by Nicholas Archambault
Widespread destruction borne from the spread of summertime wildfires has become an annual scourge for millions of residents of the American west. This summer, skies over California were blanketed for days at a time with apocalyptic orange haze, the toxic sign of wildfire activity surging across the state. The New York Times that 2020 is the worst in a series of increasingly more devastating fire seasons. As the result of blazes spread across not only California, but Oregon and Washington as well, over two dozen people have died. Over five million acres have burned across the three states, a sixfold increase from the yearly average between 1950 and 2000.
Western wildfires, however, are more than destructive crises -- they also represent a symptom of a dangerous long-term environmental trend. As the planet continues to steadily warm, the chaos of 2020's wildfire season will become the yearly norm rather than an anomaly. According to the Los Angeles Times, the ten most destructive fires on record have occurred since 1991, fueled by hotter weather, intense drought, and increasingly exurban settlement to set records that are sure to be broken within the coming years.
Fire season precipitates a chain reaction across all levels of a state. Firefighters put their lives on the line to battle the flames, governmental leaders issue evacuation directives, families confront the challenge of obtaining temporary housing. In the aftermath, ecologists seek to understand the root causes of the devastation and work with policymakers to implement more appropriate safety procedures.
A model that can evaluate and predict the size and potential damage of blazes in response to geography and changing climate conditions could save lives and inform the work of those tasked with mitigating wildfires' dangers. This project surveys a subset of nearly two million U.S. wildfires from between 1992 and 2015 and attempts to predict their sizes based on proximity to the nearest city as well as a number of meteorological conditions.
Data were collected from the U.S. Department of Agriculture and the California Irrigation Management Information System at the California Department of Water Resources.
All documentation of data, description of processes, exploratory visuals, and results are available in file, 'Report'.
NOTES: This project was completed for Yale undergraduate course S&DS 425: 'Statistical Case Studies'. Project parameters compelled me to use a simple linear model, but the results (or lack thereof) clearly demonstrate that a different model type would have been more appropriate.