BAMOS Vol 31 Special Issue October 2018 Bulletin Vol 31 Special Issue 01 2018 | Page 17

BAMOS
Special Issue
The current revolution in Numerical
Weather Prediction
Peter Steinle
Bureau of Meteorology
Numerical Weather Prediction (NWP) has underpinned much
of the improvements in forecast accuracy over the past few
decades. This has been achieved through advances in the
sophistication of numerical models, the number and quality
of observations and the way the observations are used to start
the model forecasts. Operational NWP systems are generally
maintained and developed by large teams of scientists, a
situation that goes back for several decades. It is therefore easy
to see this as a very mature field, with limited opportunities for
major developments or, for an individual or small team to make
a significant contribution. However, the continuing changes
in technology and society mean that NWP will remain a very
dynamic area offering many opportunities for new approaches
to be developed involving a wide range of contributors.
One of the major changes is the amount of observations
available from both inside and outside the meteorological
community. New satellites and radars will continue to provide
increasing information to measure the atmosphere. There are
also other technologies such as lidars that can provide even
higher density information. The other rapidly growing area are
observations from outside what may be considered “the official”
meteorological community. This includes personal weather
stations and the like, which clearly contain valuable information,
but also have quite different characteristics to standard
measurement sites, posing a challenge to existing systems. The
largest growth though may well be in indirect observations
such as air pressure from mobile phones, temperature from cars
and a host of indirect measurements of rainfall such as changes
in signal strength between mobile phone towers, and the signal
received at NBN ground stations all the way to information on
the speed of car windscreen wipers. Such observations are of
course quite different to what NWP centres are used to and so
accessing, processing and using this data poses a number of
interesting problems.
As supercomputers have grown larger and larger, the cost of the
electrical power needed to drive them is becoming prohibitive.
This is forcing vendors to look at radically different hardware
designs, including the use of very large numbers of energy-
efficient chips, such as those found in video game hardware.
This change in hardware is expected to provide a major jump
in the computational power that modelling centres can afford,
raising the prospect of much more detailed forecasts being
updated much more frequently—so not only using more
observed data, but also converting that to forecast information
much more quickly. This in turn requires a better transformation
from meteorological variables to quantities that are used in
decision making, i.e. impact forecasting.
This change in computing architecture has many consequences
for all computationally intensive applications, providing many
more research challenges—requiring not only a redesign of
models, but changing the underlying coding paradigm to
enable greater flexibility in optimising numerical systems. Such
changes can also mean that the efficiency of the underlying
algorithms within the modelling system need to change—
opening up a host of new problems across mathematics and
the physical sciences. The problems posed by these changes in
NWP systems are not only mathematical and/or computational,
but the storage and communication of the data also needs to
be reconsidered.
The combination of the above changes introduces the
feasibility of providing urban scale NWP. The value of
introducing models with resolutions around 100m is becoming
increasingly apparent. As cities grow, energy usage, production
and distribution patterns change and need to be managed
across all time scales, starting at minutes. This also includes
efficiently managing the heat and air quality around areas of
high population density—and not only the direct impact on
the energy efficiency of buildings, but also the management of
green spaces. The future needs of aviation are also much greater
than present in order to meet increased safety and efficiency
standards. As cities grow their interaction with the atmosphere
changes, as does the population’s response and so emergency
and health services will also require more accurate and timely
information.
In summary, in an increasingly technologically advanced
society there is an increasing need for highly detailed, highly
accurate weather forecasts. In turn the ability of technology to
support these forecasts means that NWP will remain a rapidly
developing field for some time to come. Furthermore, the NWP
developments will require access to increasingly more areas
of expertise. As a result, NWP remains a very active area of
development, providing a broad scope of research topics from
an increasing number of disciplines, allowing a wide range of
research groups to make valuable contributions.
17