A transdisciplinary approach to understanding the health effects of wildfire and prescribed fire smoke regimes

Prescribed burning is used to reduce the occurrence, extent and severity of uncontrolled fires in many flammable landscapes. However, epidemiologic evidence of the human health impacts of landscape fire smoke emissions is shaping fire management practice through increasingly stringent environmental regulation and public health policy. An unresolved question, critical for sustainable fire management, concerns the comparative human health effects of smoke from wild and prescribed fires. Here we review current knowledge of the health effects of landscape fire emissions and consider the similarities and differences in smoke from wild and prescribed fires with respect to the typical combustion conditions and fuel properties, the quality and magnitude of air pollution emissions, and the potential for dispersion to large populations. We further examine the interactions between these considerations, and how they may shape the longer term smoke regimes to which populations are exposed. We identify numerous knowledge gaps and propose a conceptual framework that describes pathways to better understanding of the health trade-offs of prescribed and wildfire smoke regimes.


Introduction
Landscape fire smoke is a complex and dynamic mix of energy, water vapour, gases, and aerosols. Gaseous emissions affect climate forcing and biogeochemical cycling, while aerosols affect rainfall distribution and warming of the troposphere (Langmann et al 2009) and transport of plant nutrients. The public health impacts of air pollution, including landscape fire smoke, have become more clearly characterised over recent decades (Reid et al 2016a(Reid et al , 2016b. Globally, an estimated three million deaths each year are attributable to outdoor air pollution (Lim et al 2013), while an estimated 340 000 annual deaths are attributable to smoke from landscape fires (Johnston et al 2012). Even when fires do not cause loss of infrastructure or direct human injury, large human populations can be exposed to smoke pollution causing measurable increases in mortality and morbidity (Henderson et al 2011, Faustini et al 2015, Adetona et al 2016, such that including the health impacts of smoke changes the overall economic burden of wildfire events (Johnston and Bowman 2014).
Fire is an inevitable feature of many environments on Earth, yet humans are able to manipulate landscape fire activity for a diversity of reasons (Bowman et al 2011). A key technique is intentionally burning landscapes to reduce fuel loads and hence the occurrence, extent, and severity of uncontrolled wildfires (Fernandes and Botelho 2003). When appropriately planned, escape of prescribed fires from intended boundaries occurs in less than 1% of cases (Dether and Black 2006), but when escapes occur they can result in significant damage to homes and infrastructure (Smith 2012). In many jurisdictions, smoke management is an important constraint on prescribed burning (Sneeuwjagt et al 2013) leading to an unresolved policy debate about human health trade-offs between smoke from wildfire and prescribed fires. This problem involves numerous disciplines including epidemiology, meteorology, atmospheric chemistry, fire ecology and management, but few researchers or practitioners have such transdisciplinary expertise. Our objective is to provide a succinct review of this complex problem and to propose a conceptual transdisciplinary framework that can be used to guide further research.

Smoke impacts on human health
The health impacts of landscape fire smoke are driven by: (1) chemical composition and concentrations of pollutants in the smoke plume; (2) the intensity and duration of the smoke exposure during an event; (3) the extent to which individuals can protect themselves from the exposure and (4) the number of individuals who are exposed and their underlying health status.
The epidemiologic literature has focussed on concentrations of particulate matter less than 2.5 microns in diameter (PM 2.5 ), which is routinely measured and has been widely studied (Pope and Dockery 2006). Exposure to PM 2.5 from many different sources has a range of impacts on human physiology, including: promotion of inflammation and blood coagulation; impairment of the respiratory, cardiovascular, and autonomic nervous systems; and increased risk of genetic mutations (Brook et al 2004, Barregard et al 2006, Berhane et al 2011, Danielsen et al 2011. Although the composition of PM2.5 can vary widely, there is insufficient evidence to quantify the relative influence different chemical compositions on health outcomes at this time (Levy et al 2012). In a healthy person, short term physiological changes in response to PM 2.5 are unlikely to be of clinical importance, yet they can pose significant risk to unhealthy people. For instance, increased concentrations of PM 2.5 may precipitate heart attack or stroke in a person who already has cardiovascular disease, potentially leading to death (Brook et al 2004). Likewise, smoke exposure might cause minor irritation of the eyes or throat in healthy people, but could precipitate severe breathing difficulty in a person with asthma or other respiratory disease (Johnston et al 2006). Positive associations between forest fire smoke and mortality have been documented in a range of settings including Europe . The population health impacts of PM 2.5 are observed across all PM 2.5 concentrations with no safe threshold, which highlights the potential impacts of prescribed fires, even though they produce comparatively less smoke than wildfires.
Although serious health outcomes are relatively rare, the absolute number of people who suffer serious impacts from smoke events can be substantial when large populations are exposed. The health impacts of air pollution are tied to the prevalence of risk factors in the population, such as cardiovascular and lung diseases, diabetes, older or younger age, and lower socioeconomic backgrounds (Eze et al 2015, Pope et al 2011). The same is likely true for fire smoke, although there has been less research on the topic (Reid et al 2016a(Reid et al , 2016b. Smoke exposure is known to disproportionately affect people with lung diseases (Henderson and Johnston 2012) and the evidence concerning impacts on heart diseases is emerging (Rappold et al 2012, Haikerwal et al 2015a, 2015b. A few studies have also demonstrated adverse impacts on infants and unborn babies (Jayachandran 2009, Holstius et al 2012, and on people with lower socioeconomic status (Reid et al 2016a, 2016bEnvironmental Research, Rappold et al 2012. Higher adverse health outcomes from smoke exposure have also been documented for Indigenous Australians who have worse population health outcomes compared with non-Indigenous people (Johnston et al 2007, Hanigan et al 2008. Some of the adverse health effects of smoke can be mitigated through preventative medication (Bhogal et al 2006). By definition, preventive medication must be taken in advance to reduce the likelihood of deterioration in symptoms (National Asthma Council Australia 2015), and this requires advanced notice of possible smoke impacts. Similarly, short-term reduction of personal exposure to smoke in homes can be achieved if doors and windows are closed before a smoke event. In an open house, indoor air quality will rapidly reflect outdoor conditions (Dix-Cooper et al 2014). There is evidence that creating clean air refuges using mechanical air filtration substantially reduces wildfire smoke exposure (Barn et al 2008(Barn et al , 2016. However, such devices have not been widely adopted. Moving people to areas unaffected by smoke is logistically challenging and stressful for the affected populations. The only evaluation of evacuation during severe smoke events concluded that it was protective in only 30% of cases (Krstic and Henderson 2015). Specialised face-masks can be effective in occupational settings for personal protection, however there is no evidence these are helpful as a public health protection measure for wildfire smoke episodes (Sbihi et al 2014).
Wild and prescribed fires generally result in different smoke exposures to human populations. Wildfires are unpredictable and, in temperate forest ecosystems, can occur at a low frequency (decades) whereas prescribed burning is more predictable and can occur at a greater frequency. Thus wildfire smoke affects large areas and potentially large populations, contrasted with the potential for severe local impacts of prescribed fire smoke (Haikerwal et al 2015a, 2015b). The relative health effects of these contrasting smoke exposures remain unknown. An additional research challenge lies in understanding the contribution of background air pollution from other sources, which complicates resolving the relative effects of wildfire and prescribed fire smoke.

Biomass combustion and smoke production
Differences in fire intensity, area burnt, fuel availability and type, and duration of fires all affect smoke volume, composition and dispersion. Large wildfires tend to occur during drought periods and in severe fire weather, driven by high temperatures, low relative humidity, and strong winds (Meyn et al 2007). This results in the combustion of larger fuels, with the potential to spread to the canopies of woody vegetation (Wagner 1977). By contrast, prescribed fires are typically lit during milder weather, which results in the combustion of fine surface fuels and limited damage to forest canopies (Agee and Skinner 2005). Complex physiochemical processes of pyrolysis and combustion involving the breakdown of hydrocarbons drive qualitative and quantitative differences between wild and prescribed fire smoke.
Pyrolysis involves heating fuel to enable the endothermic reactions that break up long chain hydrocarbons to produce solids and gases, and vaporise the moisture in the fuel (Sullivan and Ball 2012). Many of the gaseous compounds are strong greenhouse gases (Jain et al 2006) and can be toxic to humans (Demling 2008). Combustion is a rapid exothermic oxidation of the products of pyrolysis that releases energy in the form of heat and light. Idealised complete combustion is a process where all hydrocarbons in the original fuel, both gaseous and solid, combine with oxygen to release energy and form carbon dioxide (CO 2 ) and water vapour (H 2 O) (Rein 2013). In all landscape fires, combustion is incomplete, producing a smoke that includes gases (CO 2 , CO, H 2 O, NO x , NH 4 , SO x , CH 4 , phenols, etc) and aerosols (elemental, organic, and inorganic carbon compounds)(Akagi et al 2011). While many of these species are also harmful to human health, the population level impacts are primarily driven by the concentrations of aerosols measured as PM 2.5 (Reisen et al 2015). The combustion process also liberates small quantities of macro and micro plant nutrients and trace elements, including heavy metals. Constituent compounds in the smoke can undergo further chemical transformation in the atmosphere after emission, while aging can also alter aerosol size, composition, and concentration (Cubison et al 2011). In general, the concentration of PM 2.5 is thought to scale with the concentrations of the cooccurring organic and inorganic compounds (Ward et al 1996, Andreae andMerlet 2001).
The enormous variability in smoke composition is due to the substrate burned and factors that determine the completeness of oxidation during combustion, including the fuel moisture content, temperature of combustion, and amount of oxygen available (Hobbs et al 1996). Combustion is often divided into hightemperature flaming and low-temperature smouldering types to differentiate the chemical processes and the smoke emissions. Flaming is produced by the combustion of gases, which is rapid and involves more complete consumption of the original fuel. Smouldering is produced by the combustion of solids, which is slower and less complete. Flaming versus smouldering combustion is one of the most important determinants of potential public health impacts, because smouldering and incomplete combustion greatly increases the emissions, the suite of toxic co-pollutants, and the burn duration ( . In a fuel bed with high moisture content, the heat released by flaming combustion is often used to vaporise water in nearby fuel particles. In a fuel bed with low moisture content, combustion is more rapid and complete due to the maintenance of higher temperatures, promoting the combustion of the volatile compounds and gases. These factors are shaped by the current and antecedent meteorological conditions (Weise and Wright 2014).
Weather conditions have major influences on fire behaviour and smoke production, and fire weather is increasingly studied at a global scale (Field et al 2015, Jolly et al 2015). Relative humidity and rainfall drive landscape fuel moisture, while wind speed drives oxygen delivery and energy transport, which dictate the speed and completeness of combustion, as well as the rate of fire spread (Rothermal 1983, Dowdy et al 2010). The spatial arrangement of the fuel array also influences oxygen availability, the completeness of the combustion process, and hence the composition of emissions (Weise and Wright 2014). Densely packed fuels allow little oxygen penetration and slower, less complete combustion. Larger fuels, such as branches and logs, have lower surface-area-to-volume ratios that reduce oxygen penetration. These fuel types often undergo extended smouldering combustion rather than rapid flaming combustion. For all of the above reasons the leading fire front often features more efficient flaming combustion, while smouldering combustion dominates the flanks (Surawski et al 2015).
Assigning emissions factors to strict prescribed or wildfire categories is complicated by the range of variables driving smoke production. For example, prescribed fires may be carried out in moister fuels, leading to less efficient and greater smouldering combustion. They are also less likely to burn heavy fuels, such as logs, that may smoulder for a long time (Bertschi et al 2003). Wildfires typically have a mix of flaming fires (at the head of the fire) and smouldering fires (on the flank of the fire). Improved physical models of pyrolysis, fuel consumption and spread dynamics may be required to accurately establish smoke constituents under different burning conditions and fire behaviours.

Smoke transport
Smoke transport is shaped by many complex interactions between landscape fire and meteorological conditions, with the most critical determinant being the injection height, at which horizontal transport of the plume begins. Smoke plumes with low injection heights will disperse within the mixing layer, the lower layer of the atmosphere in contact with the Earth's surface, while smoke plumes with high injection heights may be carried aloft over long distances by winds in the upper atmosphere ( figure 1). There is a broadly linear association between fire intensity (total instantaneous energy release) and plume height (Val Martin et al 2009, Raffuse et al 2012, Peterson et al 2014, because the rate of biomass consumption drives the rate of heat release. In turn, the rate of heat release drives the plume buoyancy, such that intense, hot fires create a convection column and high injection heights (Fromm et al 2006). Intense wildfires are associated with higher injection heights than lower intensity prescribed fires, which are assumed to produce lower smoke plumes, as is the case with agricultural fires that have low fuel loads (Kaskaoutis et al 2014). The locality of impact from different fire types can be expected to be correlated with injection height, and therefore fire intensity, with lower intensity fires more likely to impact local communities and high intensity fires resulting in broader long-range but more diffuse population impacts.
The atmospheric stability and mixing height (the height above ground in which turbulence produces well-mixed air) also determines how well smoke can be dispersed (figure 1). A stable atmosphere occurs when air temperatures decrease slowly with height, whereas an unstable atmosphere occurs when temperatures decrease rapidly with height (Fergu- and convection columns that can penetrate past the mixing layer, even reaching into the stratosphere. This smoke often has minimal local impacts relative to the size of the fire, but the aerosols in the upper atmosphere have a significant climate impact (Westphal and Toon 1991), and they can be transported around the globe to have impacts in other areas (Sapkota et al 2005). (a) a low intensity fire in a pronounced inversion or stable atmosphere produces smoke that is restricted to bottom of the mixing layer; (b) a low intensity fire in a well-mixed atmosphere produces smoke that is transported regionally below the mixing layer; (c) a moderately intense fire in a wellmixed atmosphere produces smoke that penetrate through the mixing layer enabling long distance (100 kms) transport in the upper troposphere; and (d) a high intensity fire in an unstable atmosphere produces smoke that penetrate the stratosphere and disperses at a global scale.
The ability for the atmosphere to disperse smoke can be captured by the ventilation index (VI, Goodrick et al 2013), which multiplies the mixing height and the mean wind speed. This index can be used to plan optimum conditions for prescribed fire to ensure smoke transport away from local communities, and is applied in this capacity in British Columbia. High VI values indicate a high mixing height and wind speed, with good smoke dispersion conditions. Low VI values indicate a low mixing height and still air, often resulting in temperature inversions where pollution is trapped in a cold layer of air near the surface. Smoke impacts are often highest during stable atmospheric conditions and temperature inversions that form under meteorological conditions suited to prescribed burning (Price et al 2016). The atmospheric dispersion index (ADI, Lavdas 1986, Goodrick et al 2013) builds on the ventilation index by adding a measure of atmospheric stability in predicting smoke dispersion capability, and variants on this index is used in several jurisdictions in the United States to inform planned burn planning. Evaluation of these simple dispersion indices as an alternative to computationally intensive smoke modelling is an important avenue for future research.

Health and prescribed and wildfire 'smoke regimes'
We suggest that smoke emissions from wild and prescribed fires can be conceptualised as a smoke regime in the same way that fire ecologists integrate the effects of recurrent fires under the rubric of the fire regime. Understanding the totality of the health effects of a smoke regime from prescribed and wildfires requires considering numerous trade-offs. The increased fraction of fuel burnt in wildfires produces more smoke emissions. Williamson et al (2013) found the mean horizontal footprint of wildfire plumes to be six times greater than those for prescribed fires in south-eastern Australia. However, the higher intensity of wildfires means the greater likelihood of plumes reaching the upper atmosphere, where horizontal transport can carry the smoke away from the local area. Prescribed fires can be planned, to some extent, around optimal atmospheric conditions for smoke transport, but the lower intensity of the fire can result in lower atmospheric injection heights and greater local impacts on air quality. In particular, prescribed burns must avoid hot and windy conditions in favour of cool, still days to minimise the risk of fires escaping planned boundaries. This inevitably means that prescribed fires often coincide with overnight temperature inversions that trap smoke in valleys where people often live.
Prescribed burns may cause a negative feedback on subsequent wildfire area and intensity, but the nature of this feedback can only be quantified over the long term, via the influence of a prescribed burning program on the fire regime of an area. Empirical and simulation studies have shown that prescribed burning reduces the area of wildfire in the forests of southern Australia by a ratio of approximately one third, meaning each one-hectare reduction in wildfire are requires a three-hectare area of treatment (Boer et al 2009, Price and Bradstock 2011, Bradstock et al 2012b. This ratio has been termed 'leverage' (Loehle 2004, Price et al 2015. In non-forest regions with lower fuel loads and less frequent wildfire, the leverage is even lower (Price et al 2012a, Price et al 2015. Only in regions with particularly high fire frequency, such as savannas, can a 1:1 ratio be achieved (Vilen andFernandes 2011, Price et al 2012b). The implication is that prescribed burning programs will increase the overall area burnt in most regions. The inefficiency of prescribed fires at reducing wildfire area is because fuels recover quickly and wildfires are rare, such that most fuel-reduced patches are not encountered by a wildfire. If a wildfire does encounter a treated patch the intensity is reduced, which implies lower fuel consumption and smoke emission but the effects have never been quantified with respect to intensity, fuel consumption or smoke. Instead, studies have focussed on the reduction of fire severity, a measure of crown scorch (Keely 2009), which can be mapped using post-fire remote sensing. These studies have demonstrated that recent prescribed burning reduces damage to the forest crown, but the effect decreases as the time-sincefire increases and as the fire weather becomes more severe (Bradstock et al 2010, Bradstock 2012, Tolhurst andMcCarthy 2016). To quantify the full effect of a prescribed burning program on smoke pollution requires the integration of the leverage effect and the reduction in intensity effect. Bradstock et al (2012a) explored this integration in the context of predicting greenhouse gas emissions, concluding that prescribed burning programs were unlikely to reduce emissions, while acknowledging that a definitive answer depends on gathering empirical evidence on fuel consumption by prescribed and wildfire.
Given the mechanisms of feedback between prescribed and wildfire, community smoke exposure can only be compared by considering the smoke regime rather than emissions from individual fires, in other words, smoke from more frequent and less severe fires versus smoke from less frequent more severe fires. For example, Schweizer and Cisneros (2016) argue that smoke regimes in the USA would be less harmful to public health with a policy of regular prescribed burning than with a policy of total fire suppression. The comparative population health impacts of smoke regimes will be driven mostly by the number and susceptibility of people affected by smoke from each regime. Smaller fires usually have more local impacts while larger fires have greater capacity for long distance transport. Although smoke generation and population exposure might be reduced by smaller fires, the public health costs may be borne by populations that are repeatedly exposed to smoke from a local regime of prescribed fire. This may be further complicated by climate change, which is reducing the window of opportunity for prescribed burning each year and leading to more fires being set when the conditions are favourable. This compression can result in pollution events as severe as those associated with larger wildfires, exemplified by the conditions in Sydney during May 2016 (figure 2).

Modelling population exposure to smoke
Modelling smoke production and transport is critical for predicting and modelling smoke exposure from wild and prescribed fires. This is challenging because of the need to parameterise fuel type and mass, rate of and type of consumption, and energy release and atmospheric conditions (Heilman et al 2014).
The direct measurement of smoke emissions is constrained by the availability air monitoring stations, and a dense network is required for tracking and understanding dispersion. Routine monitoring is often biased towards large urban or industrial centres, with a focus on mobile and industrial sources for regulatory purposes (Johnston et al 2010). Increasingly, however, smoke monitoring networks based on large numbers of inexpensive, spatially dispersed, real time instruments are being deployed. Such networks help to improve smoke dispersion models and to quantify community impacts more meaningfully. Examples include the Florida Air Quality System (FLAQS, www.dep.state.fl.us/Air/air_quality/airda ta.htm) and the Base Line Air Network of EPA Tasmania (BLANkET, http://epa.tas.gov.au/epa/real-timeair-quality-data-for-tasmania). These networks can also provide real-time data on smoke impacts to the public. An alternative to ground-based particle monitors are remote sensing platforms on satellites, which can measure atmospheric aerosol concentration over large geographic areas. Data products such as the aerosol optical depth (AOD) or aerosol optical thickness (AOT) measure total atmospheric aerosol by comparing the scattering of light in a given pixel with its spectral signature under a low aerosol load. While AOD measures particles throughout the atmospheric column, there is often strong correlation with particulate concentrations at ground level (van Donkelaar et al 2015). These measurements are available in areas outside the ground-based monitoring network, allowing quantification of long-distance smoke transport. The polar-orbiting MODIS, AVHRR and VIIRS instruments all provide AOD measurements at varying resolutions and overpass frequencies, while geostationary satellites, including GOES satellites over the Americas and the Himawari-8 satellite over east Asia, also provide continuous AOD coverage. Other instruments use LiDAR (e.g. CALIOP) or stereoscopic imagery (e.g. MISR) to provide three-dimensional imagery of smoke plumes. Goodrick et al (2013) provide an excellent review of the range of empirical and physical models available for smoke dispersal. Empirical models use statistical relationships to describe observed smoke concentrations with respect to a range of predictor variables associated with smoke emissions including land use, meteorology, satellite location of fire and estimates of intensity. For example, Yao and Henderson (2014) developed a system for British Columbia, Canada that incorporates MODIS AOD, atmospheric venting index, measured PM 2.5 , and MODIS fire radiative power measurements to estimate smoke concentrations across the province. There was a correlation coefficient of 0.84 between model estimates and observed PM 2.5 concentrations in training cells on high-smoke concentration days. These models are simple to parameterise and require little computational power, but are often only locally applicable and can perform poorly in conditions outside the training environment. Physical models range in complexity from simple box models, such as the atmospheric dispersion index (ADI), through to grid models with a lattice of cells with pollutant concentrations and meteorological variables (Goodrick et al 2013). In between, there are Gaussian plume models, which incorporate wind impacts on plume dispersion, puff models, which disperse independent puffs of smoke with defined pollutant concentrations, and particle models, which simulate the movement of individual particles through the atmosphere. Each approach has particular strengths, weaknesses, and applications.
The most elaborate and potentially useful class of models are gridded numerical chemical transport models, such as the Australian Air Quality Forecasting System (AAQFS, Cope et al 2004). These systems incorporate a meteorological model to simulate atmospheric temperature profiles and wind, emissions sources parameterised with emissions factors for a variety of chemical species, and chemical models able to account for changes in chemistry as the smoke ages. Increases in computing power and the greater availability of data sources for validation and model improvement mean that physical models have the potential to improve our prediction of smoke impacts from landscape fire over the coming years. A fusion of multiple approaches is likely to emerge, to develop systems for accurate prediction and estimation of smoke emissions. Such models will not only be pivotal in resolving the trade-offs of wildfire and prescribed fire, but also for providing advanced public health warnings (Yuchi et al 2016). Framework for understanding smoke health impacts of wild and prescribed fires A range of research challenges in the fields of fire ecology, epidemiology, and atmospheric sciences must be resolved before the relative human health costs of wild and prescribe fire smoke regimes can be accurately estimated. A framework to understand the interrelations among the many complex factors is presented here (figure 3). To understand the full impact of prescribed burning programs on human health we must consider the risk factors for each fire (the event), the feedbacks between prescribed and wildfire over time (the fire and smoke regimes), how this translates into the interaction between smoke and populations (the exposure), and, ultimately the effects of the exposure on the population as a function of its size, vulnerability and preparedness (health impacts). This framework will be useful for developing predictive models of smoke exposure, because understanding exposure is essential for quantifying the health costs of both wild and prescribed fire smoke regimes. Predictive models are also crucial to enable fire managers to reduce smoke exposures and provide warnings to susceptible people.

Conclusion
We have provided a transdisciplinary review of the health impacts of smoke from wild and prescribed fires. This problem demands understanding recent advances in air pollution epidemiology generally, and the specific effects of landscape fire smoke. The composition of landscape fire smoke is extraordinarily Figure 3. Framework to understand the causes and uncertainties of landscape fire effects on human health. The framework is broken down into events, regimes, exposure, and impacts. Individual fire events have features that can either dampen or amplify smoke production (green and red lines). Smoke and fire regimes are the collective consequences of numerous events. Long-term smoke exposure is shaped by the interaction between fire and smoke regimes. The health impacts of smoke exposure are modulated by the distribution and demographic profile of human populations. The aggregate population health impacts are determined by the complex interaction of all the factors denoted in the diagram. There is no current assessment that considers all factors that influence the health outcomes of prescribed and wildfires. Such analysis is required to better understand the human health trade-offs inherent is the use of planned fire to manage wildfire. complex, and we use PM 2.5 concentrations as an epidemiologic proxy for the mixture because we have only incipient understanding of its overall toxicology. Modelling of smoke particle emissions demands quantification of the type of fuel and prevailing meteorological conditions, both of which influence the combustion conditions. The likely human health impacts of this smoke depend on the size of populations affected, the concentration and duration of exposure, and the effectiveness of any public health interventions. Prescribed fires can reduce the incidence of wildfires, trading off frequent short-term local-scale smoke pollution events for reductions in longer time-scale broader population exposures. This trade-off must be incorporated into any evaluation of the health impacts for specific smoke regimes but it is impossible to understand the respective health impacts of prescribed and wildfires with the available data. Instead, we propose a structured framework that outlines the necessary steps to move towards resolution of this critical public health issue.