3.1. Origin of the Historical Ambient Aerosol Distributions Used in ISO 16890
The urban and rural volume distributions that are used in ISO 16890 were taken directly from one of the most widely regarded textbooks on atmospheric chemistry and physics: Seinfeld and Pandis (2006) [30
]. The distributions are reported as trimodal distributions (shown in Figure 1
with summary statistics provided in Table 1
), although ISO 16890 assumes they are bimodal distributions with lognormal parameter fits, which are described in Table 2
in part 1 of the standard (i.e., just the two larger modes, ignoring the smallest mode). This assumption is considered reasonable, because the standard does not require removal efficiency measurements below 0.3 µm. Interestingly, the aerosol distributions in Seinfeld and Pandis (2006) [30
] are quite dated, and their original source is worth investigating further.
Seinfeld and Pandis (2006) [30
] reference the urban and rural distributions to Jaenicke (1993) [31
], who summarized prototypical aerosol distributions for various ambient environments. Jaenicke’s work has been widely used in other indoor environmental research and applications, including in a widely cited study on the size-resolved dynamics of outdoor aerosol transport and persistence in buildings [25
]. Following the literature further, Jaenicke (1993) [31
] appears to have gathered these aerosol distribution data from publications dating as far back as the 1970s. Specifically, the rural aerosol distributions reported in Jaenicke (1993) [31
] are stated to come from measurements in the lower troposphere over the high plains of North America that were conducted in 1975 and 1976 using a combination of electrical aerosol analyzer (EAA: ~0.01–0.36 µm), optical particle counter (OPC: ~0.3–40 µm), and an axially scattering spectrometer (ASSP: ~1.4–45 µm) [32
]. The urban aerosol distributions reported in Jaenicke (1993) [31
] appear to come from two publications from the 1960s and 1970s that used the same combination of aerosol monitoring equipment [33
], although it is not immediately clear exactly what was the original source of the urban aerosol data. Regardless, the reliance on static assumptions for historical ambient aerosol distributions from the 1970s or earlier raises questions about the relevance and validity of using the distributions to estimate realistic values for e
, and e
today, which this work attempts to address.
3.2. How Relevant Are the Historical Ambient Aerosol Distributions Used in ISO 16890 Today?
One of the largest known reviews of ambient aerosol distributions globally is Azimi et al. (2014) [29
], which identified a total of 194 long-term average ambient distributions (i.e., from one year or more of measurements) primarily from North America (i.e., the US and Canada) and Europe (i.e., Sweden, Finland, Norway, Denmark, Germany, France, Italy, Ireland, England, Netherlands, Switzerland, Lithuania, Hungary, the Czech Republic, and Bulgaria). The article fit trimodal lognormal distributions to each of the 194 distributions found in the literature, and reported geometric means, geometric standard deviations, and total number concentrations for each mode fit. The article also estimated PM2.5
concentrations for each distribution, assuming spherical particles and both unit and varying size-dependent density.
Azimi et al. (2014) [29
] also used the 194 ambient distributions to estimate the fractional removal efficiencies for a number of HVAC filters for integral measures of PM2.5
and total ultrafine particles (i.e., UFPs: all particles smaller than 100 nm). The work followed a similar procedure as that described in ISO 16890 and in Section 2
herein, albeit with the following exceptions: (1) 194 long-term average trimodal ambient aerosol distributions from around the world were used, rather than a single static urban or rural bimodal distribution; (2) fractional filter removal efficiency data from laboratory tests that measured down to as small as 30 nm were used (and the remaining data were fit with a model down to just a few nanometers) [35
]; and (3) in addition to estimating PM2.5
and UFP removal efficiencies assuming 100% outdoor air delivery, PM2.5
and UFP removal efficiencies for indoor particles of outdoor origin were also estimated assuming ambient aerosols penetrated through the building envelope of a typical residential building with a recirculating HVAC system. This approach allowed for estimates of the PM2.5
and UFP removal efficiency for various filters for indoor particles of outdoor origin, which is a condition that frequently occurs in many buildings when indoor sources are not present [36
]. The number and volume distributions from the 194 outdoor particle size distributions reviewed in Azimi et al. (2014) [29
] are shown in Figure 2
, with the number and volume distributions from Jaenicke (1993) [31
Although the ambient number concentrations in Jaenicke (1993) [31
] are clearly much higher in magnitude than all 194 of the more recent ambient distributions, largely due to decreasing historical trends in ambient particulate matter concentrations over time, the shape of the number distributions still share some similarities. For example, they all have a peak number concentration around 10–20 nm. However, there are also some deviations between the two sets of distributions, particularly around 0.05–0.5 µm. Conversely, the volume distributions from Jaenicke (1993) [31
] appear to be very different from the other distributions from the recent literature. However, the main reason for the differences in the volume distributions, particularly in the >2.5 µm size range, is the use of different aerosol monitoring instruments in each study. Most recent measurements have been made using a scanning mobility particle sizer (SMPS) with an upper size limit for the number concentrations of as high as ~0.8 µm, while only a few also used an optical particle sizer (OPS, with an upper size limit of ~10 µm) or an aerodynamic particle sizer (APS, with an upper size limit of ~20 µm). The combination of instruments used to measure the distributions reported in Jaenicke (1993) [31
] (i.e., EAA and OPC) yielded substantial coverage of number concentrations above 1 µm, and thus the results could be used to accurately identify the large volume peaks in the ~5–20 µm size range. Therefore, the volume distributions for the 194 distributions reported in Azimi et al. (2014) [29
] are only accurate up to ~2.5 µm. Thus, only the estimates of PM1
concentrations (and e
) are accurate for these distributions (i.e., not the PM10
concentrations and e
). Moreover, it is also possible that the volume concentrations in the 1–2.5 µm size range estimated using the Azimi et al. (2014) [29
] data are somewhat lower that reality, because most of the distributions measured via SMPS would miss some volume/mass estimates in that size range. However, this is likely a small issue, as most of the distributions still have substantial volume/mass in the 1–2.5 µm size range.
To explore the impact of these assumptions for ambient aerosol distributions on resulting estimates of e
PM, Figure 3
shows estimates of e
(but not e
) made for four HVAC filters that were originally tested for size-resolved removal efficiency in Hecker and Hofacre (2008) (rated at MERV 6, 8, 10, and 14, as shown in Figure 4
], assuming the filtration of 100% outdoor air of (1) the ambient urban and rural volume distributions from Jaenicke (1993) [31
], and (2) the average of the 194 ambient volume distributions from Azimi et al. (2014) [29
demonstrates that the use of more modern ambient distributions yields different estimates of e
PM than the ISO urban and rural distributions. Estimates of e
had larger deviations than estimates of e
. Estimates of e
deviated by ~2–6% and ~1–2% on an absolute basis between the Azimi et al. (2014) [29
] distributions and the ISO rural and urban distributions, respectively. Conversely, estimates of e
for the MERV 8 filter ranged from as low as ~32% using the average of the ambient distributions from Azimi et al. (2014) to as high as ~54% using the rural distribution from ISO 16890; the urban distribution was in between these two estimates (~40%). Estimates for e
for the MERV 6 filter ranged from ~9% (Azimi et al.) to ~12% (ISO urban) to ~19% (ISO rural) [29
]. The primary reason for these differences is that the peaks and overall shapes of the volume distributions in Figure 2
are reasonably well aligned for all of the distributions in the <1 µm size range, but not for the distributions in the 1–2.5 µm size range.
3.3. How Does the Outdoor-to-Indoor Transport of Ambient Aerosol Distributions Affect Estimates of ePM for Particles of Outdoor Origin?
Indoor aerosols are most commonly a mixture of aerosols that transport from outdoors, and are emitted from indoor sources [41
]. This section explores the impact of the outdoor-to-indoor transport of ambient aerosols alone (i.e., in the absence of indoor sources), while the subsequent section explores the impact of indoor distributions resulting from a combination of both indoor and outdoor sources.
As ambient aerosols transport indoors, they are subject to a number of size-dependent removal mechanisms that can alter the nature of the distributions, including penetration through the building envelope (if ambient aerosols are transported indoors via infiltration) and deposition to indoor surfaces [25
]. Azimi et al. (2014) accounted for ambient particle penetration and deposition in typical residential buildings that rely mostly on infiltration for ventilation air to estimate PM2.5
and UFP removal efficiencies for indoor particles of outdoor origin (again, ignoring indoor sources) [29
]. Although infiltration had a small effect on calculated removal efficiencies in that work, on average, some cases showed larger deviations than others.
Similarly, Figure 5
shows the absolute differences in estimates of e
, and e
that were made for the same ambient aerosol distributions in Figure 2
, and the same four HVAC filters in Figure 4
, although the ambient distributions were also modified to account for typical size-dependent residential infiltration factors. Differences were calculated as the estimated removal efficiency assuming that infiltration factors were applied prior to calculating e
PM minus the removal efficiency estimated assuming the filtration of 100% outdoor air. These data are intended to provide a comparison of the two calculation approaches that are used for a given ambient distribution, which allows for comparing PM10
in addition to PM1
for all three ambient distribution sources: ISO rural, ISO urban, and the average from the 194 distributions in Azimi et al. (2014) [29
]. Assumptions for size-resolved infiltration factors based on typical size-resolved penetration factors, deposition loss rates, and air exchanges rates in US residences (from Azimi et al. 2014 [29
]) are shown in Figure 6
]. Note that the literature on envelope penetration factors in commercial buildings is quite limited, so this analysis is only applicable for a typical US residential building that relies on infiltration alone. Moreover, this analysis ignores indoor sources, and thus the results are only applicable to periods of time in which indoor aerosols result only from the infiltration and persistence of ambient aerosols alone (i.e., indoor sources are ignored).
Accounting for outdoor-to-indoor transformations had only a small influence on e
for indoor particles of outdoor origin for all three assumed ambient aerosol distributions (i.e., less than 2% for all comparisons), but the magnitude of differences increased with increasing particle size classifications. This is particularly true for the urban and rural distributions used in ISO 16890, which show absolute deviations that are as high as ~5% for e
, and as high as ~15% for e
for the MERV 8 and MERV 10 filters for indoor particles of outdoor origin. These high values are largely because a substantial portion of the ISO 16890 aerosol volume distributions is in the 1 µm to 10 µm size range. Particles greater than a few µm tend to have very low infiltration factors, due to both low penetration factors through building envelopes and high indoor deposition loss rates (Figure 6
]. This is a critical distinction to consider that, if not accounted for, can lead to rather large overestimations of e
for indoor particles of outdoor origin if one uses only the 100% outdoor air assumptions in ISO 16890. It may be more appropriate to first account for transformations that occur during outdoor-to-indoor transport, rather than relying on e
PM reported according to ISO 18690, if one is to estimate e
PM removal efficiency of a filter for particles of outdoor origin for use in a building that relies on infiltration for ventilation air (which includes the vast majority of residences in the US, for example), as several studies have done recently [45
]. It is also worth noting that the absolute differences in e
were smaller for the MERV 14 filter, because the filter has a removal efficiency of almost 100% for particles larger than 3 µm, and over 90% for 1–3 µm particles (Figure 4
). Thus, the ambient transformations that most drastically affect the >1 µm size range (Figure 6
) did not influence the e
PM results as much as lower efficiency filters.
3.4. How Does the Use of Indoor Aerosol Distributions Affect Estimates of ePM?
Since indoor aerosols most commonly result from a mixture of aerosols that are generated indoors and infiltrated from outdoors, estimates of e
, and e
were made for the same four HVAC filters, but assuming that the challenging aerosol was from a small sample of indoor aerosol distributions. Although there are no well-defined or standardized definitions of “typical” indoor aerosol distributions, the literature on indoor particle size and/or mass distributions was searched for those studies that used both an SMPS or similar instrument to measure sub-micrometer particles and an APS, OPS, or similar instrument to measure super-micrometer particles simultaneously. Two seminal studies of indoor aerosols distributions in a sample of US residences were found for use in this exploratory analysis: Abt et al. (2000) [50
] and Long et al. (2001) [51
]. Abt et al. (2000) [50
] measured indoor and outdoor particle concentrations for approximately weeklong periods over multiple seasons in four single-family homes in Boston, Massachusetts, US, using a combined SMPS and APS sampling system to measure particles between 0.02–0.5 µm and 0.07–10 µm, respectively. Long et al. (2001) [51
] similarly measured indoor and outdoor concentrations for approximately weeklong periods over multiple seasons in nine single-family homes in Boston, Massachusetts, US using the same instrumentation as Abt et al. (2000) [50
]. The average number and calculated volume concentrations throughout the entire sampling periods were reported in both studies, which captured the average of a variety of indoor aerosol sources including cooking, cleaning, resuspension from occupant activities, and infiltration from ambient air, among others.
shows the average indoor distributions from each study on a volume basis. The dashed lines represent the size-resolved data using the same size bins that were used in the original studies, and the solid lines represent manual approximations of trimodal logarithmic fits to these same distributions. Manual approximations of these distribution parameters are also shown in Table 2
. Figure 8
also shows the same approximations of indoor volume distributions overlaid on top of the ambient distributions from Figure 2
for comparison purposes.
There are some clear differences in the shape and magnitude of the ambient volume distributions from Jaenicke (1993) [31
] (and ISO 16890), and the average indoor volume distributions from Abt et al. (2000) [50
] and Long et al. (2001) [51
]. For example, the peak volume concentration is ~0.3 µm for the urban ambient distribution, and ~0.2 µm for the indoor volume distributions. Additionally, the indoor distributions are relatively flat, from ~1 to ~8 µm, while the ambient distributions vary widely in shape in the same size range. These differences are influenced by the nature of the underlying emission sources in the indoor studies, which can vary widely in the magnitude and shape of their number and volume distributions [36
shows estimates of e
, and e
for the same four HVAC filters that were used in prior sections made using (i) the ISO 16890 urban and rural ambient volume distributions (assuming filtration of 100% outdoor air) and (ii) the average indoor volume distributions from Abt et al. (2000) [50
] and Long et al. (2001) [51
]. The intent of this comparison is to demonstrate whether or not using the ambient distributions in ISO 16890 can be used to accurately estimate the e
PM for indoor aerosols. Although the number of indoor distributions is small, the comparison provides an exploratory analysis that could be further refined with additional indoor distribution data.
The results in Figure 9
suggest that estimates of e
PM made for ambient distributions following the current ISO 16890 standard cannot be used to accurately predict e
PM for all of the size classifications for use cases that involve recirculating HVAC systems with indoor sources, particularly for e
. For example, using the 100% outdoor air ISO urban and rural distributions resulted in estimates of e
for the MERV 6 filter of 25.6% and 35.4%, respectively. Meanwhile, using the Abt et al. (2000) [50
] and Long et al. (2001) [51
] indoor distributions resulted in estimates of ~24% and ~16%, respectively. Even more drastically, using the 100% outdoor air ISO urban and rural distributions resulted in estimates of e
for the MERV 10 filter of ~62% and ~77%, respectively, while using the Abt et al. (2000) [50
] and Long et al. (2001) [51
] indoor distributions resulted in estimates of ~60% and ~49%, respectively. Interestingly, the estimates of e
for all four filters were reasonably similar for the two indoor aerosol distributions and the ISO urban ambient distribution (i.e., within 2% or less for all of the comparisons), suggesting that the approach in ISO 16890 may yield reasonable estimates for typical indoor residential applications, although further investigation with greater numbers of indoor distributions is warranted.