How climate data are helping to predict malaria outbreaks in Africa.
A group of physical scientists based in Reading, UK, recently made the relatively unusual move from climate prediction to epidemiology. Tim Palmer and his team at the European Centre for Medium-Range Weather Forecasts had developed and tested a mathematical model of short-term climate variability. The result was a package that could predict regional climate patterns, such as rainfall and drought, about six months in advance. But having created the model, researchers wanted to do more, and they began looking for applications.
“We wanted to go beyond the theoretical physical sciences and see how we could use this system for the practical benefit of people,” says Palmer.
The ideal opportunity presented itself when Palmer went to a London seminar on malaria, which was held by scientists from the Liverpool School of Tropical Medicine. One talk in particular stood out: a data analysis of malaria statistics. “These epidemics were very much linked to climate — especially the amount of rainfall in the previous year,” Palmer says.
The Liverpool group had robust data on malaria in Botswana, which seemed to mesh well with the Reading team's model, so the two decided to collaborate.
The fruits of this collaboration appear on page 576 of this issue, but getting to this point saw the partners wrestle with a significant language barrier. “In the early days, we would use different jargon,” Palmer says. “The problem was that words we both used sometimes had different meanings.” For example, the physical scientists used the word ‘model’ to mean a mathematical simulation. For the epidemiologists, the same word sometimes meant a simple correlation or a statistical relationship between two things.
Once the groups had cleared up their misunderstandings, they were faced with the task of reconciling two types of data. The physical scientists had global data based on grid points a few hundred kilometres apart, whereas the malaria data was more localized, often gleaned from individual communities within one of the global grid points. The teams used software and statistics to make the data match, then ran simulations based on climate information from the past 20 years to see how well the new system predicted malaria outbreaks. “It seemed to perform very well, very skilfully,” says Palmer.
In Botswana, the rainy season — and so the malaria season — begins in December. But health officials usually have to wait until February before they can make an accurate assessment of the scale of any malaria outbreak. With the new model, they should be able to tell six months in advance which areas will be hit hardest. This will give them time to take preventative action, such as spraying stagnant water, distributing mosquito nets and stockpiling antimalarial drugs.
Palmer and his team now want to use a similar approach for other diseases that have some correlation with climate — such as cholera in Bangladesh and meningitis in sub-Saharan Africa. “The interesting thing for me is that predicting these diseases can be linked directly to a physical climate problem,” Palmer says.
Rights and permissions
About this article
Cite this article
Tim Palmer. Nature 439, xi (2006). https://doi.org/10.1038/7076xia