 |
“The flexibility provided by the IMSL Library has permitted
us to build a unique and flexible post-processing system, allowing
us to compete head-to-head in the global weather services market.”
-Brian Peters, Senior Development Meteorologist,
Metra
QUICK
FACTS
| > |
Weather service developing statistical post-processing system from scratch to offer more
accurate forecasting
|
| > |
Regression, covariance matrices, Kalman routines and optimization used from the IMSL
Fortran Numerical Library running on Linux
|
| > |
More accurate forecasts have allowed New Zealand’s Metra to compete in global weather
services market
|
THE PROBLEM
In the past 10 years, the business of numerical weather prediction
has seen some dramatic change. Weather forecasting using computers
used to be the sole purview of large national weather services,
but lower-cost, high-performance computing changed all that.
“The basic concept behind numerical weather prediction has been around
for decades. Up until the 1990's, however, computing power was
limited, so the process was chiefly done by large national weather
services that had access to supercomputers. In the past decade
there has been rapid growth in the number of smaller national
weather services and private weather companies applying advanced
forecasting techniques, due to the accessibility of low-cost,
high-power
computing,” says Brian Peters of Metra, a New Zealand-based weather
services company.
Metra of New Zealand is one such company that has capitalized
on low-cost computing. Brian Peters of Metra was tasked with
building more robust Local Area Models (LAMs) for New Zealand
by using statistical post-processing. Mr. Peters needed help
with developing
applications that were flexible and cost-effective but could
also help him exploit unique ideas in weather forecasting.
Metra runs a 24
hour a day, 7 day a week operation, so the application had to
be robust, fast, accurate and easy to integrate with their existing
systems. “During
the design process I looked at many different commercially available
statistical analysis packages but found that most of them were
too "stand-alone" and did not mesh easily with
our operational environment, required too much CPU time and memory,
were not flexible enough for what we wanted to achieve, did not
have particular unique applications that we required, or were
just too expensive,” Mr. Peters says. The IMSL Fortran
Numerical Library met Mr. Peters’ list of requirements
and enabled his team to concentrate on the science, rather than
the coding.
The Solution
Mr. Peters uses the IMSL Fortran Library on Linux as part of the
post-processing stage for developing the proprietary local area
models (LAMs). Statistical post-processing gives the LAM the additional
intelligence needed to generate superior forecasts. Metra runs
the LAM
on other machines as stand-alone processes, then refines the raw
LAM output using code containing the IMSL Library algorithms to
produce higher quality forecasts.
Metra uses several statistics algorithms in the IMSL Library, including
regression, covariance matrices, Kalman and optimization routines.
Four times a day in 6 hour blocks, post-processing of the LAM is
performed concurrently on separate machines. The code utilizing
the IMSL Library runs on desktop PCs. The applications process
10’s of Gigabytes of data per day.
Metra has been very successful selling their
forecasts globally to energy companies, who need forecasts for
the next day or two, and stock exchange traders, who are interested
in 2 week forecasts. Mr. Peters has been very pleased with the
IMSL Library and technical
support: ”We have been using the IMSL Library for five years
and have had a very positive experience. Technical support has
been helpful and the software has been problem free.”
RETURN ON INVESTMENT
Using the IMSL Library has equipped Metra to offer more accurate
forecasts that are tailored to the customer’s needs, distinguishing
them from their competitors: “The
flexibility provided by the IMSL Library has permitted us to
build a unique and flexible post-processing system allowing
us to compete head-to-head in the global weather services market,” says
Mr. Peters. Internally, the meteorologists writing the forecasts
for the people of New Zealand benefit from
having more accurate Numerical Weather Prediction (NWP) guidance.
Better guidance results in more accurate forecasts that benefit
the entire nation - public, commercial, transport, agriculture,
marine, aviation, energy, etc.
Better, more accurate, weather forecasts aid in planning and in
making critical decisions about the future. Whether it is the farmer
planning to deal with a potentially damaging frost, a power plant
operator trying to schedule maintenance, a fisherman planning his
next trip, or a 747 pilot landing after flying across the Pacific
Ocean, accurate weather forecasts assist in safety and commercial
decisions throughout the nation.
|
|
|
|
Industry
Weather Forecasting
Application
Statistical
post-processing
Product
IMSL Fortran Library
Metra provides weather and information presentation services
to customers around the world from its base in New Zealand. It
has
particular expertise in supporting the media, aviation, and energy
industries, and in national meteorological service operation.
Key Benefits
| > |
Superior forecasting |
|
| > |
Lower CPU and memory usage |
|
| > |
Flexibility for innovative
development |
|
| |
|
| |
|
|
| |
|
|
| |
|
| |
| |
|
| |
|
