Demonstrating the most effective interventions to improve classroom air quality. What novel in situ tests of real-world conditions show is still missing in our guidance

Over 20 years ago a report commissioned by the European Commission identified air quality in schools as a public health priority. Despite this concern, little action was taken in the following two decades. Over the last two years as classrooms were increasingly recognised as hotspots for the transmission of SARS-CoV-2, renewed interest and resources have been made available in response to this issue. Questions remain, however, over how best to achieve safer classroom air. Our analysis assessed a range of in situ interventions to remove particulate matter (PM2.5) and carbon dioxide from inside a populated classroom. Our approach used saline spray and volunteers’ exhalations as our source of PM2.5 and carbon dioxide to explore the ability of high efficiency particulate air (HEPA) filters, natural ventilation and a recirculating A/C unit to remove these air pollutants which collectively provided a novel set of data. For a total window opening of 1.86 m2 for a 181.7 m3 classroom with a HEPA filter with a 703m3/hr clean air delivery rate, our results confirmed that outdoor air was needed to purge the room to reduce carbon dioxide levels that otherwise rose to >1000 ppm in 12 min. Cross and natural ventilation reduced levels of PM2.5 and carbon dioxide very effectively—in under 5 and 10 min respectively during low levels of outside PM2.5. We conclude that natural ventilation supplemented with the use of HEPA filters is the most effective way to reliably improve indoor air quality year-round, balancing the need to have easy to enact approaches to reduce the buildup of PM2.5, airborne viruses and carbon dioxide. These results highlight an important knowledge gap. Without having localised real-time outdoor air pollution sensing, evidence-based decisions cannot be made about how often, and for how long, windows can safely remain open to purge classrooms in times of poor quality outdoor air.


Introduction
The Lancet Covid-19 Commission Task Force on ventilation in schools highlighted the lack of in situ studies evaluating the independent impact of ventilation and air cleaning to reduce the risk of COVID-19 transmission (Lancet COVID-19 Comm 2021). This important omission was reconfirmed by a recent literature review which identified 'significant gaps within the body of literature on COVID-19 mitigation measures for naturally ventilated classrooms' (Sutherland et al 2022). Specific concerns highlighted in this review related to a lack of studies that assessed: in situ measurements of the high efficiency particulate air (HEPA) performance in a naturally ventilated classroom; in situ measurements of the same classroom's natural ventilation performance; or in situ measurements of the HEPA filter and natural ventilation measured when they were used simultaneously. that the use of HEPA filters cannot impact carbon dioxide levels and so cannot substitute for opening windows or other forms of natural ventilation. Without significant engagement and ongoing education, the introduction of any hardware (e.g. HEPA filters, carbon dioxide monitors) will likely have limited ongoing impact on improving IAQ as the human element is equally, if not more important, than the engineering considerations (Snow et al 2022).
The health and learning capacity of students in Australian schools has been found to be negatively affected by poor IAQ (Haddad et al 2021), an issue that has gained heightened interest during the COVID-19 pandemic (Hyde et al 2021). Despite generally low outdoor air pollution levels, there are regular pollution spikes and poor IAQ may stem from outdoor pollution sources that are seasonal in nature-such as bushfires, dust storms, pollen and grass fires; or sources that prevail year-round, such as emissions from traffic and industrial pollution (Mazaheri et al 2016, Cooper et al 2020, Walter et al 2020, Carmona et al 2022. These issues are prevalent across the world, with a lack of sufficient ventilation inside classrooms, as indicated by high carbon dioxide levels (Jacobson et al 2019) often compounding this IAQ problem, with carbon dioxide concentrations greater than 1000 ppm recognised to negatively impact learning ability (Snow et al 2022). For Australia, further nuance is needed to develop a standardised response at the national level with specific attention to the significant range of climatic conditions across the continent. To accommodate these various climates, schools in the most populous states of New South Wales, Victoria and Queensland already have different levels of provision of A/C and/or heating. In the state of New South Wales, for example, only a minority of the 2200 public schools have HVAC systems, whereas the majority of Queensland public schools have retrofitted wall mounted A/C systems.
Many school decision-makers face a dilemma. They can ask staff to open windows to lower carbon dioxide levels, promote good air flow and circulation, and reduce indoor sources of air pollution such as re-circulated dust or mould spores. Increasing natural ventilation in this manner, however, can enable outdoor PM 2.5 , pollen and other air pollutants to enter buildings (AIRAH n.d.). Indeed, closing windows and doors only partly prevents outdoor PM 2.5 from ingress, with some studies showing around 60%-70% of outdoor PM 2.5 entering closed classrooms (Hou et al 2015). Furthermore, new research indicates that wildfire smoke exposures and pollen are associated with increased risk of COVID-19, further complicating these decisions (Henderson 2020, Damialis et al 2021, Cortes-Ramirez et al 2022. As in many countries worldwide, consistent guidance for Australian school decision-makers for a range of related policies, such as extreme heat, is limited (Education Victoria 2020, Queensland Government 2021, South Australian Gov 2021). In addition, a systematic and integrated approach for response to acute or chronic air pollution is lacking beyond the introduction of HEPA guidance (Education Victoria 2022). Filling this gap are informal or ad hoc practices generally accepted to improve IAQ. These practices include encouraging opening windows to allow as much outdoor air to flow indoors as weather or noise will allow. Additional benefits of increasing natural ventilation in this way can include reduced transmission of airborne viruses between children, and between children and staff. Despite these potential benefits, in practice teachers tend to keep windows closed to manage noise levels, control temperature, prevent rain from entering, or improve energy efficiency (Snow et al 2022). A holistic assessment of IAQ should also consider seasonality for temperature and humidity, which in turn affect ventilation rates (Bakó-Biró et al 2012, Wargocki and Da Silva 2015).
In the face of these uncertainties, the need to understand and improve IAQ in classrooms emerged as a critical issue at the start of 2022, when schools in Australia's most populous states returned to full in-person teaching.
Specifically, concerns about indoor transmission of airborne viruses led to a new interest in, and awareness of IAQ in schools (Curtius et al 2021, Ferrari et al 2022, Sutherland et al 2022. Australian state governments' initial responses to reduce transmission in schools during the pandemic were prompt and forceful, and resultant policies and measures varied from state to state. These multifold pandemic responses prompted many decision-makers to question how to best handle ongoing IAQ issues. Central to this question is the imperative to balance measures that reduce outdoor air pollution ingress with measures that aim to that keep airborne viral transmission levels as low as possible. When in-person teaching returned in the state of Victoria, for example, public schools were provided with HEPA filters for most classrooms. The roll-out of 51 000 HEPA units began in late 2021 and extended into early 2022. HEPA filtration removes virus particles and particulate matter (Kong et al 2022). The state of New South Wales' approach differed somewhat. The state conducted a rapid assessment of public schools' existing ventilation options, such as openable windows and ceiling fans. Unopenable windows and broken fans and were immediately fixed during the assessment (Varming 2021). Some schools were found to have insufficient ventilation, and in a small minority of new 'HVAC designed in' schools, no access to natural ventilation. These schools were provided with HEPA filters for rooms unable to conform to minimum ventilation standards. In the state of Queensland, public schools underwent audits to determine whether HEPA filtration was needed, and were encouraged to maximise natural ventilation during the audit process. Tasmania, South Australia and West Australia bought and installed 4500, 4000 and 12 000 HEPA filters respectively. In the independent school sector, a more fragmented approach prevailed. Whereas schools with sufficient funding were able to implement complete HEPA filter roll-outs, others resorted to more limited HEPA filter use, identifying and installing units in poorly ventilated spaces (ABC News 2022b).
The combination of improved ventilation, added HEPA filtration, and adoption of still more layers of protection has become known as the 'Swiss cheese approach' to reducing viral transmission (Mackay 2020). As an important co-benefit, this approach reduces exposure to many other forms of air pollution and their concomitant health impacts (and associated economic and societal costs). A key benefit of these policies is a reduced exposure to agents that trigger asthma-including pollen which is significant given Australia's high childhood asthma incidence (AIHW 2009, Aldekheel et al 2022, Kong et al 2022.
To attempt to address these issues and fill in the research gap about in situ testing, we developed a set of experiments that are broadly generalisable to a range of locations worldwide. We examined the capacity of HEPA filters to remove PM 2.5 in its own right, and, as a proxy for airborne viruses in a classroom setting and compare this ability with other interventions including natural ventilation and recirculating A/C. We consider whether natural ventilation is likely to be effective enough to reduce PM 2.5 in a classroom populated with unmasked students. We also explore the relationship between carbon dioxide and PM 2.5 concentrations across these different interventions.
This research was approved by UNSW Human Ethics, approval # HC220325.

High efficiency particle air filter comparison
We tested four HEPA filters that were representative of classroom-sized units that could be bought and used in Australia as of Sept 2021. To replicate some of the state government procurement guidelines, the unit needed to weigh less than 20 kg, have replaceable filters without the need for tools, no ionisation or electrostatic precipitation and a CADR acceptable for a use in a standard sized 60-90 m 2 classroom. The comparative details of the models tested are provided in table 1. This table identifies some differences between the HEPA filters that may be of interest to schools. Specific concerns relate to noise at maximum setting, external casing material, and Wifi (remote) control. We used an ultra-low volume (ULV) Longray 3600E fogger with 0.9% saline solution to generate particles with an initial diameter of 5-20 µm, to seed a 2 m by 2.2 m by 1.9 m chamber as shown in figure 1. The HEPA filter placed inside the chamber was turned on after two minutes and the clearance rate for PM 2.5 and PM 10 was measured with a TSI 3330 optical particle counter that sampled every 30s. This process was repeated for each HEPA filter.

Comparison of classroom interventions to remove PM 2.5 from saline spray
The 67.3 m 2 ground floor classroom has a ceiling height of 2.7 m. This classroom had two mid wall windows, each with an openable area of 0.61 m by 1.53 m, and an external door with an openable area of 0.86 m by 2.48 m on the east and north wall respectively as shown in figure 2. There were three 0.19 m by 0.1 m vents on the outside facing east wall, each with openable area of 50%.
The location of the sensors around the room are detailed in figure 2. In the centre of the classroom, an Aranet4 and a Netatmo were used to record carbon dioxide levels and temperature respectively. Particulate matter (number and mass) was recorded on a TSI 3330 particle counter alongside a Clarity Node-S on a table approximately 0.5 m from the ground, with five other Clarity Node-S distributed around the room on tables approximately 0.8 m from the ground. Clarity Node-S are a lower cost sensor that have been calibrated for PM 2.5 and we distributed them around the room to explore if there were significant differences in spatial distribution of PM 2.5 during the experiment.
To reduce residue remaining on the classroom furniture from the saline spray, we covered all surfaces with plastic sheeting. The ULV fogger with 0.9% saline solution was run for three minutes prior to each intervention as detailed in table 2. Each intervention was run for around 30 min. Between each intervention, the windows and external door were opened, and three HEPA filters were turned on full power for 15 min to clear the room.

Comparison of interventions to remove PM 2.5 from unmasked people
The next series of tests were designed to better understand how well a range of interventions could remove PM 2.5 from volunteers' exhalations, and to enable us to compare these levels to carbon dioxide concentrations. We reduced the interventions tested with volunteers in the room given it was clear from the previous saline tests that when the outdoor air pollution levels were low, cross ventilation was the most effective way to reduce indoor air pollution levels. Anecdotally, however, we know that teachers would very rarely be prepared to teach with the window and doors open (even if cross ventilation does occur between classes as people exit and enter the open door while windows remain open). Further, this classroom's A/C system did not work effectively enough to reduce PM 2.5 to any appreciable degree as so we excluded it from further testing. The interventions we tested were: one HEPA, three HEPA, Stale Air and Natural Ventilation, as shown in figure 7. Because we were using a classroom that had HEPA installed already, we conducted one intervention with the HEPA filters installed 'as is' i.e. as we found it located in the corner of the classroom. We created the three HEPA test (where we brought in two more HEPA filters) to see whether there could be a benefit to adding more units around the room to increase the CADR to reduce indoor PM 2.5 . Time constraints and low outdoor PM 2.5 levels on the testing day prevented any further intervention combinations.
The simplified the set of interventions with the exhalation from unmasked volunteers (average 20 years old) are shown in table 3. Although school-aged children breathe, on average, slightly faster than young adults (18-30 breaths per minute compared to 12-16 breaths per minute), children have smaller lungs and lower lung volume, and as a consequence, exhale less total volume than adults so we did not expect this difference in breathing rate/exhaled lung volume to significantly impact our results. The same layout and approach for measurement of PM 2.5 , temperature and carbon dioxide was taken for this set of interventions.
Twenty-two volunteers sat distributed around tables, as shown in figures 3 and 4. During each intervention, the volunteers participated in spoken games to standardise the quantity and location of exhaled PM 2.5 and carbon dioxide to attempt to replicate students' exhalations during a usual classroom activity. Each intervention, detailed in table 3, was run for approximately 30 min. Because the volunteers' exhalations were the main source of indoor PM 2.5 in these interventions (unlike the saline spray test where there was no ongoing source of PM 2.5 ) we did not get the volunteers to speak prior to the start of the intervention. Between each intervention, the room's air was refreshed by fully opening all windows and external door.  Figures 5(a) and (b) shows how effectively the HEPA filters cleared the normalised PM 2.5 and PM 10 respectively in the laboratory chamber. This figure shows that all the units were effective at clearing these size fractions, however, two units with higher stated CADR were faster at clearing both fractions in the first two minutes, although by the fifth minute, all units had cleared the particles to a similar degree. The normalisation was based on the maximum number concentration value for each case. For all four tests, the maximum PM 2.5 number concentration in the chamber was 2868# cm −3 with a standard deviation of 72# cm −3 and maximum PM 10 number concentration was 3017# cm −3 with a standard deviation of 94# cm −3 . These results indicate that the use of any of these models in our classroom testing would be broadly representative of a HEPA filter that might be procured to be used in a school.

Comparison of interventions to remove saline spray
The tests occurred between 10 am and 4 pm on a still, rainless day. A Clarity Node-S monitor installed outside of the classroom, but within the school grounds, recorded an average PM 2.5 number concentration of 5.16 cm −3 during the period of testing (max: 7.54 cm −3 ; min: 2.01 cm −3 ; median: 5.00 cm −3 ; IQR: 1.74 cm −3 ). The average temperature and humidity recorded by an outdoor Netatmo weather sensor during testing time was 19.4 • C and 80% respectively. All the interventions shown in figure 6 indicate good agreement of the trends between the optical particle counter and the Clarity Node-S PM 2.5 readings. The carbon dioxide level remained around 500 ppm during, and between, all interventions for both the Aranet4 and Netatmo sensors. Given the close agreement between the nondispersive infrared sensors, we report only the Aranet4 given its shorter sampling interval. The stable carbon dioxide levels were expected given there was no introduction or removal process of this gas during any of these interventions.  The Stale Air intervention (a) was run for 40 min to enable the longer tail of PM decline to be recorded. This figure shows that after initial natural fall out of larger sized particles from the air, the levels of PM 2.5 remained relatively stable with a slow decline over time. The highest levels of PM 2.5 were seen after 40 min in this intervention. Panel (b) shows the Cross Ventilation intervention which had the fastest decline, and lowest final levels, of PM 2.5 after 30 min. Both the use of one HEPA, panel (c) and three HEPA, panel (e) showed its effectiveness of PM 2.5 removal. By 30 min, both interventions showed effective removal of PM 2.5 , albeit a faster decline and to a slightly lower level in the three HEPA intervention. The Natural Ventilation intervention, panel (d) was slightly more effective at removing PM 2.5 than either the one or three HEPA intervention after 30 min. It is important to remember that this intervention (as would be the case for the cross-ventilation intervention) was undertaken when outdoor levels of PM 2.5 were low. This result would likely not have been replicated if outdoor air pollution levels were higher. Finally, the recirculated A/C intervention, panel (f) showed the least effective PM 2.5 removal after 30 min other than the Stale Air intervention. In this case, PM 2.5 levels dropped slowly and showed slightly greater reductions compared to the Stale Air test (a).

Comparison of interventions to remove PM 2.5 exhaled by unmasked people
The tests occurred between 10 am and 4 pm on a still day with no rain. A Clarity Node-S monitor installed outside of the classroom, but within the school grounds, recorded an average PM 2.5 number concentration of 1.16 cm −3 during the testing time (max: 1.62 cm −3 ; min: 0.54 cm −3 ; median: 1.12 cm −3 ; IQR: 0.4 cm −3 ). The average temperature and humidity recorded by the outdoor Netatmo weather sensor during testing time was 18.4 • C and 92% respectively. The PM 2.5 number concentration reading was notably lower than the previous saline spray testing day but the temperature and humidity readings were similar to those observed during the previous testing day. For the intervention with one HEPA in the classroom (a), it took just under 13 min for the carbon dioxide level to reach 1000 ppm. At the end of the testing period this level was over 1500 ppm. Both kinds of PM sensors picked up a declining trend of PM 2.5 during the testing period. When we added an additional two HEPA filters, for a total of three HEPA in the room (b), we saw a very similar trend for carbon dioxide build up over the same time, however, we saw the increased effectiveness of PM 2.5 removal within the first few minutes of the testing period.
The Natural Ventilation intervention (c) is the only treatment for which carbon dioxide levels did not breach 1000 ppm (or 800 ppm-a suggested guideline for reducing COVID-19 transmission) during the test. The sensors identified a slight decreasing trend of PM 2.5 over the course of the intervention. The final test, that of Stale Air (d), showed a marked increase in PM 2.5 captured by both types of sensors over the test. The build-up of carbon dioxide was the quickest in this intervention, with the 1000 ppm threshold breached in nine minutes. Figure 7 compares results between the removal of PM 2.5 presented in figures 6(b) and (d) (cross ventilation and natural ventilation with saline spray) and 8 (c) (natural ventilation with human exhalation), normalised to outdoor concentrations of PM 2.5 . This figure shows the difference in concentration of PM 2.5 between the indoors and outdoors for the saline spray experiments and the much smaller difference between the indoor and outdoor PM 2.5 levels for the human exhalation experiments. This figure is useful to explain why there was an apparent difference between the gradient of the slope for the removal of PM 2.5 between these experiments. In the human exhalation interventions, the number concentration of indoor PM 2.5 was lower, and the number concentration outside was also low (with a measured PM 2.5 of 1 cm −3 ). That is, there was not a strong gradient of PM 2.5 between them, and consequently, there was not much transfer of indoor PM 2.5 outside. This result contrasts with what was seen in the saline spray interventions where the indoor number concentration was initially extremely high and the outdoor number concentration of PM 2.5 was 5 cm −3 .

Limitations
This study does not account for classrooms that have HVAC systems with fresh air intakes, nor did we test rooms with fans close to windows that could further increase the removal of indoor air to the outside. Any significant difference between indoor and outdoor temperatures and/or wind speed (and direction) would also likely impact the amount and direction of flow of air through classroom windows (specifically lower winter temperatures would likely increase the natural ventilation rate). If there were high levels of PM outside the window, or other air pollutants we did not test for, such as ozone, different assessments would need to be considered. We did not test DIY filters and recognise that the HEPA filters available in different countries have different dimensions, direct air flow differently and consequently are likely to produce different results. The tested HEPA filters running on their maximum setting produce noise that is significantly above a level considered disruptive to teaching in other countries, for example in Germany-where these levels are set at 40 dB (IRK 2020). Running the tested HEPAs at less that maximum settings to reduce noise levels would have reduced their air cleaning efficiency and we did not test at these levels. This issue is complicated by there being no certification process for HEPA filters at the national or international level to enable standardisation.

Discussion
This research is predicated on the fact that we have very limited knowledge of levels of IAQ in classrooms. In the Australian case, building age and maintenance problems, especially with recent extreme rainfall in the populated south and east of the country has led to many buildings suffering black mould issues, and these issues have compounded existing ingress of outdoor air pollution and natural ventilation problems. The general advice to keep windows open 'as much as possible' unless a known high level pollution concentration is outside (e.g. a highly trafficked road, pollen count warnings etc) frequently conflicts with advice to close the windows to make HEPA filtration more efficient-in an attempt to reduce transmission of airborne viruses. Further, factors encouraging windows to be closed are increasing. Some of the most recent include extreme heat, extreme rainfall or dangerous levels of bushfire smoke, as well as outdoor noise.
Despite these concerns, school principals need clear and simple guidance about how best to protect the health of their students and staff from indoor and outdoor air pollution. Tension can occur when there is poor quality air outside a school and closing windows to prevent this air entering classrooms conflicts with advice to increase natural ventilation to reduce viral transmission and dilute carbon dioxide levels. The analysis presented here shows how important it is to refresh the air in classrooms with outdoor air, ideally between every class-unless the outdoor air quality is known to be poor and further steps are needed to improve air quality. These experiments contribute to the literature by providing evidence regarding the performance of HEPA filters in classrooms designed to be naturally ventilated, and the comparative performance of HEPA filters and natural ventilation for reducing PM 2.5 and carbon dioxide. Future research directions could include assessing the intervention of natural ventilation and HEPA filtration on high PM 2.5 days and also the location of HEPA in the room during these times. This assessment would provide policy makers evidence to identify the best protective guidance to give and enable more rigorous public health advice during episodes of high outdoor pollution, such as during El Nino events, or when schools are located in areas of high traffic or near industrial sites. It also makes a case for routinely monitoring IAQ in classrooms.

High efficiency particle air filter comparison
Our chamber test comparison of classroom-sized HEPA filters that broadly met the procurement guidelines used by the Victorian and NSW state governments for use in public schools identified a range of suitable products that were very effective at reducing PM 2.5 and PM 10 . We found no major differences between the models tested in their ability to remove PM 2.5 and PM 10 . Of these larger sized HEPA filters, the guides identified units that would all provide good PM 2.5 removal (and using its removal as a proxy for removal of airborne viruses (Ueki et al 2022)). These filters would be beneficial to help reduce the concentration of bushfire smoke, traffic pollution, dust, mould, pollen and airborne viruses in classroom air. Anecdotal evidence suggests that the features affecting their ongoing use in a classroom setting relates to noise levels and the inclusion of a 'set and forget' feature. Informal feedback with teachers and associated representative bodies identified that the likelihood of a teacher turning off units increased with higher perceived noise levels. A work around for this problem would be to install two (or more) smaller units in each room. This approach would enable the necessary CADR to be achieved, whilst having an overall lower noise level. Units that can be set with start and stop times via a remote access app or on machine setting can help to remove the need for teachers to interact with the HEPA filters. Given the relative novelty of HEPA filters, a carefully developed education campaign about the benefits, and constraints, of their use would be extremely beneficial with specific education and ongoing support for facility managers about how to use and maintain HEPA filters. For example, ensuring that these machines' prefilters are regularly cleaned to extend the life of the non-reusable HEPA filter is an important and frequently overlooked maintenance practice that needs to be promoted.

Comparison of interventions to remove saline spray
The saline spray interventions highlighted just how effective cross ventilation can be to reduce PM 2.5 levels in a classroom-with the proviso that there are low levels of outdoor PM 2.5 . For schools in areas which generally have good outdoor air quality, the use of cross ventilation in classrooms during and between each class has the capacity to reduce levels of a range of air pollutants such as PM 2.5 , mould and re-circulating indoor pollutants as well as airborne viral particles and carbon dioxide. If, however, outdoor pollution levels are elevated, relying on cross ventilation alone is a suboptimal strategy. Further, during class-time, the cross-ventilation option may be available for some classrooms in more rural or remote settings in good weather conditions-assuming that noise travelling between classrooms is not a concern. During temperature extremes, heavy rain or medium or high air pollution days (caused by bushfires, dust storms, commute traffic or high pollen counts), this approach alone is not feasible-even for these schools (Duill et al 2021). It is important to emphasise the need to purge the indoor air between classes with cross or natural ventilation irrespective of the conditions outside the window if carbon dioxide levels are not to build up during the day. HEPA filtration can serve an important role in this situation in reducing indoor levels of pollutants in this 'fresh' air. The A/C unit tested in this experiment was not effective at reducing PM 2.5 . Although some newer units may have slightly better filtration, they would be unlikely to be rated at levels that could reliably remove smaller particulate matter, and so should not be considered as a mechanism to reduce indoor air pollution.

Comparison of interventions to remove PM 2.5 exhaled unmasked people
The interventions that used unmasked volunteers as a lower, albeit ongoing, source of PM 2.5 had the benefit of adding a realistic build-up of carbon dioxide. This set of interventions allowed us to compare how the PM 2.5 and carbon dioxide concentrations varied over time and between different interventions. These interventions clearly indicate the potential inadequacy of relying on carbon dioxide levels as a proxy for 'safe air' (Stabile et al 2021). These interventions demonstrated how quickly carbon dioxide concentrations increased in a room with stale air. Even with the additional mixing of leaked window air provided by HEPA filters, by around 12 min, carbon dioxide concentrations had surpassed 1000 ppm. These carbon dioxide concentrations can lead to fatigue, headaches and reduced ability to concentrate. This last point highlights the need to be cautious about relying on carbon dioxide monitoring to identify the risk levels of airborne transmission of viruses in filtered air. This message needs to be more strongly emphasised, and as facility managers or asset service officers become more aware of the nuances of IAQ, it might be beneficial for them to monitor and then respond to PM 2.5 and carbon dioxide levels.

Summary
We present a case for investing in a range of interventions needing financial and human resources in the short term to improve the health of students and teachers and reduce the costs associated with chronic and acute disease that would otherwise result. These actions would have significant flow on benefits to the wider population and the economy via reduced viral transmission in families and potential to avoid missed days of work to look after sick children, or from parents' own ill health. For classrooms that do not have existing HVAC mechanical ventilation with fresh air, cross ventilation during low air pollution days followed by a combination of natural ventilation with HEPA filtration is likely to be the best overall solution to ensure good IAQ with respect to particulate matter, gases and airborne viruses. In colder climates when windows are opened less, HEPA filters can work to supplement reduced natural ventilation.
HEPA filters cannot, however, replace ventilation with fresh air, that is they cannot be used as a stand-alone, independent measure. Further, significant investment needs to be made in ongoing education and maintenance programs to ensure they are used correctly. Similarly, an over-reliance on natural ventilation for infection control is problematic because there are limits on its use when outdoor air pollution levels are high. The complementary nature of natural ventilation with HEPA filters is likely to provide the most beneficial outcome for ensuring good IAQ at reasonable cost given our current understanding. The additive effect of this approach-i.e. natural ventilation and HEPA filters could likely be further enhanced if multiple smaller sized HEPA units are used in a room to reduce their noise levels and increase the chance that they can are located close to an infection source.
These results are focused on locations that have previously been neglected, however, the results are transferable elsewhere for locations where there are multiple people in rooms that may well be under ventilated such as meeting rooms, general practitioner's waiting rooms, restaurants or shared office spaces.

Data availability statement
The data cannot be made publicly available upon publication because they contain sensitive personal information. The data that support the findings of this study are available upon reasonable request from the authors.