The future of nowcasting
One of the major changes in the past decade has been the increase in the ability of the general
population in developed countries to receive real-time information over a range of mobile
platforms. This makes it possible to deliver location specific nowcasts to millions of users.
They use this information to make routine decisions regarding leisure and other outdoor
activities and very occasionally decisions relating to severe weather events. Mitigating damage
due to severe weather has been the motivation for developing nowcasting systems in the past,
but the focus is likely to change to providing routine nowcasting services to the general public.
The current generation of nowcasting systems are already capable of delivering products that
are useful in this context and the focus in the short term should be on developing the ability to
customize and disseminate these to a very large number of users.
Improved communications and computer capacity have also made it possible to routinely
combine data from a network of weather radars into a single large domain, and to improve
the algorithms that are used to provide quantitative radar rainfall estimates. Improvements
in the QPE will continue as the radar hardware improves and the density of the radar
networks increases. These improvements will allow for improvements in the quality of the
nowcasts in the first hour.
