Wildfires are a natural phenomenon that are important to the balance of many ecosystems. Nonetheless, they can be a threat to humans, whether leading directly to loss of life or causing environmental and economic damage. Every year in Europe, there are about 65,000 fires and these lead to the loss of some half a million hectares of countryside and forest, according to 2010 statistics by the European Commission. The vast majority of these fires are in southern Europe.

Ideally, wildfires would be forecast with lead times of one month or more, but unfortunately they result from a complex interplay of factors. One strong influence is the availability of biomass; another is the water content of that biomass, which is reduced by drought.

This dependence on meteorological conditions can be quantified with fire weather indices – estimates of wildfire risk linked to parameters such as temperature, precipitation, humidity and wind speed. In North America, fire weather indices have been used to forecast wildfires with long lead times; similar forecasts have been achieved by analysing meteorological conditions with statistical techniques.

In Europe, by contrast, forecasts with long lead times aren’t available. Part of the problem is that Europe has a relatively heterogeneous landscape where different types of land – forest, agricultural, urban and so on – occur within short distances of one another. What’s more, southern Europe, unlike the US, is fragmented into many countries, each with different approaches to environmental management. “These might be some of the reasons why wildfire prediction at a pan-European scale has not yet been assessed in the scientific literature,” said Lukas Gudmundsson at ETH Zurich in Switzerland.

Now Gudmundsson, together with colleagues from ETH Zurich and the University of Lisbon, has shown that forecasting with long lead times is possible for southern Europe, using information on droughts. The team employed a statistical technique known as logistic regression to estimate the probability that an area burned by wildfire in a summer month was above average in terms of drought index. The result was a number between zero and one, where zero is no chance of heightened wildfire activity and one is a certain increase in activity.

The researchers tested their method with wildfire and drought data for southern Europe between 1985 and 2010; they found that their retrospective predictions were statistically significant.

“Only time can tell whether our methodology will be used broadly by decision makers,” said Gudmundsson. “In my view this contribution is primarily a pilot study, demonstrating that wildfire activity can be forecast with relatively long lead times. Further work is needed to improve the predictions and to make such models operationally available.”

Improved predictions could help decision makers warn the public and allocate more resources in advance to tackle wildfires, Gudmundsson believes. But he pointed out that complete fire suppression is not the aim of modern forest management. “In many ecosystems, wildfires are an essential feature and regular smaller fires, consuming fuel, can prevent large catastrophic blazes,” he said.

The team reported the findings in Environmental Research Letters (ERL).

Related links

Related stories