Making Forecasts Friendlier

A telling forecast can influence decisions on budgeting for mosquito nets and spray, investment in drought resistant crops, and allocation of water resources for an up-coming season. But how are people using forecasts to make decisions? How do they interpret the probabilities presented in each forecast? IRI scientists are constantly thinking of these and other issues to increase the usefulness of their forecasts.

IRI publishes forecasts of seasonal precipitation and temperature for every region of the world. The forecasts are designed as a resource to help developing nations make informed decisions about water, agriculture and disease management. But the uncertainty inherent to climate prediction, plus the probabilistic nature of the forecasts can get in the way of maximizing their utility.

Currently, forecasts are presented as percent likelihoods that a region will experience dryer, wetter and normal rainfall, or hotter, cooler, and normal temperatures during a given season (see Q&A box). This information can be extremely valuable to regions that are prone to drought and flooding, are vulnerable to epidemics of temperature- and humidity-dependent diseases such as malaria, or whose economies depend on small-scale agricultural productivity.

Download Answers (PDF)

But seasonal forecasts must not only meet the needs of a diverse set of individuals, groups and governments, they must also communicate probabilistic information to a multitude of people who may interpret it in different ways.

"The same forecast may be part of the decision of one farmer to do one thing, and another farmer to do the opposite. There is no simple threshold where one must act in a certain way based on a forecast's results. One must consider what has been happening in the region recently and other non-climatic factors as well," says Simon Mason, who runs the IRI's Climate Program.

On the level of individuals, there may be further ambiguities to consider, says Sabine Marx, associate director at the Center for Research on Environmental Decisions (CRED) and adjunct research scientist at IRI. "There are psychological factors users face in dealing with this information and its uncertainty. This is a cross-cutting issue- finance, health, climate, and all of the fields which involve hedging risk deal with individuals and how they react to risk in different ways. This can factor into any decision made by government leaders or even individual farmers. Some people are risk takers, others are not. It's hard to control for that human element," she says.

"Even if you had perfect information and perfect understanding of information, the decision on how to act on a forecast would still not always be clear," say Marx.

In a recent paper published in the journal Global and Environmental Change, Marx and her colleagues delve into this issue and offer an example: a farmer experienced loss of an early-planted maize crop a few years ago, when the rainy season stopped abruptly for three weeks, causing the entire crop to die. This farmer will be extremely wary and anxious when contemplating whether to plant the current year's crop early. Even if probabilistic information presented by a forecast suggests that the crop-loss episode was rare, and that early planting makes sense this year, "the negative effect stemming from the previous experience may nonetheless prevail," the authors write.

This is a result of what Marx refers to as experiential processing, when one "relates current situations to memories of one's own or others' experience." Education and training on how to use and interpret forecasts could help users better manage decisions. "There might be a need for a separate entity altogether that does this...whose goal it is to train those in developing worlds and here in the U.S. to understand and manage forecast information," says Marx.

IRI is making strides toward these goals by tailoring forecasts to specific needs of individuals and small groups, but individual psychology and inherent uncertainty in climate prediction may never be fully overcome.

A major complaint about the way in which seasonal forecasts are typically presented is that the three categories (see Q&A box) do not give a very clear indication of whether the coming season is likely to be extremely unusual, with potentially large impacts, says Mason.

"'Below-normal' rainfall, for example, represents a one-in-three-year drought, which in many cases may not be severe enough to take any drastic mitigating action. We are busy implementing a new forecasting system that can provide indications of the possibility of more extreme climate conditions."

Another complaint is that seasonal forecasts provide information only about average conditions over a certain period. But for many applications, such as agriculture, it can make a big difference if the season's total rainfall occurs only on a few days when rainfall is very heavy, or over a much larger number of days when rainfall is moderate.

"To address this, the IRI has developed some experimental forecasts that provide information about the expected frequency of rainfall, as well as the total amounts," Mason says.

Visit the IRI's Climate Program pages to learn more about forecasts.

Rebecca Fried is a second year graduate student in the Earth & Environmental Science Journalism dual masters program at Columbia University.