Forest fire dataset could help firefighters save lives, property


A UC Riverside team led by an assistant professor of computer science Ahmed Eldawy is collaborating with researchers at Stanford University and Vanderbilt University to develop a dataset that uses data science to study the spread of wildfires. The dataset can be used to simulate the spread of forest fires to help firefighters plan for emergency response and conduct an evacuation. It can also help simulate how fires might spread in the near future under the effects of deforestation and climate change, and aid in risk assessment and planning for the development of new infrastructure.

The open source dataset, named WildfireDB, contains more than 17 million data points that capture how fires have spread to the contiguous United States over the past decade. The dataset can be used to train machine learning models to predict the spread of forest fires.

Ahmed Eldawy

“One of the biggest challenges is having a detailed and organized dataset that can be used by machine learning algorithms,” Eldawy said. “WildfireDB is the first comprehensive, open-source dataset that links historical fire data to relevant covariates such as weather, vegetation and topography. “

First responders depend on understanding and predicting the spread of a forest fire to save lives and property and to prevent the fire from spreading. They must find the best way to allocate limited resources over large areas. Traditionally, the spread of fire is modeled by tools that use physics-based modeling. This method could be improved with the addition of more variables, but until now there was no complete, open source data source that combines fire occurrences with geospatial features such as mountains, rivers, cities, fuel levels, vegetation and weather. .

Eldawy, along with UCR doctoral student Samriddhi Singla and undergraduate researcher Vinayak Gajjewar, used a new system called Raptor, which was developed at UCR to process high-resolution satellite data such as vegetation and vegetation. weather report. Using Raptor, they combined historic wildfires with other geospatial features, such as weather, topography, and vegetation, to create a scale dataset that included most of the United States. .

WildfireDB mapped historical fire data in the contiguous United States between 2012 and 2017 with spatial and temporal resolutions that allow researchers to focus on the daily behavior of fires in areas as small as 375-meter square polygons. . Each fire occurrence includes vegetation type, fuel type, and topography. The dataset does not include Alaska or Hawaii.

To use the dataset, researchers or firefighters can select information relevant to their situation in WildfireDB and train machine learning models capable of modeling the spread of forest fires. These trained models can then be used by firefighters or researchers to predict the spread of forest fires in real time.

“Predicting the spread of forest fires in real time will allow firefighters to allocate resources accordingly and minimize loss of life and property,” said Singla, the first author of the article.

The article, “WildfireDB: an open-source dataset connect wildfire spread with relevance determinants”, will be presented at the 35th Conference on Neural Information Processing Systems (NeurIPS 2021) Track on Datasets and Benchmarks and is available here. A visualization of the dataset is available here. Eldawy, Singla, and Gajjewar were joined in the research by Ayan Mukhopadhyay, Michael Wilbur, and Abhishek Dubey at Vanderbilt University; and Tina Diao, Mykel Kochenderfer and Ross Shachter at Stanford University.

Header photo: Mike newbry to Unsplash


About Author

Comments are closed.