MetService spendup to improve accuracy

MetService is promising more accurate forecasts this year.

The state-owned enterprise is spending hundreds of thousands of dollars improving computer models that guide forecasters in their predictions.

Forecasts are generally between 80 per cent and 85 per cent accurate in New Zealand, a difficult country for weather prediction.

However, MetService wants to lift that rate.

National weather services general manager Norm Henry said the three-dimensional weather research and forecasting (WRF) model currently used by forecasters had 8km by 8km grid squares with weather data at each corner.

The aim was to reduce the grid size, increasing the amount of data and providing more precise forecasts.

"It takes all the observations we have, as well as older model data, and does its best possible interpolation," he said. "The plan is to implement a 4km grid, which is more complex, as cutting the grid spacing in half requires around eight times as much computing power."

Computer upgrades costing hundreds of thousands of dollars were scheduled to be in place by the end of the year, Henry said.

Global computer models gave MetService forecasters a "broad-scale picture of the weather, where the main fronts and highs and lows are, and how they're moving".

"But by doing local modelling, we're effectively down-scaling that. What happens when you down-scale is you get a much better representation of the local terrain.

"It goes without saying the geography of New Zealand is complex, the coastline and the terrain, and that drives a lot of local-scale effects – sea breezes, the height of snow in a winter storm in Canterbury, the amount of rain in the catchment areas around the southern lakes."

Over the past 10 to 15 years there had been a "steady increase in ability" to predict weather more accurately and further ahead, he said.

"I'd be pretty cautious with any forecast beyond five to seven days, but that leads into another interesting aspect of this. We'll never rely on a single model, but instead we try to look at a range of possible model solutions.

"Forecasters spend a lot of time comparing them to each other and assessing how they are performing compared with current observations. Otherwise you can get badly led astray. There's this notion that you don't want to put your eggs in one basket."

- © Fairfax NZ News