Big data key to bringing hyperlocal weather forecasts to Georgia farmers

Big data key to bringing hyperlocal weather forecasts to Georgia farmers

Researchers say tools would help Flint River Vally farmers conserve water and increase long-term crop yield

Farmers in southwest Georgia's Flint River Valley could one day get accurate, hyperlocal weather forecasts just for their individual farms up to three days in advance.

A team of researchers from the Flint River Soil and Water Conservation District, the U.S. Department of Agriculture, the University of Georgia and IBM is using sophisticated big data tools to analyze large volumes of meteorological, geographical, historical and other data. The goal is to model weather behavior with a higher degree of accuracy and localization than can be done today.

Such forecasting would help farmers make more informed irrigation, seeding, harvesting and fertilizer scheduling decisions, which in turn would enable them to conserve water and increase long-term crop yield.

"With data and data-driven solutions, we are looking at the next generation of agriculture," says David Reckford, director of The Flint River Partnership project. "We are beginning to apply a level of science to the system that will allow us to grow more with less."

The Flint River Valley is an important part of Georgia's agricultural industry. Farms in the 27-county region contribute roughly $2 billion annually in farm-based revenue.

About 10 years ago, the region's water conservation authority teamed up with the U.S Department of Agriculture, University of Georgia, and other local, regional and state-level agencies to promote water conservation practices among farmers in the area. The effort has already paid dividends.

A so-called Variable Rate Irrigation (VRI) technology developed by researchers at the University of Georgia moved quickly from concept to commercial product as the result of the work done by the Flint River Soil and Water Conservation District and the same groups involved in the data analytics effort. The GPS-based technology allows farmers to set irrigation sprinklers so that the nozzles turn water off over areas that don't need water and turn them back on over areas that need irrigation.

Reckford is confident that hyperlocal forecasts enabled by big data analytics and sophisticated modeling technologies will one day yield similar benefits.

Weather forecast models in the U.S. typically have a horizontal resolution of 12 kilometers, meaning they are based on data gathered from grid points spaced 12 kilometers apart. By building weather models with a 1.5 kilometer resolution, the Flint River Partnership project is looking to provide farmers in the area with much more granular weather information.

That would lead to more informed decisions regarding irrigation, seeding, harvesting and fertilizer application, Reckford said.

Central to the effort is IBM's Deep Thunder short-term forecasting and customized weather modeling technology. The technology, which is built around a parallel processing supercomputer, is deigned to help organizations forecast weather down to a square kilometer -- and even smaller -- geographic area.

Deep Thunder was created jointly by IBM and the National Oceanic and Atmospheric Administration (NOAA) as part of a project launched in 1996. After years in IBM's research labs, the technology is now sold to utility companies, city governments and others as a cloud-based offering that's used for precision weather forecasting.

The Flint River project will ingest and crunch data from numerous sources, including NOAA, several weather stations set up in the region by the University of Georgia, and the Earth Network WeatherBug system. The system will also evaluate land use, vegetation, topography and other geographic data from the U.S. Geological Survey and NASA.

Researchers from the Flint River Soil and Water Conservation District, the Department of Agriculture, the University of Georgia and IBM are using sophisticated big data analytics to produce detailed weather forecasts.

"The amount of data that goes into each forecast is many tens of gigabytes," said Lloyd Treinish, chief scientist of IBM's Deep Thunder project. But once the extraction and filtering and quality control work is done, the amount of data that to be analyzed is reduced by an order of magnitude, he said.

A full 72-hour forecast will be about 320 gigabytes but what gets disseminated to the farmers is much less.

Farmers will be able to track weather conditions that apply specifically to them, in 10-minute increments, up to 72 hours in advance, Trennish said,

Using a desktop or mobile browser, farmers will be able to view site-specific forecasts from IBM's Deep Thunder weather forecasting portal. Some of the forecasts will be available as high-definition videos while others will be in the form of detailed two-dimensional animations of rainfall patterns, cloud movements and soil moisture evolutions, Treinish said.

Farmers will also be able to view the information numerically in spreadsheet form.

Farmers will be able to track thunderstorms, temperatures and wind speed variances for their specific locations during different times of the day. They will know with more certainty if a rain system will produce an eighth or a quarter inch of water, or if the wind speeds would prohibit chemical applications.

Some day soon, Reckford wants the forecasts pushed directly to farmers in the field. Instead of having farmers visit the Deep Thunder portal to access forecasts, Reckford wants them to be able to receive forecasts directly on their smartphones and tablets.

The information presented via the Deep Thunder platform will help farmers make more informed decisions and present them with a variety of options. "This is really about engaging decision makers by providing them with high quality information," Treinish said. "We provide detailed information about the impacts and choices that are driven by a weather event, but [the farmers] are the ones making the decisions."

Both Reckford and Treinish, though, remain somewhat cautious when it comes to predicting the accuracy of the forecasts that will be generated by Deep Thunder.

The models generated by Deep Thunder will need to be validated against historical data collected by the University of Georgia weather stations over the years, to asses the accuracy of the forecasts over the long term.

"Accuracy means different things to different people," Treinish said.

Some farmers for instance, are less concerned about rainfall predictions than temperature forecasts. "Are they worried more about an inch of rain or about 90 degree heat or about too much wind to put their fertilizer down?"

The forecasts will never be perfect, Treinish conceded. "The key is we are reducing the error rate," to about half compared to the usual forecasts, he said.

Jaikumar Vijayan covers data security and privacy issues, financial services security and e-voting for Computerworld. Follow Jaikumar on Twitter at @jaivijayan or subscribe to Jaikumar's RSS feed. His e-mail address is

See more by Jaikumar Vijayan on

Read more about big data in Computerworld's Big Data Topic Center.

Follow Us

Join the newsletter!


Sign up to gain exclusive access to email subscriptions, event invitations, competitions, giveaways, and much more.

Membership is free, and your security and privacy remain protected. View our privacy policy before signing up.

Error: Please check your email address.

Tags business intelligencesoftwareapplicationsdata miningApp DevelopmentBusiness Intelligence/AnalyticsFlintUniversity of Georgia

Show Comments