Capital Press

PROSSER, Wash. -- AgWeatherNet, Washington State University's network of automated weather stations, is testing a national, state-of-the-art forecasting model to better predict adverse weather for growers, particularly in areas of tree fruit orchards.

The system might be available online in one to three years depending on continuous funding, said Gerrit Hoogenboom, the network's director and WSU professor of agrometeorology.

WSU is requesting three more years of research funding from the Washington Tree Fruit Research Commission, Hoogenboom said.

AgWeatherNet provides current weather conditions and historical data from 137 stations, mainly in irrigated portions of Central Washington.

Adding weather predictions by station and larger areas will allow growers to make better decisions, Hoogenboom said.

The system will be used to warn of frosts, freezes, hail and, with inputs from other models, can help in pest and disease control decisions, he said.

AgWeatherNet gives monthly weather summaries, weekly weather outlooks and warnings based on data from all its stations, but does not predict air temperature, dew point temperature, wind speed or wind direction, Hoogenboom said.

Accurate predictions are a challenge, given Central Washington's varied terrain, he said.

The new Weather Research and Forecasting (WRF) model helps with that because it is a next-generation "mesoscale numerical weather prediction system" designed to serve both operational forecasting and atmospheric research needs.

Mesoscale is the study of weather systems ranging from a few miles to hundreds of miles wide. It can be applied for storm-scale research and prediction, air quality modeling, wildfire simulation, hurricane and tropical storm prediction, regional climate prediction and more.

"It can provide us with hourly weather predictions at a very high spatial resolution -- close to a 2-by-2-mile grid," Hoogenboom said.

WRF was tested during freeze and frost events last Oct. 25-27, Feb. 24-27 and April 7.

Results were processed on a high-performance computer purchased with part of $95,000 in funding from the research commission.

"The model performed well for low elevations in Eastern Washington and providing 24- to 48-hour predictions," Hoogenboom said. "Longer out was more difficult. We are trying to develop a parameter scheme so it will do better."

The next step is testing the model for real time weather predictions by storing results and comparing them with actual AgWeatherNet observations, he said.

Recommended for you