With a PhD in statistical astrophysics, David Corliss leads a data science team at Fiat Chrysler. He serves on the steering committee for the Conference on Statistical Practice and is the founder of Peace-Work, a volunteer cooperative of statisticians and data scientists providing analytic support for charitable groups and applying statistical methods in issue-driven advocacy.
The year 2020 has certainly been one for the record books. In a year in which the Data for Good community has striven to address so many concerns, the wildfires in the western United States have been one of the most dramatic. I remember seeing the smoke from the fires dye the moon a deep red-orange—and I’m in Michigan!
As severe as the wildfires have been this year, they might have been even worse were it not for an amazing event at Stanford: the 2020 Big Earth Hackathon. Now in its third year, the annual hackathon has teams of 1–4 Stanford students compete to develop data-driven solutions for important environmental use cases. Projects are judged on several characteristics, including defining the specific problem to be addressed, project impact, degree of completion, and quality of presentation. The core of the Big Earth Hackathon is the creativity and innovation of the solution.
Due to the COVID-19 pandemic, the 2020 hackathon moved forward in a virtual format. Teams participated in the Wildland Fire Challenge, focusing on one of the following areas:
- Equality and Fairness: investigating how the fires have affected different subsets of the population
- Prediction and Analysis: modeling which locations and types of buildings are at most risk and the effectiveness of fire response effort, as well as predicting health effects due to smoke
- Mitigation: evaluating mitigation strategies to see which are most effective or becoming more effective as characteristics of the fires change over time
Hackathons often show off the best innovative ideas and technology, and this one was no exception. Four winners were recognized, each offering a creative approach to a difficult problem. One project broke up geography into 500 m squares and analyzed historical fire data to predict the risk of a new fire in each landscape segment. Another used image recognition to provide automated detection of new fires as they started. Locations were flagged by the algorithm as possible new fires were investigated by a rapid response team. This presented a common challenge: reducing false positives in the image recognition algorithm without increasing false negatives. The student team, in collaboration with Fireball International, developed an SQL-Python pipeline to preprocess the images and then apply the algorithm. This high-volume pipeline facilitated processing of larger data sets with more detailed images for comparison. This allowed the image detection algorithm to detect new fires earlier and faster.
Other projects looked at the impacts of fires. One team developed an application called DamageMap, analyzing images from aerial surveys after a fire to automate assessment of fire damage to buildings. Machine learning was used to replace a slow, laborious, and error-prone process to provide first responders and disaster response teams fast and accurate assessment of fie damage. Another project studied the climate effects of wildfires by estimating the carbon footprint. As wildfires have become more severe and numerous in recent years, the amount of greenhouse gases produced has increased dramatically. The application estimated the amount by county and created an interactive map to visualize the impact.
At hackathons and other events, we can see new technology emerging in real time and become part of the process. Applying the latest technology to D4G projects, the most innovative solutions are brought to bear on the most important issues we face as a society. All have a place in learning, developing, sharing, and applying new developments for the greater good.
Get Involved
With the start of the new school year, the August Stats4Good column encouraged Data for Good projects and papers. Now that classes are underway, a number of student competitions have been announced. These programs are not specific to D4G research but are excellent places to showcase your analytic projects making an impact for the greater good.
The ASA Consortium of Sections (GSS/SRMS/SSS) 2021 Student Paper Competition also includes research from recent graduates presenting work done as a student. Competition winners will be featured in the section business meetings at JSM 2021 in Seattle, and winners will receive $1,000 to support their attendance. Applications may be submitted through December 19.
Another program, coming from the ASA and Consortium for the Advancement of Undergraduate Statistics Education (CAUSE), is accepting submissions for the next round of the Undergraduate Statistics Project Competition through December 18. There are two categories: the Undergraduate Statistics Class Project Competition, where the work needs to be for an introductory- or intermediate-level statistics class, and the Undergraduate Statistics Research Project Competition, which focuses on research not done for a class, such as REUs and capstone projects.
For students working in SAS, the 2021 SAS Global Forum Student Symposium competition is for teams of 2–4 students working with a faculty adviser. Team applications are being accepted through November 15, with papers due by February 15.
While these competitions and many more are open to all subjects, not just Data for Good, they offer an excellent opportunity to showcase work, network with other researchers, and establish a practice of “doing well by doing good.”
Leave a Reply