Holy smokes : measuring wildfire smoke and its impacts using machine learning and causal inference
Abstract/Contents
- Abstract
- Responding to global environmental change and adapting to future climate uncertainty requires accurate measurement, monitoring, and understanding of environmental impacts and estimates of the effects on society. In this dissertation, I focus on wildfire smoke-related air pollution and apply a variety of methods to better quantify the harms from this environmental stressor. In my first chapter, I develop a computer vision method to characterize the spatial extent of wildfire smoke from geostationary satellite imagery with high temporal resolution. When applied to external validation data, the trained model is able to segment smoke plumes even with noisy input labels and is able to capture more within-EPA-monitoring-station variation in fine particulate matter (PM2.5) than existing annotations of wildfire smoke extent. In my second chapter, I use causal inference approaches to examine the effect of wildfire smoke exposure on student learning outcomes as measured by test scores across nearly all public school districts in the US from 2009-2016. I find that smoke exposure negatively affects student learning and that the effects are more negative for younger students, present across test subjects, and affect communities with varying levels of economic disadvantage and racial-ethnic composition. In my third chapter, I develop a new wildfire severity metric that links source fires to their downwind smoke PM2.5 impacts and allows us to quantify the societal harms of individual fires. I find that the proportion of smoke transported across county or state boundaries has increased in recent years and that 7 out of the top 9 most severe smoke-generating fires originated in California and that 6 are from the 2020 fire season. This approach enables assessment of whether specific fire characteristics affect smoke toxicity and helps clarify the growing transboundary impacts to local air quality, which has historically relied on local emissions regulation in order to reduce local exposure to pollutants. These chapters contribute to our understanding of wildfire-related harms and inform better decision making to adapt to future environmental impacts.
Description
| Type of resource | text |
|---|---|
| Form | electronic resource; remote; computer; online resource |
| Extent | 1 online resource. |
| Place | California |
| Place | [Stanford, California] |
| Publisher | [Stanford University] |
| Copyright date | 2023; ©2023 |
| Publication date | 2023; 2023 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Author | Wen, Jeff |
|---|---|
| Degree supervisor | Burke, Marshall |
| Thesis advisor | Burke, Marshall |
| Thesis advisor | Ermon, Stefano |
| Thesis advisor | Lobell, David (David Brian) |
| Degree committee member | Ermon, Stefano |
| Degree committee member | Lobell, David (David Brian) |
| Associated with | Stanford Doerr School of Sustainability |
| Associated with | Stanford University, Department of Earth System Science |
Subjects
| Genre | Theses |
|---|---|
| Genre | Text |
Bibliographic information
| Statement of responsibility | Jeff Wen. |
|---|---|
| Note | Submitted to the Department of Earth System Science. |
| Thesis | Thesis Ph.D. Stanford University 2023. |
| Location | https://purl.stanford.edu/nq450qz8484 |
Access conditions
- Copyright
- © 2023 by Jeff Wen
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...