Custom Seasonal Forecasts

The IRI's Climate Predictability Tool enables scientists from the national meteorological services of developing countries to quickly produce their own seasonal climate forecasts. Countries can then use the information to make decisions on water allocation, estimate crop yields and disease outbreaks, and forecast energy requirements, for example.

An example forecast map produced by the CPT for southern Africa

Users download and install the Windows program onto their own computers and can use it in two ways. First, they can create their own forecasts from scratch, using historical rainfall and sea-surface temperature data from their own collection or from the IRI's Data Library. "They might want to see how well an El Nino phenomenon can predict rainfall in their region, for example," says Simon Mason, who heads IRI's Climate Program and is the CPT's principle designer.

The other option is to 'downscale' global climate models created at institutions such as the Met Office Hadley Centre or the European Centre for Medium-Range Weather Forecasts. Typically, the forecasts they produce are at resolutions of 150 to 300 km, which are too coarse for regional applications, says Mason. "Some small countries such as Lesotho and Swaziland and small islands of Indonesia aren't even seen by these models," he says. But by employing statistical methods, the CPT can downscale, or tease out finer spatial detail from, the GCMs. "One advantage of statistical downscaling is that computationally, it's much easier to run, making it ideal for poor countries that don't have major computing resources or staff. We designed the CPT specifically for these countries," he says.

Keeping true to IRI's mission, the idea and development of the CPT grew out of response to a problem. Since 1997, Mason and other colleagues from the IRI have participated in Climate Outlook Forums in Africa, which are designed to train meteorologists on the best available forecasting methods. "They used to spend most of the time trying to create a seasonal forecast, which made the whole process slow, frustrating and expensive," says Mason. "We needed to provide people with a user-friendly interface to first make the forecasts and then to be able to communicate those forecasts in easier to understand ways."

So, building extensively on some existing, open-source code that performed complex mathematical calculations known as singular value decompositions, Mason and his colleagues developed the CPT and began training meteorologists on how to use it. The tool allows them to make forecasts now in a matter of hours, not days. Users can generate maps, charts and data files of raw values or values expressed as probabilities. "All of this means the meteorologists can go back to their countries and produce reliable seasonal forecasts once a month rather than a few times a year," he says.

Ousmane Ndiaye, a graduate research assistant at IRI, went to Niamey, Niger to train meteorologists on using the CPT during a recent West Africa COF, called PRESAO. He says the flexibility of the tool and its output choices makes it attractive to many different groups.

"During the COFs, we invite not only climate scientists but also end users such as health ministry workers, cotton growers and nongovernmental organizations," he says. The groups are encouraged to give their input as to whether or not the forecasts are relevant to their work and if they are meeting their needs for climate information.

"We try to come up with an output that's going to be most useful to them. It could be as simple as a map that shows areas of expected dryness or wetness. We leave it up to them to interpret the map in their respective domains, whether it be for malaria or growing coffee," says Ndiaye.