Recently, scientists from the University of California in the United States published research in the International Journal of Forest Fire Prevention, saying that the use of new technologies such as machine learning can predict the scale of fires.
Researchers use Alaska as a research area because the state has been hit by a series of simultaneous fires in northern forests over the past decade, posing a threat to human health and fragile ecosystems. According to reports, in the first decade of the 21st century, the average annual burned area in northern Alaska reached 7,670 square kilometers, the highest in 150 years.
The researchers analyzed a hypothetical scenario in which dozens of fires broke out simultaneously. This may sound extreme, but in recent years this situation has become very common in parts of the western United States, because climate change has caused hot and dry ground, and the risk of fire has greatly increased.
Using machine learning algorithms, they have developed a model that can help predict whether wildfires will evolve into small, mesoscale or large scale fires. The core of this model is the decision tree algorithm. By substituting climate data and key details about atmospheric conditions and the type of vegetation around the point of fire into the model, researchers can predict the final scale of the fire development with an accuracy rate of 50 %. Among them, a key variable in the model is the water vapor pressure difference, that is, the humidity in the area during the first 6 days of the fire; the other main consideration is the proportion of black spruce tree species in the forest.
Black spruce is the dominant species in northern Alaska coniferous forests, with slender drooping branches, and the seeds can adapt to the environment after a fire. Black Spruce's survival strategy is to destroy all surrounding vegetation in the fire, to make room for future generations to reduce competition. "Within a 2.5-mile radius from the point of fire, black spruce is an important factor in determining the development of the fire," said Sean Corfield, the study's first author.
Under the "feeding" of new data, the algorithm can continuously learn iteratively, so as to quickly calculate the critical value used to identify the occurrence of a fire. "Only a few of these fires will evolve into large-scale fires, and most of the burned areas are caused by them, so we use this new method to focus on identifying specific fires with the greatest risk of loss of control," said Kofield.
Affected by climate change, the frequency of wildfire events that can be expected in each season will increase sharply. The use of new technologies such as artificial intelligence to predict wildfires could benefit the fire department. This information is useful for decision makers who are responsible for allocating an inadequate firework human resource. Timely and effective fire-fighting measures can not only protect people's lives and property, but also play an increasingly important role in protecting ecosystems.
(Source: International Journal of Forest Fire Prevention Compilation: Wu Peng Editor-in-Chief: Cui Guohui)