BAMOS Vol 31 Special Issue October 2018 Bulletin Vol 31 Special Issue 01 2018 | Página 22

22
BAMOS
Special Issue
Figure 1. HIWeather Conceptual Warning Process: Can we bridge the five valleys of death? Can we
attribute value to each actor? Bridges represent inter-disciplinary and/or inter-agency communication.
Source: Met Office.
1987 and all that!
Brian Golding
UK Met Office
While congratulating AMOS on its 30th birthday, we also
remember the 30th anniversary of a weather event that
transformed weather warnings in the UK. On the night of 16th
October 1987, the worst storm since 1703 hit southern England.
Early forecasts were excellent, but as landfall approached,
predictions of a more southerly track made rain the primary
hazard for England, rather than wind. In the event, several
exposed sites in southern England recorded Storm Force 11
winds, resulting in 18 deaths, loss of 15 million trees, thousands
of people without power, and broadcast media reduced to
skeleton services by loss of transmitting infrastructure. The
insurance cost was estimated at £2bn. Subsequent analysis of
the storm would lead to identification of the sting jet in mid-
latitude depressions. However, more immediately, an inquiry
concluded that observations to the southwest of the UK should
be enhanced, that enhanced computer power was needed to
implement higher resolution Numerical Weather Prediction
(NWP) models, and that the warning service itself should be
reviewed. At that time, “flash” messages were issued to the
public through radio & TV when hazardous weather near urban
centres was already happening or about to happen within 3
hours. These warnings had been correctly issued at 01:30, but
few people were awake to receive them. There were no generic
arrangements for informing emergency managers.
The result of the review was the UK National Severe Weather
Warning Service which has evolved into the format used today
and is widely copied. From the start, it covered the whole
country, was available to emergency responders as well as the
media, and warnings were issued much earlier. Initially, early
warnings, portrayed in map form, were issued when confidence
exceeded 60%, and more specific flash messages up to 24 hours
ahead, when confidence exceeded 80%. The current service
is both probability and impact based, each warning area
characterised by a box in the 4x4 probability vs impact matrix,
with the headline risk highlighted by colour coding. Early “alerts”
at low probability are issued as soon as a threat is identified,
often as much as 5 days ahead, especially when the impact
would be high. Warnings are communicated in a multi-layered
form using colour, risk matrix, map, text and supported by social
media, press releases etc. Development of this service has been
supported by the spectacular advances in NWP since 1987: the
global model grid length has decreased from 150km to 10km
and the regional model from 75km to 1.5km; there have been
huge advances in satellite observing capability; variational data
assimilation and ensemble prediction have been introduced.
Together these have resulted in a 4-day forecast today having
equivalent accuracy to a 1-day forecast in 1987.
Looking ahead, the WMO World Weather Research Programme
has launched the multi-disciplinary HIWeather (High Impact
Weather) project to raise resilience to weather-related hazards,
worldwide, by tackling the main research gaps in all stages
of the warning chain. The new science of km-scale NWP has
major challenges in characterising and reducing error growth,
especially near convective clouds. Such information is required
for data assimilation and ensemble prediction as well as in the
development of improved parametrisations. While warnings are
now impact-based in many countries, automated guidance to
support them remains limited. Turning a weather forecast into
an impact forecast requires, first, that the hazard be predicted,
whether it be flash flood, wildfire, heat wave, snow, etc, and
second that the human impact be assessed, requiring knowledge
of the exposed population and their vulnerabilities. Finally, the
warning must be communicated through media and in formats
that enable recipients to make good decisions. In addressing
these challenges, HIWeather has developed a concept of the
warning process as a value chain in which the expert links are
separated by five “valleys of death” over which bridges need to
be built. To optimise the process, it is necessary to understand
the sensitivity of the eventual decisions made by recipients
of the warning to the contribution of each actor in the chain,
through appropriate evaluation processes. HIWeather is using
this concept as a unifying principle in Forecast Demonstration
Projects over the next few years, both to evaluate new capability
and to build capacity in developing countries.