Size-resolved fluorescent biological aerosol particle concentrations and occupant emissions in a university classroom



This study is among the first to apply laser-induced fluorescence to characterize bioaerosols at high time and size resolution in an occupied, common-use indoor environment. Using an ultraviolet aerodynamic particle sizer, we characterized total and fluorescent biological aerosol particle (FBAP) levels (1–15 μm diameter) in a classroom, sampling with 5-min resolution continuously during eighteen occupied and eight unoccupied days distributed throughout a one-year period. A material-balance model was applied to quantify per-person FBAP emission rates as a function of particle size. Day-to-day and seasonal changes in FBAP number concentration (NF) values in the classroom were small compared to the variability within a day that was attributable to variable levels of occupancy, occupant activities, and the operational state of the ventilation system. Occupancy conditions characteristic of lecture classes were associated with mean NF source strengths of 2 × 106 particles/h/person, and 9 × 104 particles per metabolic g CO2. During transitions between lectures, occupant activity was more vigorous, and estimated mean, per-person NF emissions were 0.8 × 106 particles per transition. The observed classroom peak in FBAP size at 3–4 μm is similar to the peak in fluorescent and biological aerosols reported from several studies outdoors.

Practical Implications

Coarse particles that exhibit fluorescence at characteristic wavelengths are considered to be proxies for biological particles. Recently developed instruments permit their detection and sizing in real time. In a mechanically ventilated classroom, emissions from human occupants were a strong determinant of coarse-mode fluorescent biological aerosol particle (FBAP) levels. Human FBAP emission rates were significant under quiet occupancy conditions and increased with activity level. Fluorescent particle emissions peaked at a diameter of 3–4 μm, which is the expected modal size of airborne particles with associated microbes. Human activity patterns, and associated coarse FBAP and total particle levels varied strongly on short timescales. Thus, the dynamic temporal behavior of aerosol concentrations must be considered when determining collection protocols for samples meant to be representative of average concentrations using time-integrated or ‘snapshot’ bioaerosol measurement techniques.


Bioaerosols, or particles of biological origin, are widely present in indoor environments (Després et al., 2012). They are important for the health and comfort of occupants (Ege et al., 2011; Mendell et al., 2011; Mitchell et al., 2007), and for the aesthetics and maintenance of the built environment (Borrego et al., 2010; Portnoy et al., 2004). Bioaerosol characterization may also be applied as a forensic tool for source detection and identification (Castillo et al., 2012; Fierer et al., 2010). However, techniques to measure bioaerosols that are most widely used involve substantial sample-processing steps, and are limited by minimum biomass requirements or saturation effects (Caruana, 2011; Després et al., 2012). Consequently, common methods cannot readily provide the temporal resolution needed to elucidate dynamics that occur on short timescales, especially for extended monitoring durations.

Laser-induced fluorescence of individual particles has been demonstrated to be a useful technique for measuring bioaerosols in real time. One recently introduced commercial particle counter is the ultraviolet aerodynamic particle sizer (UV-APS; model 3314, TSI, Inc., Shoreview, MN, USA). The UV-APS autofluorescence excitation and emission wavelengths of 355 and 420–575 nm, respectively, are characteristic of reduced pyridine nucleotides and flavins (Agranovski et al., 2004a; Harrison and Chance, 1970), and are, nominally, indicators of cellular metabolic activity (Agranovski et al., 2003b; Eng et al., 1989; Hairston et al., 1997; Li et al., 1991). However, the interrogation of particles at a given excitation wavelength provides detection that is not specific at the molecular level (Pan et al., 2007; Pöhlker et al., 2012). For example, dormant and nonliving biological materials, such as bacterial spores (Brosseau et al., 2000), pollen, and cell-wall compounds and pigments (Bliznakova et al., 2007; Pöhlker et al., 2012, 2013), have also been observed to fluoresce at the target wavelengths. Potential abiotic interferents, such as soot, certain types of secondary organic aerosol, humic and fulvic acids, and minerals may also contribute to the fluorescent signal (Agranovski et al., 2003a; Bones et al., 2010; Lee et al., 2013; Pöhlker et al., 2012; Saari et al., 2013).

Previous studies utilizing the UV-APS were conducted in a laboratory (Agranovski et al., 2003a; Jung et al., 2012; Kaliszewski et al., 2013; Kanaani et al., 2008), in outdoor air (Huffman et al., 2010, 2012; Schumacher et al., 2013), and in a swine confinement building (Agranovski et al., 2004b). Results from these studies suggest that while biological materials other than live microbes contribute to the fluorescent signal in many settings, interference from abiotic species is likely to be weak in the supermicron particle size fraction (Healy et al., 2012; Huffman et al., 2012; Stanley et al., 2011; Toprak and Schnaiter, 2013). Hence, supermicron particle fluorescence at the UV-APS wavelengths has been interpreted as a lower-limit estimate for the abundance of primary biological aerosol particles, or PBAP (Huffman et al., 2010; Pöhlker et al., 2012), and coarse-mode fluorescent particles have been referred to as fluorescent biological aerosol particles, or FBAP. To date, little research has been reported utilizing the measurement of fluorescent particles in common occupied, indoor environments.

The present study was undertaken to investigate sources and dynamics of coarse-mode (>1 μm) fluorescent biological and total aerosol particles in indoor air in a common setting. A university classroom was chosen as the test site both for convenience and to focus on the influence of occupants, as humans are known to play a prominent role as a source of coarse PBAP in densely occupied indoor spaces (Fox et al., 2005; McKernan et al., 2008; Qian et al., 2012; Scheff et al., 2000). Previous research using quantitative bioaerosol measures to evaluate the influence of occupants on indoor aerosols under normal, occupied conditions has been limited to either integrated sampling periods of several hours or days for molecular, chemical, and direct-counting methods (Chen and Hildemann, 2009; Fox et al., 2005; Qian et al., 2012; Toivola et al., 2002), or short (2–20 min), sequential snapshots for culture-based methods (Bartlett et al., 2004; Brandl et al., 2008; Lehtonen et al., 1993; Luoma and Batterman, 2001; McKernan et al., 2008; Scheepers et al., 2012; Scheff et al., 2000). Size-resolved emission rates of bacteria and fungi attributable to human occupancy of an indoor environment have only been reported in one previous investigation (Qian et al., 2012), which presented average estimates based on a single, 4-day period in a university classroom. The methods employed in that study did not provide any time resolution for exploring bioaerosol dynamic processes in the studied environment.

The objective of the present study was to utilize continuous measurements, collected at high time resolution, to evaluate the influence of the common states of sitting (and writing) and lightly walking on indoor bioaerosol concentrations. The approach incorporates the influence of emissions from the respiratory tract, the shedding of particulate debris from human skin, hair, and clothing, plus suspension of particles from the floor and other contact surfaces. To assess seasonal changes, weeklong intensive study periods were repeated four times throughout a year. The classroom was also monitored over many days when unoccupied. Indoor and outdoor levels of carbon dioxide, temperature, and relative humidity, and size-resolved (total) particle levels were simultaneously sampled. Occupancy was assessed through direct observation supplemented by sensor measurements. A material-balance model was applied to quantify FBAP number emission rates as a function of particle size, using a steady-state approach during class sessions when levels were stable, and an integral analysis approach for rapidly changing conditions between classes. To our knowledge, this is the first study reporting FBAP measurements in a common occupied indoor environment.


Study site

A classroom on the fifth floor of a university building in an urban area in the western United States was selected as the study site. The room was chosen based on ease of access, a consistent pattern of use that included occupied and unoccupied periods, and supplemental criteria including hard tile flooring, no green plants, no visible water damage, and a single-pass ventilation system. Smoking was not permitted in the building. Activity in the room was typical of a college classroom: students were seated and took notes during lectures, and a lecturer presented information using a chalkboard, overhead projector, and/or computer projector. The duration of each of several daily lecture periods was typically 50 or 80 min, and occupancy during classes ranged from 20 to 80 students. During 10-min intervals between classes, there was considerable motion as students entered and exited. The room's two doors, which opened to an interior hallway, were usually closed during class periods. Custodial cleaning was conducted on an approximately daily basis in the early morning. Cleaning involved wiping the chalkboard and desks, sweeping, and, less frequently, wet mopping of the floor.

The room volume was assessed as 670 m3 based on physical dimensions (length, width, height). Nine percent of the measured volume was subtracted to account for large objects including furniture, HVAC ducts, and people, to thus convert the measured volume into an approximate ‘ventilated’ volume estimate of 610 m3 (Hodgson et al., 2005). The mechanical ventilation system operated from 8:00 to 20:45 during weekdays. Supply air was filtered with standard, pleated, MERV 11 filters that were replaced on a regular schedule. Natural ventilation made an insignificant contribution to total air exchange owing to the absence of exterior doors and windows. Furthermore, the floor, ceiling, and four walls of the room each divided it from other interior spaces in the building. Consequently, infiltration of outdoor air was expected to contribute little to the ventilation rate when the mechanical system was operating.

The air-exchange rate (AER) was determined to be 5 ± 0.5 per hour (mean ± standard deviation) when the mechanical ventilation system was on, and 0.4 ± 0.05 per hour when it was off. The AER was evaluated through CO2 tracer decay tests (utilizing deliberate releases) and from analyzing the change in the CO2 level in the classroom during the transition from occupied to unoccupied conditions. The state (on or off) of the mechanical ventilation system was continuously monitored via an anemometer (Model 9545; TSI, Inc.) that was mounted beneath an air supply vent in the room.

Experimental protocols and instrument details

Observational monitoring under occupied conditions was conducted during four approximately weeklong periods in March, April, September, and November 2012 (total weekday sampling duration = 475 h). The classroom was also monitored during unoccupied periods in May, July, and November (total = 201 h), and outdoor air was sampled before and after indoor sampling periods for ~24 h in each test (total = 261 h). Sampling dates and times are summarized in Table S1 in the online supporting information.

During observational monitoring, a UV-APS measured time-resolved number concentrations of FBAPs and all airborne particles with an aerodynamic diameter between 1–15 μm (resolved into 36 size groups). The device was configured to sample air during 285 s out of every 5 min. Autofluorescence intensity of particles was recorded on a scale from 0 to ≥63 (arbitrary units). The total flow rate of 5 l/min included a sheath flow rate of 4 l/min and an aerosol sampling rate of 1 l/min. The UV-APS data were analyzed via user-written software based on the Igor platform (Wavemetrics, Inc.; Portland, OR, USA; Huffman et al., 2010). The instrument was operated from within an air-cooled, foam-lined enclosure placed inside the classroom that was designed to muffle instrument noise and provide for instrument security. The inlet extended from the enclosure vertically into the room via electrically conductive silicon rubber tubing (length 0.3 m, inner diameter 17 mm).

Additional time-resolved (once every min) data were gathered through supplementary particle and co-pollutant sensors deployed indoors and outdoors during the indoor sampling periods. These included optical particle counters (model GT-526; Met One, Inc., Grants Pass, OR, USA), which measured levels of particles larger than 0.3 μm in diameter (resolved into six size groups); carbon dioxide monitors (model 820, LI-COR Biosciences, Lincoln, NE, USA; model 8554 Q-Trak Plus, TSI, Inc.); and temperature and relative humidity data loggers (HOBO U10-003; Onset Computer Corp., Bourne, MA, USA). Data collected by supplementary sensors were used to investigate determinants of indoor FBAP levels.

Indoor sampling was conducted from a location at the rear of the classroom, at an elevation of 1.2 m, corresponding approximately to the breathing height of a seated person. Outdoor sampling was conducted via inlets protruding from a window in the same building (length 1 m, bend radius ~0.5 m) under a shade that provided protection from sunlight and rain. A supplementary particle monitor was placed in a second location in the classroom, within the block of student desks and in closer proximity to occupants, to acquire information on the spatial variability of particle levels in the room.

Direct observation by the field research team yielded information about building features and dimensions, occupant activity patterns in the classroom (at 1-min resolution approximately between the hours of 9:00 and 17:00), and potential particle emission sources. Additional information on occupancy patterns was obtained through the use of ‘state’ sensors (HOBO U9; Onset Computer Corp.), which indicated when doors were opened or closed, and a light/movement sensor (HOBO UX90-006; Onset Computer Corp.) positioned in the room.

Quality assurance

Instrument calibrations and performance checks were conducted throughout the period of field monitoring. The UV-APS size and fluorescence response were tested with monodispersed polystyrene latex (PSL) particles with physical diameters in the range 0.6–5 μm (Duke Scientific Corp., Fremont, CA, USA; Polysciences, Inc., Warrington, PA, USA). Data were post-processed to correct for a consistent additive offset (0.2 μm) observed in the size response, based on the calibration curve presented in Figure S1a and Table S2 of the online supporting information. The accuracy of the adjusted UV-APS output was assessed by comparing it with the response from a second aerodynamic particle sizer (model 3312, TSI, Inc.; Figure S1b). The UV-APS flow rate, laser, and high-concentration status flags were within standard limits in all cases. The fluorescence response was within manufacturer specifications. Additional details are reported in the online supporting information Data S1.

Side-by-side (SBS) testing of the optical particle counters (OPCs) was conducted on multiple occasions, and at least once before and after each of the indoor observational monitoring periods, with the instruments co-located for a few hours in a laboratory or in the monitored classroom. The averaging time and metric chosen for matching instrument responses were 1 min, and number concentration of particles larger than 1 μm, respectively. One of the OPCs (labeled OPC3) was designated as the reference instrument, and the objective was to adjust data from other OPCs to match as closely as possible the response of the reference device, so as to reduce the bias in ratios of concentrations measured by various instruments during the study. Table S3 summarizes slopes, or ‘adjustment factors’ from SBS tests. Average values from this table are used to adjust OPC1, OPC2, and OPC4 data.

Flow rates of all active instruments were checked with an external flow meter (Defender 510, BIOS International Corp., Butler, NJ, USA) and were corrected when they were outside of manufacturer-specified normal ranges. The particle instruments were confirmed to measure zero particle concentrations when air was sampled through a particle filter placed at the inlet. The response of carbon dioxide monitors was tested using calibration gases at 0 and 1000 ppm. When the response was outside the manufacturer-specified normal range, data were post-processed to correct the observed bias. Carbon dioxide data collected outdoors with the Q-Trak during periods of high relative humidity were omitted from consideration, owing to a strong and apparently spurious association between the two parameters.

UV-APS data analysis

Particle counts recorded by the UV-APS were processed by evaluating number concentrations of particles in the 1–15 μm size range (referred to in this paper as coarse particles) with fluorescence intensities ≥0 (total particles, NT), ≥2 (fluorescent biological aerosol particles, NF), and ≥20 (highly fluorescent particles, NF20). Huffman et al. (2010) provide a rationale for using a fluorescence intensity cutoff of 2 for quantifying NF. The threshold of 20 (arbitrary units) for highly fluorescent particles was chosen based on a visual inspection of the fluorescence intensity distribution of particles sampled in the occupied classroom. A characteristic feature of this distribution was the presence of a large fraction of particles that saturated the fluorescence detector, causing an apparent bimodal fluorescence intensity distribution as shown in Figure S2. As a consequence, the estimate of the fraction of particles that was ‘highly fluorescent’ was robust, meaning that it did not depend strongly (i.e., the outcome was approximately constant within a factor of 2) on the choice of cutoff intensity between 20 and 63 (arbitrary units). For the UV-APS sampling duration and flow rate employed in this study, the minimum detection limit (MDL) was 0.00021 particles/cm3. Below approximately 10−3/cm3, concentration changes lacked smoothness as they changed in discrete multiples of the MDL for each additional particle detected.

Data interpretation

Occupant emission rates of size-resolved total particles, FBAP, and highly fluorescent particles were estimated for the combined processes of shedding and resuspension, and for conditions representative of high vs. low classroom activity, using a material-balance approach (Equation (1)) applied to data from the occupied sampling periods. The analysis assumed well-mixed conditions and steady values of dynamic particle parameters. The analysis was conducted for eight lumped particle size groups i (distilled from 36 UV-APS size bins between 1 and 15 μm; Table S4), as a way to ensure that a large proportion of measured concentrations were sufficiently above the detection limit to be robust.

display math(1)

In Equation (1), Nnet,i(t) is the increment of indoor number concentration (particles per cm3) at time t associated with occupants; Ei(t) is the occupant emission rate by processes of shedding and resuspension at time t (particles/h); V is the volume of the indoor mixed zone (cm3); a is the air-exchange rate (1/h); and ki is the first-order particle deposition loss-rate coefficient (1/h).

The net indoor particle number concentration Nnet,i(t) was evaluated as the indoor concentration at time t minus the expected contribution from outdoor particles. For highly fluorescent particles, the contribution from outdoors was evidently low, based on measurements made during the unoccupied sampling periods, and was assumed to be zero. For FBAPs, the contribution from outdoors was represented with an empirically determined ‘indoor concentration of outdoor particles,’ evaluated separately for each weeklong monitoring period and each particle size group (Table S5). The approach was based on the observation that indoor FBAP levels were substantially and consistently greater during occupied periods than for unoccupied conditions. The indoor concentration of outdoor particles was estimated as the product of a constant factor, the indoor proportion of outdoor particles (IPOPi), multiplied by the daily (9:00–17:00) mean outdoor concentration measured before and after the weeklong period being considered.

Particle size-specific estimates of the IPOPi, as presented in Figure S3, were evaluated as the mean indoor/outdoor ratio (based on 1-min data) of total particles measured with optical particle counters during the unoccupied period Unocc1. These estimates were assessed for reasonableness by comparing measured results with modeled values computed using Equation (2), which is adapted from Riley et al. (2002), neglecting natural ventilation, infiltration, recirculation, and based on = 5/h, size-specific deposition coefficients (ki) as reported in Figure S4, and size-specific filtration estimates (ηi) for MERV 11 filters.

display math(2)

Modeled and measured (in parentheses) IPOPi estimates were 0.19–0.34 (0.46), 0.13–0.3 (0.20), and 0.09 (0.10) for 1–2 μm, 2–5 μm, and >5 μm particles, respectively. A best-fit curve (based on a power-law relationship and measured values) was used to estimate IPOPi values for the eight UV-APS lumped-size groups. The largest OPC size group was assumed to have an upper limit of 15 μm for the purposes of this calculation.

Quantitative estimates of size-specific deposition loss coefficients (Figure S4) were inferred from FBAP trends observed during cleaning episodes that occurred regularly during observational monitoring periods. Cleaning resulted in pronounced increases in NF, followed by a period when the room was unoccupied with the ventilation system off. As such, cleaning episodes represented convenient, repeated instances from which to estimate NF decay rates. The deposition loss coefficient was evaluated as the negative slope of the natural log of NF,net,i(t) vs. time during the immediate post-emission period, minus the air-exchange rate. The estimates were expected to be lower bounds of ki during classroom-occupied conditions owing to the increased surface area associated with the presence of people, increased airflow caused by their movement, plus the higher airflow rate induced by the operation of the ventilation system. Notwithstanding these limitations, the estimates are within the range of previously reported values (Thatcher et al., 2002; You et al., 2013).

Emissions associated with two major categories of activities – lecturing and transitions between classes – were assessed using two different approaches. During lectures, particle levels in the classroom typically began high owing to the high activity level during the preceding transition and then settled to quasi-steady concentrations as the class session proceeded. A steady-state approximation was invoked to evaluate the emission rate Eij (in units of particles/h) using Equation (3), where Nnet,avg,i,j is the average Nnet,i(t) during the last 15 min of lecture class j:

display math(3)

This approach was applied for determining emissions of both FBAPs and highly fluorescent particles. However, total particle emissions were not evaluated using this approach owing to the difficulty of obtaining robust Nnet,i(t) estimates for total particles. The incorporation of a constant influence from outdoors was not valid for total particles due to the more pronounced influence of outdoor levels on indoor concentrations; and a modified approach based on incorporating time-varying outdoor total particle trends could not be applied because we lacked simultaneous outdoor APS measurements. The carbon dioxide emission rate per lecture session was also evaluated, on the basis of CO2net calculated as CO2in-CO2out, and based on treating CO2 as chemically inert (kCO2 = 0). The outdoor CO2 data acquired during Occ1 and Occ2 observational sampling periods (with a Q-Trak monitor) lacked the accuracy needed for emission rate assessments. For these two measurement periods, the mean CO2out from Occ3 and Occ4 emission rate evaluations (407 ppm) was substituted in place of measured CO2out levels.

In contrast to the quiet occupancy conditions during lecture classes, during transitions, increased occupant movements resulted in pronounced total aerosol and FBAP peaks whose onset coincided with the end of a lecture class. Emissions (quantified in units of particles per transition event) were estimated for total particles and FBAPs, using a peak integration approach: the area under the peak (above the baseline) of the corresponding particle time series, multiplied by × (ki a). For total particles, the method yielded robust results only for particle sizes larger than ~2.5 μm. For smaller particles, a low peak-to-baseline ratio caused results to be sensitive to small changes in peak integration parameters such as the choice of assigned peak duration and baseline. For consistency, only two-way transitions (corresponding to the ~10-min period between the end of one class period and the start of another when the two were scheduled in succession) were analyzed and reported.

Lecture classes (= 49) and transitions (= 24) observed during the full observational monitoring period were considered as two single groups, because no significant differences in mean daily concentrations based on time-of-year were observed. The few class sessions when the classroom's auxiliary ventilation system was turned on (= 6) were not included in these results. Also excluded from the results were non-lecture classes held in the room during the observational monitoring period, most notably discussion sections, laboratories, and examinations, as they were associated with a broader range of activity types. Particle number emissions are also reported in mass units, based on the assumptions that particles were spherical, had a density of 1 g/cm3, and that within each lumped particle size group, the mass-weighted size distribution, dM/d(logda), was constant when plotted versus log da.

Results and discussion

Fluorescent biological aerosol particle concentrations

Figure 1a presents illustrative time series plots of the daily profile for size-resolved FBAP number concentrations from the portion of the day when the ventilation system was on and classes were in session. Analogous profiles from an unoccupied day are presented in Figure 1b for comparison. Sharp, intermittent peaks in NF, NF20, and NF/NT were associated with transitions between lectures (indicated with dotted lines) when human movements in the room were most vigorous. In contrast, CO2 levels were generally increased during occupancy but did not increase consistently during each transition. Within-day changes in indoor/outdoor total coarse particle ratios (I/O) depended strongly on outdoor NT dynamics; this parameter was not, by itself, a reliable indicator of occupancy patterns. Strong activity-specific influences from cleaning and the intermittent operation of the ventilation system on indoor particle and co-pollutant levels were evident when the period considered was expanded to include a few hours before and after the ventilation system was in operation, as is illustrated in Figures S5a and S5b.

Figure 1.

Illustrative time series images (5-min resolution) of indoor size-resolved coarse FBAP number concentrations (NF) on weekdays with the ventilation system on: (a) when classes were in session vs. (b) when the room was unoccupied. Vertical dotted lines denote a transition between class periods. The traces depict indoor CO2 levels; and size-integrated (1–15 μm) NF, highly fluorescent particle number concentrations (NF20), ratios of FBAP to total particle number concentrations (NF/NT), and indoor/outdoor total particle number ratios (I/O). Means for the periods portrayed are reported in parentheses. FBAP, Fluorescent biological aerosol particle

Figure 2 presents daily, 8-h (9:00–17:00), average number concentrations of size-integrated (1–15 μm) FBAP and highly fluorescent particles. The figure is visually indicative of a pattern that was consistent across all times of year: indoor levels exceeded outdoor levels during occupancy, and outdoor levels exceeded indoor levels during unoccupied times. The daily average NF during occupied days (GM = 0.039/cm3, GSD = 1.3, = 18) was eight times as high (< 0.0001; Mann–Whitney U-test) as the average during unoccupied conditions (GM = 0.005/cm3, GSD = 1.6, = 8). For NF20, the occupied/unoccupied ratio was even greater (>30), as NF20 levels in the classroom were typically near the detection limit when occupants were absent. The finding of higher daily 8-h mean NF levels outdoors (GM = 0.013/cm3, GSD = 1.4, = 9), compared to indoors in the absence of occupants, was consistent with the pattern for NT (Figure S6). It almost certainly reflects the influence of the building ventilation system in filtering particles of outdoor origin, and the apparent lack of significant indoor coarse particle sources in this environment other than occupants. No seasonal differences in mean indoor FBAP levels were observed, and changes in NF and NF20 among the days sampled were small relative to differences between occupied vs. unoccupied states and between indoor vs. outdoor samples. In contrast to the trend for FBAPs, total particle levels (Figure S6; Table S6) displayed greater variability across sampled days, with an inverse pattern of outdoor levels exceeding the indoor levels, even during occupancy.

Figure 2.

Daily, 8-h (9:00–17:00) means, showing the change in (a) NF and (b) NF20 across days, and between indoor-occupied (18 days), indoor-unoccupied (8 days), and outdoor (9 days) conditions. Dotted horizontal lines at arbitrary concentrations are added to emphasize the pattern of occupied > outdoor > unoccupied across all months sampled

Size distributions of daily (9:00–17:00) mean NF, NF20, and NT corresponding to occupied and unoccupied conditions are compared in Figure 3. The fluorescent particle data exhibited a single prominent mode of diameter around 2–4 μm. There is evidence of a second weak mode in the 1–2-μm-diameter range. Lognormal distributions provided good fits to the median number concentrations of particles larger than 2 μm. Figure 3 shows the fit and lists the mode, geometric standard deviation (σg), and amplitude (A), that is, the modeled modal y-axis value, for each distribution. Based on modeled distributions, occupied conditions were associated with a statistically significant increase in the modal diameter of FBAP from 2.3 ± 0.1 μm (mean ± s.d.) to 2.9 ± 0.05 μm. The shift in size is consistent with previous research showing that airborne larger particle fractions are more strongly associated with occupants (Chen and Hildemann, 2009). Highly fluorescent particles were observed to be larger than FBAP, with an occupied period mode of 3.6 ± 0.06 μm. FBAP concentration peaks in the 2–4 μm range have been consistently reported, based on measurements with the UV-APS and the Wideband Integrated Bioaerosol Sampler (WIBS), in outdoor urban and forest environments (Gabey et al., 2011; Huffman et al., 2010, 2013; Schumacher et al., 2013; Toprak and Schnaiter, 2013), inside and outside a swine confinement building (Agranovski et al., 2004b), and in a laboratory investigation of fungal spores (Kanaani et al., 2008). Huffman et al. (2010, 2012) attributed the size mode of diameter ~ 3 μm for outdoor FBAP to fungal spores or agglomerated bacteria. An exception to this pattern is the reported modal diameter of 1.2 μm for fluorescent particles in urban Manchester, UK (Gabey et al., 2011), suggested by those authors to be influenced heavily by anthropogenic sources.

Figure 3.

Size distributions of daily (9:00–17:00) mean number concentrations of NF (upper panels), NF20 (middle panels), and NT (lower panels) measured during (left) occupied conditions and (right) unoccupied conditions. The mean, median, interquartile range (IQR), and 5th to 95th percentile range across sampled days are shown. The mode, geometric standard deviation (σg), and amplitude (A) of lognormal distributions that provide the best fit to concentrations of particles larger than 2 μm are shown for fluorescent particle distributions. Note that the y-axis scales are not consistent in each panel

The median 8-h daily average indoor-occupied NF/NT ratio observed in this study, 5%, is similar to the campaign mean outdoor NF/NT of 4% reported by Huffman et al. (2010). However, a different picture emerges when the dependence of NF/NT on particle size is evaluated (Figure S7). For indoor-occupied samples, the mean NF/NT rose steeply to 70% for 8 μm particles, whereas it increased less strongly, to <30%, for the indoor-unoccupied air sampled in the present study, and for the outdoor air evaluated by Huffman et al. (2010). A similarly low NF/NT ratio of 30% was observed for indoor-occupied samples if highly fluorescent particles were excluded from the analysis. In contrast, Huffman et al. (2012) observed similarly high NF/NT ratios in the Amazon, where median values were reported to peak at 80–90% for particle diameters of ~ 4 μm. The fraction of particles that was ‘highly fluorescent’ was not reported in that study. Our observations of a consistent, strong signal from particles whose fluorescence saturated the instrument sensor have not been previously observed or reported from studies using the UV-APS for outdoor observational monitoring, suggesting the highly fluorescent particles may have distinctive indoor sources.

Human emissions

Figure 4 presents size-integrated (1–15 μm) FBAP and highly fluorescent particle number emission rates during lecture class periods, as a function of corresponding carbon dioxide mass emission rates. The number of students in each lecture class is depicted with a color or gray scale. The expected pattern of increasing CO2 (r2 > 0.6) and increasing fluorescent particle emission rates (r2 > 0.3) with an increase in class size was observed. Geometric mean (GSD) per-occupant emission rates were 1.6 × 106/person/h (1.4) for NF, 0.73 × 106/person/h (1.4) for NF20, and 18 g/person/h (1.2) for CO2. For comparison, emissions during examination classes are presented in the supporting information Data S2. Geometric mean (GSD) CO2-normalized emission rates, which avoid uncertainties associated with assumptions that the room is well mixed and that the air-exchange rate is invariant with time, were 89 × 103 (1.4) particles per metabolic g CO2 for NF, and 40 × 103 (1.4) particles per metabolic g CO2 for NF20. Size distributions of CO2-normalized emissions, presented in Figure 5, exhibited a bimodal character. Lognormal distributions fitted to median concentrations above 2 μm showed occupant-emitted NF and NF20 during lectures had size modes of 3.1 and 3.7 μm, respectively.

Figure 4.

Estimated particle number emission rates (E) from lecture classes (= 49) of (a) FBAP and (b) highly fluorescent particles, shown as a function of corresponding carbon dioxide emission rates, with the class size indicated by means of a color or gray scale. The expected pattern of higher particle and CO2 emissions during larger class sizes was observed. A best-fit line (single-parameter linear regression, intercept = 0) for each group of estimates is shown. FBAP, Fluorescent biological aerosol particle

Figure 5.

Size distributions of CO2-normalized emission rates of FBAP (left panels) and highly fluorescent particles (right panels) by count (upper panels) and mass (lower panels). These emission rates are associated with occupancy as determined from the final 15 min of each lecture period. The mode, geometric standard deviation (σg), and amplitude (A) of lognormal distributions that provide the best fit to concentrations of particles larger than 2 μm are shown for count distributions. FBAP, Fluorescent biological aerosol particle

Figure 6 shows that the estimated number of FBAPs (1–15 μm) and total particles (larger than 2.6 μm, NT2.6+) emitted during transition events was a strong function of the number of people who walked in or out during each transition. Single-parameter linear regression best-fit slopes ± standard deviations (and the interquartile range of corresponding mass emission rates), which indicate the mean particle emissions per person per transition, were 0.78 ± 0.04 × 106/person for NF (44–64 μg/person), 0.41 ± 0.02 × 106/person for NF20 (29–45 μg/person), and 0.68 ± 0.04 × 106/person for NT2.6+ (49–80 μg/person).

Figure 6.

The total number of (a) NF and NF20, and (b) NT larger than 2.6 μm, emitted or suspended during transition events, plotted against the number of people who entered or exited the room during the transition. Transitions are defined as the ~10-min period between two lecture classes. A best-fit line (single-parameter linear regression, intercept = 0) for each group of estimates is shown and indicates the mean number (×106) of particles emitted per person during a transition

Figure 7 presents size distributions of estimated per-person particle number and mass emissions during transitions. One transition was excluded from consideration owing to its unique signature (Figure S8). Lognormal distributions fitted to median concentrations above 2 μm (2.6 μm for total particles) showed occupant-emitted NT, NF, and NF20 during transitions had size modes of 3.3, 3.6, and 4.2 μm, respectively. A comparison between transitions and lectures suggested that, on average, a person sitting in a room for about 30 min emitted a similar number of FBAPs as a person walking into or out of the room. The latter process occurred within the span of a few minutes, including the time it took for a person to settle down. Both sets of emission processes exhibited only small differences in mean values of the proportion of FBAPs that were highly fluorescent, and in the size distribution of emitted particles. Transitions were associated with less well-mixed conditions observed within the classroom with the optical particle counters, but only slightly so. The median ratio of total coarse particles measured near the door vs. in the core of the room was 1.1 (30 min average) for transitions, and 1.0 (15 min average) for lectures.

Figure 7.

Size distributions of per-person emissions, during transition between lectures (= 24), of FBAP (left), highly fluorescent particles (middle), and total particles (right) evaluated on a number (top panels) and mass (lower panels) basis. FBAP, Fluorescent biological aerosol particle

Qualitatively, the trends observed for FBAPs are consistent with patterns in levels of biological species sampled in schools using molecular and culturing techniques and chemical markers (Bartlett et al., 2004; Brandl et al., 2008; Fox et al., 2003, 2005; Meadow et al., 2014; Qian et al., 2012). The patterns include the following features: (i) a tendency for occupied levels to exceed outdoor levels which in turn exceed indoor-unoccupied levels; and (ii) positive associations with CO2, activity strength, and between occupancy and the fraction of the mass and number of particles that are of biological origin.

Fluorescent biological aerosol particle emission rates have not been previously reported, but our estimates are consistent with prior research on coarse particle emissions indoors associated with human occupancy. We inferred the total number of coarse particles emitted per person per particle size during lecture classes as the NF emission rate divided by the median-occupied ratio of NF/NT. This analysis indicated that the 75th percentile of lecture class emission rates were 0.9×, 1×, and 1.8× of the values reported by You et al. (2013) for the particle size groups 1–2 μm, 2–5 μm, and >5 μm, respectively, for the case of one person sitting still in a polyester suit in a controlled environmental chamber. The ‘strong activity’ estimates for the number of particles larger than 2 μm emitted per person per minute in the recent study of You et al. (2013) are 0.7× our estimate of the median number of particles larger than 2.6 μm emitted per person per transition.

Are occupant-associated FBAPs bioaerosols?

Coarse particles emitted by humans are expected to include human cells, microbial cells, clothing and personal care products, building material fragments, and particles of outdoor origin. Potential biological fluorescent components are microbial cells, human cells, pollen fragments, and other plant-derived fragments. Both human skin and skin-associated microbes have been observed, in the presence of occupants, to be increased in indoor air, and also in the dust constituting a reservoir for resuspension (Fox et al., 2008; Hospodsky et al., 2012; Rintala et al., 2008; Täubel et al., 2009; Weschler et al., 2011). While the limited available data (Clark and Shirley, 1973; Qian et al., 2012) suggest emissions of skin and microbes may be sufficient to account for the observed FBAP emissions, significant uncertainties preclude a definite identification.

Potential non-biological fluorescent species that may be found in classroom air are fluorescent whitening agents added to clothing and detergents (Burg et al., 1977; Leaver and Milligan, 1984), and organic materials emitted from building materials, or tracked in by occupants, such as soil-based humic acids. Teaching-associated chalk dust (Majumdar and William, 2009) is expected to have a weak contribution based on our own results from an analysis of aerosolized chalk. Results showed that <6% of chalk dust particles smaller than 6 μm fluoresced. The fraction increased to close to 50% of 10 μm particles (although the fluorescence intensity of the particles remained low), indicating the potential for weakly fluorescing minerals to interfere with the low-fluorescence response associated with large particles.


The number concentration of coarse FBAP in a classroom was increased by an order of magnitude above background during occupancy. Lecture classes and movements between classes were associated with mean NF source strengths of 1.6 × 106/person/h and 0.78 × 106/person per transition, respectively. Movements between classes were associated with a small increase in the modal diameter of emitted particles. In general, size distributions of estimated emissions for FBAPs larger than 2 μm conformed well to lognormal distributions with modal diameters of 3–4 μm. A mode of similarly sized particles, observed in previous work conducted in outdoor environments, has been linked to fungal spores or agglomerated bacteria. A second weak mode in the 1–2 μm particle size range was also observed and may be evidence of single bacterial cells and non-biological species. Day-to-day and seasonal changes were small compared to the variability attributable to the presence and activities of occupants, and to the operational state of the mechanical ventilation system. The results contribute to a growing body of evidence that human-associated emissions are likely to be important contributors to coarse bioaerosols in densely occupied indoor environments.

Although occupants were identified as the dominant proximate source of coarse FBAP indoors, additional research is needed to evaluate the ultimate source and composition of the particles, and to evaluate whether the fluorescent signal can be decomposed according to biological origin, including microbes and human skin, and to non-biological sources, such as chalk, soil, and clothing fibers. Future work could also elucidate the relative influences of shedding and resuspension under a range of conditions, so as to evaluate the generalizability of the present findings beyond the university classroom investigated.


This work was funded by a grant from the Alfred P. Sloan Foundation in support of the Berkeley Indoor Microbial Ecology Research Consortium (BIMERC). Elizabeth Heredia contributed to the data collection effort. We gratefully acknowledge the valuable input received from Denina Hospodsky, and from Allen Goldstein, Rachel Adams and other members of BIMERC. Thanks are also expressed for the cooperation of professors, students, and custodial and facilities staff who facilitated monitoring at the site. J. A. Huffman acknowledges internal faculty support from the University of Denver.