The MWT warehouse fire was one out of hundreds of recycling facility fires in the U.S. in 2023 [8]. These facilities contain a large mixture and volume of recycled material that can be highly flammable, resulting in persistent, high temperature fires and the release of harmful toxics, such as VOCs. As these facility fires are increasing throughout the U.S. there is a significant need to establish strategies that can sensitively and rapidly detect the emitted VOCs [8]. In addition to identification, there is a need for rapid quantification and health assessment of the detected VOCs to help maintain safe emergency responses and protect the health of local residents. Here we demonstrate that PTR-MS coupled with NTA can facilitate rapid identification and hazard assessment of VOCs emitted following a disaster.

First, we compared spatial variation of benzene and HCN, highly toxic VOCs and common combustion products of plastic materials [9, 11, 26,27,28,29]. Air monitoring conducted by the U.S. EPA on April 13th and 14th at sites near the MWT fire detected elevated, but still considered safe, concentrations of benzene and HCN with concentrations ranging between 0.5–2.0 ppm [29, 30]. While benzene and HCN concentrations were below the limit of detection for the air monitoring conducted by the EPA on April 15th, we detected levels of benzene and HCN near the incidence site and bordering the evacuation zone. Notably, we also detected benzene and HCN at our background sampling site, approximately 40 miles away from the incidence site. Concentrations of benzene near the incidence site and bordering the evacuation zone were significantly different than concentrations detected at the background, however, average detection concentrations of benzene across all sampling visits were within national ambient concentrations. Average ambient benzene concentrations ranging between 0.10–1.70 ppb have been detected across the US [31, 32]. While there was lack of a consistent pattern between benzene levels and sampling location, HCN concentrations suggest a pattern between detectable concentrations and proximity to the incidence site. Specifically, HCN concentrations increased as distance to the incidence site decreased. Interestingly, a similar trend in HCN concentrations was detected in a residential area near a gold mining facility, another source of HCN emissions, with concentrations ranging from 0.16 to 8.56 ppbv [33]. Observation of this pattern for HCN, but not benzene, could be due to the high volatility of benzene and the variety of benzene emission sources, resulting in background concentrations not specific to the MWT fire [34]. As fires at recycling facilities are increasing in frequency in the US [4], these data suggest monitoring HCN concentrations at higher sensitivity may be a viable strategy to investigate the spatial distribution of VOCs released from the incidence site. Together, these data highlight the ability of this approach to detect and evaluate ambient concentrations of VOC in communities and highlight the increased sensitivity of the PTR-TOF over traditionally used field-deployable equipment.

In addition to benzene and HCN, plastic combustion also releases high concentrations of other potentially toxic VOCs [9]. However, predicting the identity of all potential combustion products from such variable, non-specific substrate mixtures is challenging. Here, we demonstrate that coupling NTA with the highly sensitive PTR-MS is a rapid and viable approach to generate a comprehensive list of the emitted VOCs following the combustion of variable substances. Using this approach, we detected 46 VOCs that were above local background concentrations, including benzene and HCN, across sampling sites within and bordering the mandated evacuation zone. We continually detected the largest number of total and unique VOCs at the sampling site closest to the MWT fire (Site 1), approximately 0.1 miles away. Across both site #1 visits, the identity and estimated average concentration of detected VOCs varied, illustrating the variability in VOC emissions from fires and supporting the need for frequent air monitoring throughout the duration of disasters. The number of total and unique VOCs across the Site 2 visits were consistently fewer than those detected across the site #1 visits. These data suggest that the 0.5 mile radius evacuation mandate may have been successful in decreasing the exposure of members of the local community to potentially toxic chemicals. However, it is also important to highlight that the decreased number of total and unique VOCs detected across the site #2 visits also coincided with a sampling time 6–8.5 h after the initial Site 1 Visit 1. Although still elevated in comparison to the number of total and unique VOCs detected across the site #2 visits, there was also a decrease in the number of total and unique VOCs detected at Site 1 Visit 2, approximately 8 h after the initial Site 1 Visit 1 stationary sampling. Thus, this information together may suggest the decreased number of VOCs detected at the Site 2 visits may also be a result of the overall reduction in VOC emissions from the MWT fire. Future studies conducting mobile air monitoring could improve this limitation by enabling data collection at Site 1 and Site 2 within closer time frames to help better investigate the relationship between VOC detection and proximity to the incidence site with potentially fewer confounding factors due to the status of the active fire. Notably, the EPA did not detect any of these 46 VOC on the same day of our monitoring, across any of their monitoring sites. These findings align with the growing field of evidence suggesting NTA approaches can improve broad chemical identification in disaster situations [1, 5].

The ability to generate a comprehensive hazard assessment of the broad list of chemicals emitted following a disaster is another important advantage of NTA approaches. Here, we demonstrate that the Hazard Comparison Module can be used to rapidly identify the potential acute human health hazards of each detected VOC [21, 22]. In brief, the Hazard Comparison Module enables users to identify chemical and hazard information to evaluate the potential health effects of chemicals. Through examining the 46 VOCs with estimated average concentrations higher than the local background concentrations, eye and skin irritation had the highest number of VOCs classified as “very high” – “high” hazard. A smaller number of VOCs were also associated with “acute dermal toxicity”, further suggesting exposed skin as a vulnerable target of the emitted VOCs from the MWT fire. Interestingly, while all detected concentrations were within safe limits, skin irritation was noted by several residents. As these safety limits were established for single chemical exposures, this discrepancy highlights the need for the investigation of chemical mixture exposure effects on human health and/or highlights the importance of including susceptible groups into these assessment studies. Notably, eye and skin irritation were also among the common symptoms reported by the residents in East Palestine following the Ohio train derailment [14]. Future studies investigating the health outcomes of the first responders and residents, such as that which occurred following the East Palestine train derailment [35], would provide valuable, complimentary data to the hazard assessment performed in this study. Moreover, a similar assessment of the health outcomes of the Richmond residents and first responders, within and bordering the evacuation zone, could be used to improve our understanding of the health risk of exposure to these VOCs for susceptible individuals, including pregnant women, children, and the elderly.

In combination with hazard assessment, the rapid identification of the exposure level of chemicals released during disaster scenarios is important for risk characterization, enabling responses that can improve the protection of first responders and the local community [1, 22, 36]. Thus, we estimated the exposure concentrations and emission relationships of the detected VOCs to support future risk assessment. All detected VOCs had elevated, but still considered safe, estimated average exposure concentrations. Of the 46 VOCs detected across the sample sites, 5 had estimated detection concentrations at least 1.8-fold higher than background concentrations during at least one site visit. Notably, certain VOCs in this group, such as C2H6N2, had fold changes consistently over 100-fold higher than background levels. In addition, multiple VOCs in this group (HCN, formic acid, acetone, propanal, and propylene oxide) were classified as either a “very high” or “high” hazard for one or more of the human health effects evaluated in the hazard assessment. 3 of these 5 elevated compounds, HCN, C3H6O, and CH2O2, have also been previously associated with plastic combustion [7, 27, 37,38,39]. While these highly elevated levels of potentially hazardous VOCs suggest they may be significant contributors of any observed adverse health effects, it is also important to consider the total exposure burden to improve hazard assessment. Therefore, we evaluated the spatial and temporal emission relationships of the 46 elevated VOCs and identified 5 large groups of at least 4 VOCs that were significantly, positively correlated. These data highlight VOCs that were likely emitted together, providing helpful information for future studies evaluating the exposure effects and risk assessment of real-world mixtures.

As no study is without limitations, we would like to acknowledge a few limitations specific to this investigation. First, we applied a default reaction rate to quantify compounds that we did not have calibration standards for. This approach (i.e., nontargeted analysis) can introduce uncertainty in absolute quantification, as reported by others [25, 40]. Second, although the PTR-TOF 4000 uses “soft” ionization technology, fragmentation of larger VOCs into products that can artificially inflate smaller VOC detection levels can occur due to the ionization procedure and other factors [41]. These additional contributing factors can be the specific identity of the parent VOC and/or atmospheric conditions related to time-of-day and seasonal conditions, as well as the extent of urbanization of the sampling location. Given these uncertainties, we have focused on relative abundance of VOCs in comparison to background levels (i.e., site #0, approximately 40 miles away from the incidence site). This is a useful approach to inform complex VOC mixtures when external calibration, known k rates, and determination of fragmentation are not possible or available, particularly in the case of rapid emergency responses. Third, there are additional tools for resolving VOCs at even higher resolution [40]. However, the instrumentation applied in this study (PTR-TOF 4000) still offers high sensitivity and resolution (>4000) feasible for conducting field studies in a mobile platform. This mobile application of the PTR-TOF 4000 is important, as it permits rapid data collection that is essential when responding to environmental emergency events.

Overall, the large and persistent MWT warehouse fire in Richmond, IN, prompted the need to conduct air monitoring measurements to identify the emitted air pollutants and assess the potential human health risk to the local community. We set out to complement the monitoring efforts initiated by the EPA through stationary air monitoring within and bordering the mandated evacuation zone. By coupling together NTA with the highly sensitive PTR-MS, we generated a comprehensive list of 46 VOCs elevated across sampling sites in Richmond, IN. By leveraging the EPA Hazard Comparison Module, we also generated a hazard assessment of the VOCs to identify most likely the human health effects from exposure to these compounds. Lastly, we determined chemical exposure levels and identified significant VOC emission relationships to support future risk assessment of the VOCs associated with the MWT warehouse fire. Together, these findings support the applicability of NTA coupled with hazard assessment as a valuable tool to identify unknown chemicals and their hazard potential in disaster scenarios.