Visualizing Wildfire Occurrence & Media Coverage
I'm Ava Hu, a graduating second year in the UC Berkeley School of Journalism. I recently completed my year-long fellowship at DSE and I'm excited to share what I've worked on while on the team.
For my spring semester project at DSE, I built a data visualization workflow to compare wildfire reality with media coverage across the United States between 2020 to 2024. My goal was to examine whether news attention aligns with where wildfires actually occur, and to identify states that may be over-covered or under-covered relative to their wildfire activity.
Methods
To begin the project, I researched both wildfire datasets and news scraping tools. With support and guidance from DSE staff, I selected the Monitoring Trends in Burn Severity dataset from MTBS as the primary fire data source because it provides a consistent national record of wildfire incidents and burned areas. I filtered the data to wildfires between 2020 and 2024, simplified the geometries for web performance, and exported the results into a GeoJSON file for visualization.
For compiling media coverage, I tested several scraping and news data tools, including the New York Times Article Search API, GDELT, SerpAPI, DuckDuckGo search, and NewsData.io. I initially used the New York Times API to build a trial visualization, and later tested Google News API and SerpAPI. Because access limits and historical search restrictions made free tools less practical, I ultimately selected SerpAPI.
By the end of the scraping process, I collected 4,120 wildfire-related articles from 2020 to 2024. After deduplication and removing government announcements and international news, 2,732 articles had identifiable locations, or about 66.3% of the dataset. DSE staff also reviewed and backstopped the code throughout this process, helping me check the scraping workflow and improve the reliability of the dataset. View my Github repository here.
Modeling & Visualization
To help visualize my findings I created several versions of the visualization in Flourish. The final version uses a choropleth map and proportional dots: darker shades show higher wildfire intensity, while dots show the number of news articles connected to each state. I also created a linear regression chart to show the mismatch more clearly. In that chart, states above the regression line are relatively over-covered, while states below the line are relatively under-covered. View both below.
Results & Next Steps
My preliminary findings suggest that media coverage does not always closely follow wildfire frequency, as California received the highest number of wildfire articles. However, the regression model above shows that some states received more coverage than their wildfire counts alone would predict, while others appeared under-covered. For example, states such as Arizona, Montana, Idaho, Nevada and Alaska had comparatively lower media attention relative to their wildfire counts. This suggests that national wildfire coverage may be shaped not only by the number of fires, but also by population, political attention, economic damage, proximity to major media markets, and the visibility of certain landscapes or communities.
The significance of this project is that it turns a broad question about climate coverage into something measurable and visual. Wildfire is often treated as a seasonal or regional disaster story, but this project shows how data can help reveal patterns in which places receive sustained attention and which places remain less visible. The project also creates a foundation for future reporting: it can help journalists identify overlooked wildfire regions, ask why certain communities receive less attention, and examine how media narratives may shape public understanding of climate risk.
My next step is to strengthen the analysis and prepare the project for publication or presentation. I hope to pitch the project to news organizations or develop it into an interactive story that combines the map, the regression chart, and a short written analysis of under-covered wildfire states.
Thank you, DSE!
It's been a very rewarding year, and I'm excited to share that after graduation I'm moving to Los Angeles to work at the Los Angeles Times as a Reporting Fellow, where I'll focus on data journalism.
In case you missed it - last semester I shot and produced an explainer video on Kigali Sim, a new open source tool that DSE and the United Nations developed to help reduce greenhouse gas emissions. Watch the video below!
