The daily data of hourly means of temperatures and the derived onset, dissipation and duration of inversions were utilized to separate the measured outdoor PM_2.5 concentrations into two categories, namely those (a) occurring during a surface-based inversion (SBI) event and (b) those occurring during hours without an SBI (Table 4).

Following the EPA guidance, we calculated the 24-h mean outdoor PM_2.5 concentrations from hourly mean concentrations. To assess the exposure on a monthly basis we averaged over the daily 24-h means of the respective month (2^nd column in Table 4). Since inversions only existed for parts of the day (see e.g. Figure 4) and started or ended not necessarily to the full hour, means for times with and without inversions were calculated for each month based on the 5-min mean raw data that were kept after the QA/QC. These monthly means for times with and without the presence of SBI are up to about factor 2 higher than the monthly means calculated from 24-h means calculated in accord with the NAAQS (Table 4).

The difference between the monthly mean of 24-h means and the monthly means for times with or without inversions in the following tables is that the latter two cover not the full amount of hours of the months. This means that they are expressed in terms of means for the times with or without inversions. The monthly mean of 24-h average PM_2.5 concentrations is given by

PM 2.5 ¯ = Δ t no PM 2.5 , no ¯ + Δ t SBI PM 2.5 , SBI ¯ Δ t SBI + Δ t no (1)

where PM 2.5 , SBI ¯ and PM 2.5 , no ¯ are the concentrations during times with and without inversions, respectively. Furthermore, Δt_SBI/(Δt_SBI + Δt_no) and Δt_no/(Δt_SBI + Δt_no) are the ratios of the times with and without inversions to the time in the month. Note that Δt_SBI = 1 − Δt_no..

On average, outdoor air quality improved from March to May (Figure 5). Due to increasing air temperatures emissions from heating went down as spring progressed. After the onset of the wildfire season (end of May), ambient monthly PM_2.5 concentrations increased in June (Table 4). Also wind speeds increased as compared to the earlier months and picked up dusts from unpaved roads and the Yukon and Porcupine riverbanks. Nevertheless, March to June monthly mean PM_2.5 concentrations remained below 25 µg·m⁻³. Monthly mean outdoor

PM_2.5 concentrations reached 55.3 µg·m⁻³ in July due to wildfires (Table 4). This means the 24-h NAAQS was exceeded and led to unhealthy for sensitive groups to hazardous conditions on various days. In August, the rain season set on for which outdoor air quality improved. In September, outdoor air quality slightly decreased because decreasing air temperatures caused emissions from heating. Furthermore, inversion durations increased (Figure 4).

On average, monthly mean PM_2.5 concentrations were higher during surface-based inversions than at other times (Table 4). During times with and without SBI, PM_2.5 concentrations decreased from March to May as less heating was need because of increasing outdoor temperatures. In June, the wildfire season set on and led to increases in concentration during both SBIs and inversion-free times. Concentrations during time with and without inversions differed least in September (3.2 µg·m⁻³) followed by May (8.7 µg·m⁻³). The difference was largest (42.0 µg·m⁻³) in July (cf. Table 5, Table 6) which can be explained by advection of wildfire smoke.

Over the timeframe of available data, the US NAAQS of 35 µg·m⁻³ for the 24-h mean was exceeded nearly 45% of the time. The WHO recommended 24 h-value of 25 µg·m⁻³ not to be exceeded for three consequent days was violated as well.

Note -.- indicates no data.

Note -.- indicates no data available.

When looking at hourly means, PM_2.5 concentrations were higher than 25 µg·m⁻³ nearly 91% of the time during SBI and about 80% of the time when no SBI existed at Ft. Yukon. During the 2017 strong fire season in the Yukon Flats [13], 5-min mean outdoor concentrations were occasionally a magnitude or more, higher than the hourly means. The maximum hourly mean PM_2.5 concentrations over all four sites in March to September 2019 were 71.5 µg·m⁻³, 79.8 µg·m⁻³, 63.6 µg·m⁻³, 289.3 µg·m⁻³, 792.3 µg·m⁻³, 22.8 µg·m⁻³, and 66.7 µg·m⁻³, respectively. Outdoor air quality was on average better than in the record fire season of 2017 (cf. [13]).

On average over all indoor sites, a diurnal course can be detected in office/commercial buildings and at both heights for residences (Figure 6). Typically, indoor concentrations were higher during the day than at night which broadly coincides times of no SBI and with SBI (Table 5, Table 6). This finding is expected as people move around in their residences thereby re-suspending particles into the air that had already settled on the floor or other surfaces. Various daytime activities like cooking, smoking or feeding of woodstove could lead

to notable emissions of particles. In late June and July, mosquitoes become an annoyance in the Yukon Flats and are fought with mosquito pics. Mosquito pics contribute to PM_2.5 concentrations in June, July and August (JJA). These activities cannot be readily controlled nor documented due to privacy issues as well as indigenous cultural distrust of being monitored.

In residences, indoor monthly mean concentrations at 0.61 m were lowest in June and highest in July (Table 7). Indoor concentrations at 1.52 m were lowest in July and highest in April and May. On average over all residences, indoor concentrations were higher at 0.61 m than at 1.52 m height (e.g. Figure 6(a)). Averaged over all residencies, monthly mean PM_2.5 concentrations were 12.5 µg·m⁻³, 40.1 µg·m⁻³, 75.3 µg·m⁻³, and 46.1 µg·m⁻³ higher at 0.61 m than at 1.52 m in April, May July and August, respectively. This distribution can be partly explained by gravitational settling. The hourly mean values at those heights showed weak correlation (21%). These findings mean that exposure to high concentrations was on average over all residences higher when laying down than when standing or walking around (cf. Table 7).

The exposure at both levels was higher than the median of 6.65 µg·m⁻³ found by [11] for households with woodstoves in New England. The values of the lower and upper range of mean hourly indoor PM_2.5 concentrations (Figure 6) were lower than the 112 - 416 µg·m⁻³ observed by [25] in homes in Beijing, China in winter. PM_2.5 concentrations in office/commercial buildings in Ft. Yukon were on average very unhealthy for some hours on a daily basis in June and August and were even hazardous most of July. Exposure of employees, clients and customer was much higher than found for these groups in other places. Measurements in an office building in Guangzhou, China, for instance, showed only 3% and 1% of the time very unhealthy and hazardous condition in JJA [10]. In Dublin, Ireland, indoor PM_2.5 concentrations during working hours were typically below 25 µg·m⁻³ with highest values in naturally ventilated shops [9].

Note -.- indicates no data available.

Indoor PM_2.5 concentrations were influenced not only by continuous and temporal indoor emission sources, the residents’ activities and the degree of mixing related to the residents’ motions, but also by exchange between inside and outside air.

In contrast to all other months, June concentrations were lower (14.4 µg·m⁻³) at 0.61 m than 1.52 m, which can be explained by transport of polluted air from nearby wildfires into the community. Typically, in wildfire smoke, PM_2.5 concentrations decrease from the maximum values towards the ground as well as with height. Consequently, outdoor air has higher concentrations at the heights of vents than at the gaps between doors and the floor.

In comparison to outdoors, indoor concentrations of PM_2.5 were usually much larger in magnitude (cf. Tables 4-6). Monthly mean ratios of indoor to outdoor concentrations (I/O) in residences were 4.8, 5.3 and 12.2 at 0.61 m in March, April and September, respectively. Ratios were lowest in May, June and July with 1.4, 2.0 and 2.4, respectively. In August, I/O reached its maximum of 21. In August, the increased humidity and rainfall scavenged particles from the air thereby reducing outdoor concentrations while indoor emissions accumulated in empty homes. The increased humidity also leads to swelling of particles. This means PM_2.5 concentrations may decrease, when the grown particles exceed 2.5 µm in diameter. The high I/O ratios can be explained by indoor emissions from wood burning for cooking and heating, recreational smoking and use of mosquito pics. In Guangzhou and Dublin, for instance, I/O were below 1 [9] [25].

In the analysis of the impact of inversions on indoor concentrations, the question of air exchange arises. To compare indoor and outdoor concentrations of PM_2.5 during times with and without SBI the temporal lag between indoor and outdoor concentrations has to be determined first. We applied a lag-time correlation method [26] at various temporal lags using the means over the hourly outdoor concentrations and those averaged over all hourly indoor concentrations as input data. The method revealed a maximum correlation for a lag-time of 1-hour and 2 hours for residences at 0.61 m and 1.52 m heights (e.g. Figure 7), respectively. It showed a time-lag of 1 hour for office/commercial buildings.

When people are not home in their residents, the floor to door gaps, and other interior-to-exterior connections (e.g. cooking hoods, vents) have a greater influence at floor levels where air settles and is cooler than at 1.52 m. The 1.52 m level has a longer lag-time due to dead air and diffusion time.

Based on these findings and the determined onset and ending times of the SBIs within each 24-hour period (e.g. Figure 4), we adjusted the timing of indoor concentrations to account for the lag-time for the residences at both heights and the office/commercial buildings. Doing so synchronized the indoor concentrations with that outdoors for times with and without inversions.

On average over all residencies, indoor concentrations were higher during times without inversions than during SBIs at 0.61 m height in March, April and September (Table 5). From May to July, indoor concentrations at 0.61 m height were lower during times without inversion than during times with inversions. Averaged over the few available March days and all residences, the means were

33.8 µg·m⁻³ and 137.4 µg·m⁻³ under conditions of no inversions and SBIs, respectively. The mean of the few available data in September during times of SBIs was 76.2 µg·m⁻³, while not enough data existed to calculate a mean for non-inversion conditions.

These findings can be explained as follows. In April and September, heating is still and already again needed. Woodstoves release PM_2.5 and its precursor gases during operation. Typically, they are used for heating during the day and are fed once more prior to bedtime. Once all wood is burned, these emissions stop. When indoor temperatures fall below the monitor-set threshold, the furnace turns on. The emissions from diesel heating mostly leave via the chimney. Furthermore, settled particles are not re-suspended when the residents are sleeping. Consequently, PM_2.5 concentrations were lower during nighttime than daytime. Since, most inversions occurred at night (cf. Figure 4), indoor concentrations were lower during inversions than no inversion conditions in the heating season (Table 5). People are awake the majority of the time during the day when no surface inversion exists. Consequently, due to their activities PM_2.5 concentration levels were highest indoors during this time.

To examine the relation between construction types and building use, we averaged the hourly mean indoor PM_2.5 concentrations for each construction type and measurement level. Furthermore, we calculated hourly concentration means for residencies and office building by averaging over all respective sites.

Typically, hourly mean values at 0.61 m and 1.52 m heights correlated weakly (27.4%) and hardly (1.5%) for frame houses and cabins, respectively (Figure 7(b)). This means that indoor air quality of frame houses is stronger related to outdoor air quality than indoor air quality of cabins. On average, in frame houses, concentrations were higher at 0.61 m than 1.52 m, while the opposite was observed in cabins (cf. Figure 6(b), Figure 8). Daily 24-h mean indoor PM_2.5 concentrations at both heights for the greater part exceeded the NAAQS frequently (Figure 8).

On average, cabin indoor PM_2.5 concentrations at 0.61 m height were higher during times without inversions than during SBIs (Table 5). The same is true for frame houses in April and September, while the opposite was true from May to August (Table 5).

At 1.52 m height, means of both cabin and frame house indoor PM_2.5 concentrations were higher for times with than without inversions in all months (Table 6). Cabins provided the lowest exposure at both heights (Table 5, Table 6). No data were available at 0.61 m height for cabins in March and September. Only few PM_2.5 data were available at 0.61 m height for frame houses in March. This means the March values for frame houses are the same as those for residencies, namely 33.8 µg·m⁻³ and 137.4 µg·m⁻³ during times without and with SBI, respectively. The mean indoor PM_2.5 concentrations over all September data during inversions were 76.2 µg·m⁻³ at 0.61 m height.

These insights suggest that in frame houses, sleeping on top-bunks seems to be beneficial for health except during wildfires. Also cabins seem to be superior over frame houses in keeping outside PM_2.5 concentrations at bay.

Recall no measurements were made at 0.61 m height in office/commercial buildings because nobody sleeps in them. In April and May, concentrations at 1.52 m were higher in residences than office/commercial buildings both during times with and without SBI (cf. Table 6). At night, office/commercial buildings are closed meaning that particulate matter settled. Consequently, 1.52 m concentrations went down as compared to the times without SBI, i.e. during the day. The higher concentrations in residences than office/commercial buildings during times without SBI can be explained by the different kind of indoor activities, emissions sources and emission rates at home and work.

In JJA, monthly mean and 24-h mean PM_2.5 concentrations in office/commercial buildings exceeded those in residences (cf. Table 6, Table 7, Figure 8, Figure 9). For both, concentrations were higher during SBI than during times without SBI (Table 6). The reasons for this shift in concentration behavior are as follows. In JJA, residents were involved in activities related to the various harvest seasons

(e.g. fishing, berry picking, hunting). Hence, they stayed less time inside or even stayed outside of town in seasonal camps. Consequently, resuspension due to inside activities were reduced in residencies as compared to April and May, when residents spent more time inside. Absence also meant less emissions from mosquito pics as compared to offices/commercial buildings.

As aforementioned, in summer, the population quadruples due to tourists and fire fighters stationed at Ft. Yukon. Thus, office and commercial buildings have not only more public traffic in JJA than April and May, but also longer business hours to serve the increased demand. Furthermore, more emissions occur inside due to increased amounts of merchandise, use of equipment and cooling. Mosquito pics serve to hold mosquitoes at bay during the day. The increase in customers also leads to an increase in activities for restocking. During the day, customer traffic led to air exchange between indoor and outdoor air. At night, no such ventilation existed. Thus, the increased indoor emissions accumulated inside. Consequently, inside concentrations during the nighttime SBI exceeded those observed at daytime.

Various studies indicated that in Fairbanks, wood burning is the major cause for exceedances of the outdoor NAAQS (e.g. [27] [28]). Simulations with the Weather Research Forecast model coupled with chemistry packages (WRF-Chem) showed that using both low sulfur fuel and EPA certified wood stoves reduced the concentrations of PM_2.5 as compared to the old devices used in Fairbanks [28]. Therefore, we examined the impacts of temperature monitors on indoor air quality of office/commercial buildings. All residences had a monitor and burned both wood and diesel, while office/commercial buildings only burned fuel.

In office/commercial buildings without temperature monitor, monthly and 24-h mean PM_2.5 concentrations were up to more than two orders of magnitude higher than in office buildings with monitors during JJA (Table 7, Figure 9). This means that using a temperature monitor could reduce exposure of employees, clients and customers in summer. In April, office and commercial buildings without monitor have lower PM_2.5 concentrations than those with monitor. This finding can be explained by the monitor and hence furnace reacting to changes in indoor temperature caused by opening and closing of doors during business hours. However, in both cases, PM_2.5 concentrations remain in the healthy range.