Striving for the holy grail of accurate weather forecasting

It’s Weatherman’s Day today. Or Weatherperson’s Day, to be more politically correct. Celebrated mainly in the US, today is the day to recognise people working in the field of meteorology and weather forecasting – those people partaking in the massive task of trying to forecast the weather patterns which, let’s face it, are getting more and more crazy and chaotic each year. The date of 5 February was chosen because it is the birthday of John Jeffries, one of the first weather observers in the US – he collected daily weather data from 1774 to 1816.

Weatherperson’s Day links up thematically with World Meteorological Day that we celebrate in May, but where that day focuses on the weather itself, today we focus on those working to forecast it.

Rain! They said it was going to RAIN! (© All Rights Reserved)
Rain! They said it was going to RAIN!
OK, so the weather forecasters don’t always get it right…
(© All Rights Reserved)

From predicting whether it’s going to rain tomorrow, to developing seasonal weather forecasts, weather forecasting is a hugely complex and computationally intensive endeavour. As such, weather services are often some of the main users of supercomputers around the world. Weather forecasting activities include gathering raw weather data, analysing the data and developing intricate computer models to simulate natural weather systems. One of the (many) challenges of weather modelling is that natural weather and atmosphere systems are near-chaotic – small changes in boundary conditions can result in huge changes in outcome.

These difficulties, however, do not deter the good guys and gals at weather services around the globe who continually strive for more accurate and more timely weather forecasting. According to the US National Weather Service, for example, “lead time for flash flood warnings improved from 22 minutes in 1993 to 78 minutes in 2008. Accuracy over the same time period increased from 71 percent to 91 percent. Lead time for tornado warnings has increased from 6 minutes in 1993 to 13 minutes today. Tornado warning accuracy increased from 43 percent to 72 percent. Winter storm accuracy in 2008 was 89 percent with an average lead time of 17 hours. Since 1990, the Tropical Prediction Center’s 24 to 72 hour tropical storm forecast track errors have been reduced by more than 50%.”

These improvements are quite significant, and can potentially be the difference between life and death for communities in the path of an extreme weather event.

So spare a thought today for all the weather people through the ages who have dedicated their working lives to the holy grail of accurate weather forecasting. After the scorchingly hot, dry weather we’ve had for weeks now, they predicted rain for this week, and lo and behold, it has rained! 🙂

John Backus and the development of high-level computer programming languages

Today we’re celebrating the birthday of John Backus (3 Dec 1924 – 28 Oct 1988), American computer scientist and the leader of the team who invented the Fortran programming language (at the time called FORTRAN) while working at IBM in the mid 1950s.

Fortran was the first so-called ‘high-level computer language’, which means it was capable of converting standard mathematical formulas and English-based expressions into binary code used by computers. The language is particularly suited to scientific computing and numeric computation. Over the years, many improvements were made to the original Fortran language, with versions known by a sometimes strange series of numeric identifiers – FORTRAN, FORTRAN II, FORTRAN III, FORTRAN IV, FORTRAN 66, FORTRAN 77, Fortran 90, Fortran 2003 and Fortran 2008.

FORTRAN was the first widely used high-level computer language, providing an interface between equations and expressions understandable to humans,  and binary code used by computers.(© All Rights Reserved)
FORTRAN was the first widely used high-level computer language, providing an interface between equations and expressions understandable to humans, and binary code used by computers.
(© All Rights Reserved)

Despite being one of the oldest computer languages, it has been one of the most enduring, and after more than half a century it is still a preferred language for computationally intensive applications such as weather prediction, computational fluid dynamics and finite element analysis. One of the reasons for Fortran’s longevity is that some of the later Fortran compilers in particular are capable of generating very fast and efficient code, which can make a big difference when solving large, complex mathematical computations. It is still the primary language for used on many supercomputers, and many of the floating-point benchmarks to test the performance of new processors are still written in Fortran.

As a high-level language, Fortran has also provided an impetus for the development of numerous subsequent computer languages such as ALGOL, COBOL and BASIC.

The IEEE awarded John Backus the W.W. McDowell Award in 1967 for the development of FORTRAN. He also received the National Medal of Science in 1975 and the ACM Turing Award in 1977 for his contributions to the design of high-level computer programming systems.