The study conducted by TRANSCOM was performed in an actual Boeing 767 aircraft with 37 rows. Measurements were collected within 8-10 rows of the injection source location. The system was moved to other cabin sections for subsequent trials. The Airliner Cabin Environment Research (ACER) laboratory study was conducted in a Boeing 767 cabin mock-up of 11 rows. The mock-up cabin was constructed using parts from an actual salvaged Boeing 767 aircraft including the passenger seats, the cabin supply air system, linear diffusers, and personal ventilation gaspers. In addition, actual Boeing 767 cabin cross-section and outline were carefully considered while designing the mock-up to accurately resemble the aircraft cabin geometry.

The ACER cabin mock-up is enclosed within a 24.3 ft x 32.2 ft by 16 ft (7.4 m x 9.8 m by 4.9 m) wooden enclosure as shown in Figure 1. The mock-up has a total inside length of 30.84 ft (9.41 m) and width of 15.48 ft (4.72 m), and a total volume of 2366 cubic feet (67 cubic meters). All measurement instruments were located in two hallways with the sample lines in the cabin leading out to the right and left sides of the mock-up cabin. Two hallways on each side of the mock-up allow access to the outside of the cabin walls and provide space to house equipment and instrumentation. This void, including hallways and areas above and below the mock-up chamber, serves as an exhaust air plenum for the cabin. Once exhausted from the cabin mock-up, air is pulled from the plenum through the exhaust fans. Figure 2 shows an inside view of the cabin. The cabin has 11 rows of seats numbered 32 to 42 to match the US Transportation Command (TRANSCOM) study and distributed longitudinally with 7 seats in each row. The seats are in a 2-3-2 configuration, separated by two aisles as designed in a standard Boeing 767 aircraft. The rows are numbered from the front of the cabin at the front wall to the back of the cabin at the rear wall. The 7 seats columns are labeled A through L from left to right. This naming style was adopted to match the one used in the TRANSCOM study (Kinahan et al., 2021) which was provided to them by United Airlines.

The study conducted by TRANSCOM was carried out in a functional Boeing 767-300 aircraft. The interior dimensions of the cabin cross-section, as reported in the manufacturer’s manual (Figure 3), are identical to the ACER mock-up cabin. Figure 4 shows the cabin during the TRANSCOM study. The aircraft, operated by United Airlines, normally has 37 rows of seats divided into 3 different cabin classes with a total cabin volume of 9320 cubic feet (264 cubic meters) (United Airlines, 2022). The business class (section) has 10 rows of seats with 30 seats total. The economy plus (FWD-MID section) has 7 rows of seats with 46 seats total. The economy class(AFT section) has 20 rows of seats with a total of 138 seats. Hence, the maximum capacity of the aircraft is 214 passengers.

The ACER mock-up cabin, with 77 seats, was supplied with 1400 CFM (40 m³/minute) of conditioned air. Fresh air is drawn in from outside and passed through the dehumidifier unit into the air conditioning unit, which maintains the supply air at 60 °F (15.6 °C). The air is then delivered to the cabin through the main supply air duct running in the top centre, along the cabin length. The supply air is then delivered to the linear diffusers that run along the cabin interior length as shown in Figure 2 in the same configuration as the Boeing 767 aircraft. The air is exhausted at the floor level along the whole length of the cabin through the side walls to the side hallways. The supply air system achieves 35 ACH and keeps the average age of particles in the cabin at about 1.7 min. For these tests, the dehumidifier was not used. Additionally, only fresh single-pass air was used for these tests. Since HEPA filters are normally used on recirculation air, the air is believed to be pathogen free. However, tracer gas would not be removed by these filters if recirculation were employed in the mock-up. Therefore, using 100% outside air is the appropriate configuration for using tracer gas to accurately simulate pathogen dispersion.

For the tests conducted in the TRANSCOM study, the reported air exchange rate for the tested Boeing 767 airframe was 32 ACH, or 4979 CFM (141 m³/min). This yields an average age of particles in the cabin of about 1.9 min. The environment control system (ECS) achieved approximately 50% of the air exchange through HEPA-filtered recirculation, and 50% through fresh bleed air. The cockpits and cabins had separate supply systems with no mixing between them. The air was supplied to the cabin at a temperature between 51.5 and 67.0 °F (10.8 °C and 19.4 °C).

The study conducted by ACER utilized a tracer gas source and several gas analyzers to record the gas concentrations in radial directions at different distances away from the source. Tracer gas was introduced as an aerosol surrogate at the breathing level at different hypothetical contagious passenger locations and sampled radially up to 3.35 m from the source for 30 min per location. Two different tracer gas injection methods were used. For the first method, a 1 inch (25.4 mm) inner diameter copper tube was used as a continuous injection source to represent a breathing source. For this source type, the tracer gas consisted of CO₂ injected at 0.18 CFM (5 L/ min) mixed with He injected at 0.11 CFM (3.07 L/min) to achieve neutral air buoyancy. The second tracer gas injection method was an automated coughing manikin designed to simulate a coughing passenger using a CO₂ tracer gas. The coughing manikin has a 0.15 ft³ (4.2 L) CO₂ gas coughing volume and a peak flowrate of 24.16 CFM (11.4 L/s) with the entire cough lasting approximately 1 s.

A sampling tree was used to decrease the number of required gas analyzers needed to sample the tracer gas at different locations inside the cabin. The sampling tree is an 11 ft (3.4 m) metal rod composed of 4 equidistant sampling ports as shown in Figure 5 and is used to collect samples at 4 different locations sequentially. The seats in the mock-up were equally spaced as shown in the cabin schematic in Figure 6 allowing for the usage of the sampling tree to measure at each consecutive seat row. Each port is controlled with a normally closed (NC) SMC Pneumatics NVKF334V-3G 2-way solenoid valve connecting each of the 4-sampling ports to a common outlet port that was connected to the inlet of a single CO₂ analyzer. The ports were opened sequentially, one at a time, allowing for isolated sampling of 30 min in each location. Each sampling line was made of 304 stainless steel welded tubing with an inside diameter of 0.2 inch (5 mm). A fifth normally open solenoid valve was mounted to the manifold in reverse orientation. This allows the tube connected to the common rail of the manifold on one side and to the CO₂ analyzer on the other side to constantly be pulling a sample through to the CO₂ analyzer when all the sampling ports were closed. For each tracer gas injection location, the sampling tree was positioned in several orientations around the source. Several non-dispersive infrared (NDIR) gas analyzers (WMA-4 and WMA-5, PP Systems International, Inc., Amesbury, MA, USA) were used to sample the tracer gas inside the cabin. Three WMA-4 analyzers were calibrated to a full-scale of 2000 ppm and have an analog voltage output with the range 0-5 V which corresponds to 0-2000 ppm of CO₂. The fourth WMA-4 analyzer and the WMA-5 analyzer were calibrated to a full scale of 3000 ppm with the range 0-2.5 V which corresponds to 0-3000 ppm of CO₂. A sixth gas analyzer was a custom-made Edinburg Gascard analyzer with a 0-2000 ppm range and was used to measure the tracer gas concentration in the supply air. Each gas analyzer was operated at 1 lpm. The basic principle for all three analyzers is the same. The analyzer output signals were digitally converted using an Agilent 34970 A DAC. Table 1 provides further instrumentation details.

A typical test in the ACER study was conducted in the following sequence:

The TRANSCOM study was conducted in four days with two days reserved for ground testing and two days reserved for inflight testing. Of the two ground testing days, one was reserved for simulating inflight conditions on the ground to achieve more replicates in additional seats without the need to fly the aircraft. Hence, the only difference between hangar and inflight tests in the TRANSCOM study is the elevation at which the tests were conducted. The focus of the TRANSCOM study was to look at the potential for super spreader events that would result from aerosol transmission. In the TRANSCOM study, Fluoresbrite Plain yellow-green (YG) polystyrene latex (PSL) microspheres (Polysciences) sized at 1 lm were generated using a PulmoMate (fluorescent tracer). The flowrate was approximately 6 lpm (S. Kinahan, personal communication, June 6 2022). To sample the fluorescent tracer particles, a suite of instantaneous biological analyzer and collector (IBAC, FLIR Systems) discrete particle detectors was used. The instrument samples at 3.5 lpm, and reports tracer counts per second. The fluorescent tagged microspheres were generated for one minute in a breathing pattern using a timing circuit for 2 s on and 2 s off. The total number of released particles was 180 million particles for both breathing and coughing tests. The output of the nebulizer cup (Hudson Micro Mist) was plumbed through a tripod mounted manikin head and reached a velocity of 1.43 m/s at the manikin’s lips for breathing tests and 12.84 m/s for coughing tests. Table 2 shows the number of tests included in this paper from both studies. While both studies included a larger number of conducted tests, only the tests with matching source location, and experimental conditions were selected for comparison and listed in Table 2. For example, the TRANSCOM study included terminal testing conditions with the jetway attached to study the exposure levels during passengers loading and loading. In addition, a ground air cart replaced the aircraft ECS in some tests to study different supply air mechanisms. The terminal tests configuration from the TRANSCOM work was not considered since the hangar and inflight tests are with the testing conditions that closely match these in the ACER tests. To conduct a realistic comparison, only the tests without the source mask from the TRANSCOM work were considered in this study. Furthermore, since the TRANSCOM study (Silcott et al., 2020) noted that ‘overhead gasper supply (on or off) does not make a significant impact on aerosol risk’, tests with gaspers on and off from the TRASNCOM study were considered.

Thermal manikins were used in the ACER study to simulate passengers and cabin thermal load. Each manikin was wrapped with 10 m wire heater elements to generate about 102 W of heat to account for the average sensible thermal output of a resting adult, 70 W (ASHRAE, 2008), in addition to heat from inflight entertainment systems, avionics, and personal electronics such as laptops. On the other hand, cabin heat load was not considered in TRANSCOM tests performed in the hangar and inflight.

To establish a similar basis for comparison, some pre-analysis procedures were applied to each dataset. For the ACER data, the first step was to express the sampled tracer gas concentration data in terms of potential exposure as a percentage of mass released (%) rather than tracer gas concentration (ppm). The second was to normalize the data to a typical resting passenger inhalation flowrate of 7.5 lpm similar to the TRANSCOM study rather than the instruments sampling rate. Eq. (1) was used to convert the ACER data to potential exposure percentage and to normalize the data to a resting passenger inhalation flowrate of 7.5 lpm.

The TRANSCOM dataset was expressed in terms of potential exposure percentage versus distance from the source rather than seat number to establish a source-distance decay comparison. To do so, a CAD model of a Boeing 767 cabin was used to measure the distance between tracer particles release and collection locations as shown in the appendix in Figure A1. Airlines have control over modifying the interior of their aircraft cabins including seat pitch and width. The dimensions of the United Airlines Boeing 767 cabin seats were obtained from United Airlines website (United Airlines, 2022). Inspection of Eq. (1) shows that the ratio of the inhalation rate of a person or the intake rate of a sampling device to the total emission rate of the tracer into the space will result in small values when expressed in units of percent. These low values do not indicate low risk, and this is easily illustrated by considering that the tracer quantity would have much larger values if reported in units of ppm. Therefore, small differences in percent exposure can fall above or below important exposure thresholds and should not be thought of as being ‘within the noise’ of an experiment.

Table 3 shows a summary comparison between the two studies considered in this work.

To study the effect of the measurement direction around the source on the exposure inside the aircraft cabin, the measurements collected at the same distance and direction in the two datasets were compared. Since a sampling tree was used for the ACER dataset, the tracer gas measurements in each test were made in a single radial direction from the source. The sampling tree was rotated by multiples of 45° from the source to cover all possible radial directions. The TRANSCOM average exposure percentage at each distance and direction from the source was obtained from the raw data reported by TRANSCOM. The average exposure was calculated by averaging the exposure percentage of the measurements in the same sampling tree directions used in the ACER study. Figure 7 shows a sample comparison of the exposure percentage at the matching distance and orientation from the source (located at distance zero) between the two studies. In this figure, each sampling tree direction test in the ACER tests was repeated twice while each direction in the TRANSCOM dataset included three test replicas. In both studies, a breathing source was located in seat L37. The insert figure on the left in Figure 7 shows the five sampling tree orientations from the ACER dataset which were considered as basis for comparison. The insert figure to the right shows the matching measurement directions from the TRANSCOM dataset. The legends show the five measurement directions for each dataset. The solid legend lines with square markers represent the ACER dataset, and the dashed lines with circle markers represent the TRANSCOM dataset. By filtering the TRANSCOM dataset, only the measurements collected in those five directions, and at distances matching those used in the ACER dataset were considered. A similar comparison was constructed for all breathing and coughing source locations with the source in window and middle seats.

Figure 7 shows the source-distance decay model for both tracer types. The lines connecting the average measurement for each tracer type are to demonstrate the source-distance decay trend and not a curve fit trendline. The figure shows the general exposure decay away from the source at all orientations for both tracer types. The exposure in the lateral direction was the best match between the two datasets. For the rest of the sampling tree directions, exposure is better matched between the two studies in the farfield compared to near the source.

Figures 8–10 show the average exposure percentage versus distance from the source for both ACER and TRANSCOM datasets without regard to measurement directions. The experimental sets in Table 2 are divided into two general categories: tests conducted using a breathing source and tests conducted using a coughing source. Each category is further broken down to two additional categories. Namely, tests conducted with the source in a window seat and tests conducted with the source in a middle seat. For this analysis, both datasets were filtered to include all the measurements collected at all distances in all directions from the source.

Factors that could explain variations in gas exposures and aerosol dispersion characteristics in the two datasets include inherent differences between tracer gases and solid particles, along with other experimental conditions. The TRANSCOM study reported a maximum deposition rate of 0.005% of the emitted particles, scaled to a square foot for tests conducted in the inflight breathing mode in the Boeing 767 aircraft. For the same dataset, the minimum, average, and maximum exposures were 0.0003%, 0.0041%, and 0.0114%, respectively. The small numbers in and of themselves do not necessarily indicate low risk. On the other hand, due to the nature of gases, deposition was not a measurable quantity in the ACER study. Furthermore, the CO₂ tracer gas was mixed with helium to achieve neutral buoyancy in the cabin air. The particle deposition effect clearly was a major factor in the exposure values being higher in the ACER study than in the TRANSCOM study.

The ACER tracer gas study and the TRANSCOM study were conducted by independent groups. Data from the two studies were collected in a cabin with 11 rows at ground level (ACER) rather than a real operating plane (TRANSCOM), sometimes at ground level and sometimes in flight. Actual flights would include solar radiation, ambient air conditions, cabin pressurization and aircraft motion. Still, the TRANSCOM study deviated from actual passenger flights in not accounting for passenger thermal plume effects in tests conducted inflight and in the hangar, while acknowledging that ‘heating vs. cooling modes had the potential to drive airflow direction differently.’ The TRANSCOM study did use some heating blankets in their terminal testing mode, but these data are not considered in the current paper because some testing conditions differed greatly, such as the cabin door being open.

Experimental errors in each study could also explain differences in the reported exposure data. In the TRANSCOM study, in addition to the variability due to the statistical nature of particle movement, instrument error in the number of particles counted for the release and the seat sampling would of course directly affect the calculated exposure, although the current study presumes no knowledge of such errors, if any. Fluorescent particles were counted over an interval of six minutes. Based on the average residence time in the cabin, this might not be long enough for all the released particles to be either captured by the instruments, deposited on surfaces, or removed in the exhaust. Experimental procedures from the ACER study could also have created errors that influenced the results. Those might include a measured CO₂ injection flowrate lower than the actual, a CO₂ gas analyzer reading higher than the actual, and a supply airflow measurement lower or higher than actual lab conditions. Any of these possible errors would have affected the exposures, as source injection percentages. A total of 54 experiments were conducted, and a detailed error analysis is reported in Mahmoud (2021). Figure 12 is included to summarize the uncertainty at the six positions along the sampling tree. Noteworthy is how the uncertainty was highest where the tracer gas measurements were lowest—farthest from the source.

While all the above factors could affect the comparison, the dominant factor likely is the inherent variability of the dispersion phenomena. From one test to the next for the same location and conditions in the ACER study, variations in results were far larger than could be explained by experimental errors were common. Additionally, for a given location, the concentration profile could be very different in different directions, especially front to back even though the cabin ventilation flows were designed to be balanced longitudinally and there are no intended net longitudinal flows. Additionally, smoke visualization was used to qualitatively study the flow patterns in the cabin mock-up (Shehadi et al., 2018). This visualization showed that the patterns are chaotic and unpredictable– unlike the animated flow arrows and laminar descriptions in some popular media stories (Grondahl et al., 2021; Read, 2020). For example, complete reversal in direction locally is not uncommon and can last for many seconds, even minutes. Local flow patterns, particularly those in the vicinity of the source can have a large impact on the local dispersion of tracers. Additionally, the visualization shows that the injected markers tend to flow in plumes rather than being uniformly dispersed. Given this plume nature of the dispersion, small differences in location can have large effects on the concentrations of injected markers. A cabin study emphasizing long measurement durations (Grote, 2023) observed extreme variation in tracer results among tests of the same injection types, notably in the continuous injection regime. These behavioral shifts in carbon dioxide response provide some evidence for the presence of what may be called metastable flow structures. For these reasons, it is unrealistic to expect close agreement between small numbers of tests conducted in different cabins at different times or in different facilities. However, general comparisons about the nature of the dispersion can be made.

The overall uncertainty, expressed as standard error of the time series average, in the ACER tracer gas samples ranged from 6% closest to the source to 20% at the farthest sampling points. The 42 IBACs used in the TRANSCOM study were pre-calibrated against a referee IBAC, and the fluorescent particle counts matched to within an average variance of ±10%. These samples had a standard error of the mean of 0.0014 percentage points in the economy cabin sections and 0.0010 percentage points in first class, as reported in Kinahan et al. (2021). The standard error of the mean of the number of particles released (1.8x10⁸), was 1x10⁷, and the standard deviation in the number of released particles was 1.7x10⁷. The TRANSCOM study did not report the overall uncertainty, which includes factors such as the supply air flowrate. Yet, the error could be within uncertainties expected in simulating particle dispersion using tracer gas, according to Ai et al. (2020).

Despite both studies utilizing coughing manikins with a similar cough release velocity, the dispersion pattern and the calculated exposure percentage differed by tracer type. To achieve the cough velocity, TRANSCOM used a mouthpiece that accelerated the flow from the breathing velocity up to the coughing velocity, while using the same duration and cycle for coughing as for breathing: one-minute emission duration, cycling as two seconds on and two seconds off. Conversely, ACER used a coughing manikin that produced a cough every 75 s, with every cough lasting one second. Although both studies tried to get close to simulating an actual human cough, neither method generated tracer that accurately represents particles generated during a real cough.

By comparing the dispersion characteristics between the two datasets, a few points can be highlighted. The general source-distance decay model seems to be similar for both datasets, in that tracer concentration decreases as distance from the source increases. This behavior was evident in all directions around the source, as shown in Figure 7. Notably, in some test conditions in both studies, tracer concentrations further away exceeded measurements nearby the source. An example is the measurements collected with a breathing source in the middle of the cabin, in the inflight-AFT mode in the TRANSCOM study, as shown in Figure 9. Similarly, for both coughing source locations in both studies, Figures 10 and 11 show a similar phenomenon. The cough momentum could explain increased concentrations at locations farther from the source in those tests. On the other hand, for the breathing source tests, the inherent airflow pattern inside aircraft cabins, which tends to have swirls and eddies, could direct the tracer gas in a specific direction causing local anisotropic elevated concentrations. A detailed analysis, including a smoke visualization of this phenomenon in the ACER cabin, is reported in Mahmoud (2021), and a similar phenomenon was described in other studies, such as (Silcott et al., 2020).

For both studies, measurements collected at the same distance but in different orientations around the source were not necessarily the same, nor were they the same at a single location in subsequent trials. This is typical for measurements collected nearby the source and not farther away from the source where less variation between individual measurements were recorded. Figure 7 shows that the average exposures for two equally spaced seats (F39 and F35) from the source in the TRANSCOM study are 0.033% and 0.010%, respectively. Observations in the ACER study also varied at most of the source-distance values shown. Zhang et al. (2021) used a CFD model of a seven-row Airbus A320 aircraft to look at SARS-CoV-2 spread in the cabin. They reported that the adjacent passengers and the passengers in the back two rows are affected more than the others when a source passenger breathes normally. Also, Lei et al. (2017) showed that passengers in aisle seats had a higher norovirus infection risk from fomites than others when traveling with a virus carrier in the same aircraft cabin, and the predicted infection risk from the fomite route for aisle seat passengers was 2.2 times higher than that for non-aisle seat passengers. It is worth mentioning that variations between different airframes and testing conditions could result in different dispersion patterns. Hence, a generalized dispersion model would not likely describe all the transmission scenarios.

By looking at Figures 8 and 9, the exposure amounts and distribution patterns differ for window vs. middle sources, for both datasets. The authors of the TRANSCOM studybelieve that ‘while there is a measurable difference in middle vs aisle or window seat, results show that this relative exposure risk is not generalizable and depends on location and seat throughout the cabin’ (Kinahan et al., 2021). Yet, it might be essential to quantify the impact of these exposure differences in the case of highly communicable diseases, since large reductions in local particle counts compared to the total released is to be generally expected in indoor environments.