A machine learning model can evaluate the effectiveness of different management strategies — ScienceDaily
Wildfires are a growing threat in a world shaped by climate change. Now researchers at Aalto University have developed a neural network model that can accurately predict the occurrence of fires in peatlands. They used the new model to assess the effect of different fire risk management strategies and identified a number of interventions that would reduce fire occurrences by 50-76%.
The study focused on the Borneo province in Central Kalimantan in Indonesia, which has the highest density of peat fires in Southeast Asia. Drainage to support agriculture or residential development has made peatlands increasingly vulnerable to recurrent fires. Bog fires not only endanger lives and livelihoods, but also release significant amounts of carbon dioxide. However, prevention strategies have encountered difficulties due to the lack of clear, quantified links between proposed actions and fire risk.
The new model uses measurements taken before each fire season in 2002-2019 to predict the distribution of peatland fires. While the results on Moore can be broadly applied elsewhere, a new analysis would need to be undertaken for other contexts. “Our methodology could be used for other contexts, but this particular model would need to be retrained on the new data,” says Alexander Horton, the postdoc who conducted the study.
Researchers used a convolutional neural network to analyze 31 variables, including land cover type and indices of pre-fire vegetation and drought. After training, the network predicted the probability of a peat fire at each point on the map and provided an expected distribution of fires for the year.
Overall, the neural network predictions were correct 80-95% of the time. While the model was usually correct in predicting a fire, it also missed many fires that actually occurred. About half of the fires observed were not predicted by the model, making it unsuitable as an early warning forecasting system. Larger clusters of fires were usually well predicted, while individual fires were often missed by the network. With further work, the researchers hope to improve the performance of the network so that it can also serve as an early warning system.
The team used the fact that fire forecasts were usually correct to test the effect of different land management strategies. By simulating different interventions, they found that the most effective and plausible strategy is to convert shrubland and scrubland into swamp forests, which would reduce fire incidence by 50%. If this were combined with the blocking of all but the large drainage channels, fires would decrease by 70% overall.
However, such a strategy would have clear economic disadvantages. “The local community desperately needs long-term, stable cultivation to boost the local economy,” says Horton.
An alternative strategy would be to create more plantations as good management will drastically reduce the chance of fires. However, the plantations are among the main drivers of forest loss, and Horton points out that “the plantations are largely owned by larger companies, often based outside of Borneo, which means that the profits do not flow directly back into the local economy, but going beyond providing work for the local workforce.’
Ultimately, fire safety strategies must balance risks, benefits and costs, and this research provides the information to do so, explains Professor Matti Kummu, who led the study team. “We tried to quantify how the different strategies would work. It is more about informing policy makers than offering direct solutions.”
story source:
Materials provided by Aalto University. Note: Content can be edited for style and length.