http://www.stuff.co.nz/stuff/thepress/0 ... 30,00.htmlForecasters review system after missing storm
06 November 2006
By PAUL GORMAN
The MetService is taking another look at how it forecasts heavy snow after it failed to predict June's crippling Canterbury snowstorm.
The June 12 weather bomb brought up to a metre of snow to parts of Mid-Canterbury, caused an estimated $100 million of damage to farms and the economy, and left thousands of Cantabrians without electricity in the subsequent cold snap, some for as long as three weeks.
In some places the storm brought snow even deeper than the great July 1945 fall.
The MetService came under immediate fire for failing to warn the public of heavy snow to sea level until there was half a metre of it lying across parts of the plains.
Now, the state-owned enterprise has reviewed its efforts as it said it would when it admitted it could have done better, even though it also said then it was possible nothing could be learnt from what might have been a "freak event".
National forecast centre manager, Peter Kreft, told The Press the service was planning some changes with the way it forecast snow in Canterbury.
They included installing a new computer calculation for predicting snow in the kind of "over-running situations" that produced the June storm.
It would also have new methods to look at temperature changes with increasing height above the ground, and tweaking computer models used by forecasters in making predictions.
"Over-running" was where a relatively warm airstream moved over cold air closer to the surface, and was the culprit for most of the heavy snow along the South Island's east coast.
The service was also going to investigate a special system next year that would help it predict extreme weather events, Kreft said.
"In terms of whether an event like this could be forecast better next time, it pays to remember that it was both an extreme and a rare event. By definition, such events ... are difficult to identify," he said.
"Forecast systems which produce successful predictions of rare events also produce lots of false alarms. Forecasts of severe weather which turn out to be false alarms are often not well-received by users – it generally costs them in some way."
In the two years to the end of September the MetService's detection rate for heavy snow had been 92 per cent, with a 22% false alarm ratio, Kreft said.
A post mortem of the storm to see if clues had been missed revealed that computer models had not accurately forecast precipitation amounts and freezing levels, indicating that temperatures close to sea-level would remain about 6deg.
With that temperature, forecasters had not been "unduly concerned at first" about snow reaching sea-level, he said.
Interesting they are calling it a 'freak event' even though Blueskies put out a better forecast prior to the event and the fact that local farmers before the storm could tell the forecasts were wrong. I guess local knowledge held the upper hand here.