Grantee Research Project Results

2024 Progress Report: Partnering for Resilient Opportunities To Eliminate Cumulative Toxic (PROTECT) Health Effects from Wildfire PM2.5 in Environmental Justice Communities

EPA Grant Number: R840481
Title: Partnering for Resilient Opportunities To Eliminate Cumulative Toxic (PROTECT) Health Effects from Wildfire PM2.5 in Environmental Justice Communities
Investigators: Thakur, Neeta , Balmes, John R. , de la Rosa, Rosemarie Michelle , Holm, Stephanie , Chow, Fontini , Noth, Betsey M , Chan, Wanyu , Kirchstetter, Thomas W , Basu, Rupa
Institution: University of California - San Fransisco , California Office of Environmental Health Hazard Assessment , Lawrence Berkeley National Laboratory , University of California - Berkeley
EPA Project Officer: Hahn, Intaek
Project Period: December 1, 2022 through November 30, 2025
Project Period Covered by this Report: December 1, 2023 through November 30,2024
Project Amount: $1,330,536
RFA: Cumulative Health Impacts at the Intersection of Climate Change, Environmental Justice, and Vulnerable Populations/Lifestages: Community-Based Research for Solutions (2021) RFA Text |  Recipients Lists
Research Category: Watersheds , Endocrine Disruptors , Environmental Justice , Human Health

Objective:

This proposal seeks to 1) estimate the health effects of sub-daily exposure to wildfire-specific PM2.5 in California, including across social vulnerability factors; 2) understand community recovery from short-term health effects following exposure; 3) understand indoor infiltration of wildfire smoke and the mitigating effect of housing quality and behaviors on health effects; and, 4) identify acceptable community-relevant mitigation interventions.

Progress Summary:

We continue to make progress on all objectives during this past performance period. For Objective 1, We have been able to implement a data processing pipeline that is able to process hourly smoke forecast data from NOAA’s HRRR-Smoke results and observational data from EPA AQS data and PurpleAir data. In addition, we have successfully tested and validated two weeks of hourly values from 2018 and 2020 that include wildfire events and have generate hourly wildfire-specific PM_2.5 estimates.

Figure 1. Hourly Purple PM2.5 data is within 15% accuracy compared to regulatory PM2.5 measurements

This includes reformatting all file types to be readable in the GSI assimilation system, downsizing analysis to PM2.5-only to reduce data processing time, and applying corrections to PurpleAir data resulting in better correlation with nearby regulatory monitors (see figure 1). 

Lastly, to test our entire assimilation process, we have focused on running the system for the entire year of 2019 on an hourly basis. To do so, we have transferred and transformed HRRR-smoke forecasts to our computing system at UC Berkeley. We have also confirmed accurate and complete observational datasets from EPA and PurpleAir for 2019. This dataset will be used as our initial case study for finalizing exposure metrics and testing the health model to be developed through Objective 3. We have also built the exposure assessment tool that will determine the zip code level exposures from the gridded assimilation output of this process. To do so, we are using population centroids for each census tract and matching to the nearest grid point and assigning each centroid with an assimilation PM2.5 value, then taking the spatial average to calculate the average PM2.5 per zip code, again at the hourly level. We then calculate the following exposure metrics: daily average PM2.5, number of hours above the 24-hour EPA PM2.5 standard per day, the hourly maximum per data, the 2-hour average maximum per data, and the 4-hour average maximum per day.

For Objective 2, to generate infiltration factors from the acquired PurpleAir data that covers the state of California and extends from 2016–2023 year, we have been able to deploy a classification model to successfully identify diurnal patterns to categorize sensors as indoor or outdoor and have examined the distribution of indoor and outdoor sensors across California. For the monitors we have been able to classify appropriately, we have established the processing pipeline for an have calculated infiltration factors for valid indoor-outdoor monitor pairs across the study period and geography using modified random component supercomposition and are in the process of refining our model and its sensitivity to measurement outliers.  Resulting estimated infiltration factors will be summarized at the zip code level and used in Objective 3. 

Under Objective 3, we completed a pilot study of mean daily wildfire-specific PM_2.5 and estimated the relative risk of asthma-related emergency department (ED) visits and hospitalizations among a pediatric population cared for at UCSF Benioff Children’s Hospital Oakland over the period extending from 20172020. Through this pilot study, we were able to understand appropriate lags for each event and show cumulative risk in the days that follow wildfire smoke events. We will extend our pilot using the 2019 generated hourly estimates and examine for associations with statewide ED and hospitalization data. 

Lastly, under objective 4, we have established community stakeholder groups (CSG) in each community (San Francisco, Richmond, and Fresno). We have had initial meetings with the Richmond CSG and will have our first meeting with the Fresno CSG in March/April 2025 with planned survey distribution in both communities in June 2025. For San Francisco, through a series of facilitated meetings, three priority interventions were identified to support community resilience during weather events. These include: weatherization/decarbonization of low-quality housing; interventions that support individuals to stay-inplace during events; and communication tools. We are in process of collecting data for these first two areas. For communication tools, CSG members expressed desire to have low-tech tools to provide useful information and resources to communities. Together, we identified that a postcard campaign – including physical mailing campaign as well as social media outreach – would be most desirable and allow each postcard to focus on single events or interventions. These postcards aim to provide high yield information for community members to stay prepared and safe during weather events. Postcards also include results from our community survey relevant to educational material and tailored community resources, such as information on local respite centers and a QR code linking to more in-depth resources on extreme weather preparedness, including city and state-wide strategies, emergency response plans, and guidance on accessing support services during extreme weather events. Example postcards are included as Figure 3.

Figure 3. Sample Postcards for distribution.

Future Activities:

During the next performance period we will continue progress on all objectives. During the next performance period we will continue progress on all objectives. Specifically, under objective 1, using our ensemble model, for the time period extending from 2016–2022, we will produce a 3km x 3km hourly gridded model of PM_2.5 concentrations. These results will be compared and validated against the EPA IMPROVE network to quantify accuracy and precision.  For objective 2, we will generate infiltration factors estimated at the census tract and zip code level using the acquired PurpleAir data and household characteristic information. For the health analyses proposed under objective 3, we anticipate receipt of the state-wide Health Care Access Information (HCAI) data on ED visits and hospitalizations for respiratory, cardiovascular, and neurovascular conditions spanning the study period in the next 1-2 months, alternatively, we have HCAI data in hand for the period extending from 2016-2000. This data will undergo a series of data checks and cleaning as proposed in our QAPP. We will then conduct a series of pilots, modeling health outcomes using single wildfire smoke events (e.g., 2018 Camp Fire) or a single region using multiple years of data. This will inform the final model selection, which will be done in year 3 of the grant period. For objective 4, we will continue our key informant interviews to understand barriers to accessing and implementing weatherization services in San Francisco as well as work with our SF CSG to identify interventions that support individuals to stay-in-place during weather events. We will continue to meet with the formed CSGs in Richmond and Fresno, CA and anticipate distributing our community survey in these communities.

Journal Articles:

No journal articles submitted with this report: View all 3 publications for this project

Supplemental Keywords:

Wildfire smoke, health effects, community engagement, cumulative risk, infiltration.

Progress and Final Reports:

Original Abstract

2023 Progress Report