The NZERTF is 3-bedroom, 3-bathroom home with a basement (380 m³), main house (1^st and 2^nd floors; 624 m³)⁵⁹ and attic (99 m³), and S/V ratio of 2.7 (m⁻¹) associated with the superficial surface areas (1681 m²) of painted walls, painted ceilings, and wooden floors in the main house. The building was tightly constructed with low-VOC emitting building materials, particularly those with no formaldehyde additives (NAF) and with low acetic acid emissions.^59,60 The NZERTF was constructed of NAF paints/coatings, adhesives/sealants, insulation binders, and plywood, and with materials tested using VOC emissions testing standards set by the California Department of Public Health and/or the California Air Resources Board (CARB).⁶¹ Furthermore, gypsum board was mechanically fastened without use of adhesives and only adhesives/sealants with limited quantities of acetic acid were used when needed.⁶¹ The building ventilation system was operated at an outdoor ACR of 0.24 h⁻¹ under normal conditions to meet American Society of Heating, Refrigerating, and Air-Conditioning Engineers (ASHRAE) Standard 62.2. A variable heat recovery ventilator (HRV) exchanged indoor air with outdoor air. A heating and air conditioning (HAC) system constantly circulated air among rooms of the main house and basement with an internal circulation rate of about 1.3 h⁻¹. Filters rated at a minimum efficiency reporting values (MERV) of 7 and 8 were used in the HRV and HAC systems, respectively. Air circulated between the attic and the house via passive vents. The house temperature and RH were generally 24 °C and 30 % under normal unperturbed conditions.

The CASA experiment was designed to assess how air concentrations and surfaces responded to perturbations in conditions. There were 8 types of perturbation experiments conducted: cooking, cleaning, or injections, i.e., pesticide products, a mixture of volatile organic compounds (VOCs) called “CASA cocktail,” wood smoke, ammonia (NH₃), carbon dioxide (CO₂), or ozone (O₃). Detailed description of perturbations performed inside the NZERTF are summarized in Supplemental Information (SI) Section S1. Specific days were dedicated to each experiment type where activities or injections were made repeatedly 2 to 4 times under normal outdoor ventilation (windows and doors closed). Some experiments were repeated under both low and high RH conditions on different days. Wood smoke injections occurred late in the campaign to avoid changing the house background before other experiments. Prior to, and in between, perturbation experiments, the house was operated in a standard mode to allow the building conditions to stabilize and allow researchers to measure the house background, which reflects emissions from the building itself.

Indoor air was sampled for WSOC inside the dining room on the 1^st floor of the NZERTF (Section S2, Figure S1) using a MC, PILS, and TOC analyzer (Figure 1) located on the enclosed porch connected to the test house (Figure S1-S2). Sample lines were extended into the dining room of the house through drilled holes in the blocked door shared between the porch and dining room in the locations noted on Figure S1. The sampling inlets (Figure S3) were positioned 2 m above the floor and extended 2 m into the dining room. A 7.6 m long 0.9525 cm inner diameter (ID; 3/8 inch ID) polytetrafluoroethylene (PTFE) tube was used to collect 25 L min⁻¹ of indoor air into the MC. A 12 m long 0.9525 cm outer diameter (OD; 3/8 inch OD) refrigeration grade copper tube was used to collect 13 L min⁻¹ of indoor air into the PILS from the same location. Indoor air was drawn continuously by the MC and PILS throughout each experiment. Occasionally equivalent sample lines were extended outdoors to enable collection of residential outdoor WSOC concentrations.

WSOC_g was continuously scrubbed into a 1.2 mL min⁻¹ flow of water by pulling 25 L min⁻¹ of filtered indoor air through a 190 mL modified mist chamber, also known as Cofer scrubber,^{57, 58} equipped with a 10 mL water reservoir (Figure 1). Water exiting the MC was analyzed semi-continuously for TOC, alternating at 6 min intervals with TOC analysis of the PILS sample flow. For some experiments, only the MC was sampled, and near-real-time results are reported here. MCs use a continuous, fine aqueous mist providing large air-water surface area to efficiently scrub water soluble gases from the sampled air into a refluxing water reservoir that can be analyzed.^62,63 Field deployed MCs have been mostly operated as integrated samplers both outdoors^26,64 and indoors^2,29, but at least one study used a system of two MCs in parallel for real-time sampling of WSOC outdoors.⁶⁵ For the present study, a single MC was modified and configured to operate as a real-time sampler to capture WSOC_g dynamics when coupled to an on-line TOC analyzer. To convert the MC into a real-time sampler, a new exit nozzle was added opposite from the MC inlet nozzle to accommodate a continuous 1.2 mL min⁻¹ outlet flow (peristaltic pump-controlled) that was supplied to the TOC analyzer. The nozzle nearest the inner capillary tube where water is drawn into the mist was used as the inlet nozzle (Figure S4). The MC was initially loaded with 10 mL of deionized water daily and the meniscus level of bulk water inside the mist chamber was maintained by controlling the water flow rate at the MC inlet nozzle using a peristaltic pump set between 1.2 mL min⁻¹ to 1.4 mL min⁻¹ to replace water removed in exit flow and loss to evaporation.

The MC sampled particle-free indoor air by placing a quartz fiber filter (QFF; 47 mm) upstream; this filter provided a low-pressure drop that does not interfere with creating a mist as described by Duncan et al. 2019.²⁹ Particle-free air entering the glass mist chamber encountered a refluxing water mist from the 10 mL reservoir of bulk water. A hydrophobic Teflon membrane (47 mm) at the air flow exit prevented loss of water droplets; which refluxed back into the water reservoir.⁶³ The MC was operated with water reservoir and mist droplet residence times of 8.3 min and 0.43 sec, respectively; the droplet residence time refers to the time a sprayed droplet in the mist is suspended and in contact with sampled air.

Several quality control procedures were implemented (Section S3). Briefly, dynamic blanks were collected by running the MC at the same water flow rate while setting the air flow rate to zero (i.e., without generating mist). The WSOC_g limit of detection (LOD), estimated as three times the standard deviation (σ) of the dynamic blank, was 0.5 μg C m⁻³. The collection efficiency of WSOC_g in MC samplers is largely controlled by Henry’s Law partitioning constants (K_H).^2,63,66 The mean (n=10) collection efficiency of the MC operated in continuous sampling mode was 57 % for the WSOC_g mixture found in indoor air inside an empty university seminar room/teaching kitchen at UNC, determined by operating two real-time MCs in series. It was 63 %, 75 %, 74 %, and 41 % for glyoxal (K_H: 4,000,000 M atm⁻¹), formic acid (K_H: 8,000 M atm⁻¹), acetic acid (K_H: 4,000 M atm⁻¹), and acetone (K_H: 31 M atm⁻¹), respectively. The 57 % collection efficiency, determined in the university teaching kitchen, was used as an effective collection efficiency to correct measurements from the NZERTF under the assumption that the relative composition of WSOC_g was similar in both locations. Experimental set-up (Figure S5 A), sample data (Figure S5 B), collection efficiency calculation (Equation S1) and tabulated collection efficiency results (Table S1) are shown in Section S3. Based on the TOC accuracy (±15 %), air flow uncertainty (±5 %) and relative standard deviation of dynamic blank concentrations (±1 %), the uncertainty for WSOC_g measurements was estimated, by propagation of error, to be ±16 %.

WSOC_p was continuously measured using a PILS and TOC analyzer at 6 min intervals, alternating with TOC analysis of MC sample flow. The PILS uses supersaturated water vapor to condense water on aerosol particles, increasing particle diameters and impaction collection efficiencies. The aerosol is impacted onto a liquid water flow that can be analyzed.^67,68 PILS instruments have previously been coupled to on-line TOC analyzers to measure WSOC_p in outdoor environments, and the method is well-detailed in previous literature.^69-71 The PILS sampled indoor air through a parallel-plate carbon strip denuder (Sunset Laboratory Inc.) to remove organic gases (breakthrough: 11 %; Section S3) and a 2.5 μm aerodynamic diameter cut point round jet impactor (Brethchel Manufacturing Inc.) that removed larger particles, so that the instrument collected WSOC_p in PM_2.5 into a 1.5 mL min⁻¹ aqueous flow. The aqueous sample flow was filtered of insoluble particles using a 0.5 μm in-line liquid PEEK (polyetheretherketone) filter before introduction to the TOC analyzer.

Measures to describe the quality control of WSOC_p are summarized here and provided in more detail in Section S3. For the dynamic blank, a filter was placed at the PILS sample air inlet to remove particles from the sampled air. The LOD was 2.0 μg C m⁻³ calculated as 3σ of the dynamic blank. These LOD values are higher than reported for outdoor WSOC_p measurements,⁶⁹ likely because more variable (and higher) indoor WSOC_g concentrations resulted in more variable (and higher) dynamic blank values. WSOC_p measurements were corrected for WSOC_g breakthrough. The PILS collection efficiency is reported to be 97 % in the particle size range of 0.03 μm to 10 μm.⁶⁷ The on-line PILS-TOC method was previously shown to agree within 10 % of traditional integrated filter methods;^72,73 and have an estimated precision of 10 %.⁶⁹

The TOC was calibrated one week prior to the start of the study using commercial potassium hydrogen phthalate (KHP) calibration standards (range: 0.25 ppm_m to 5 ppm_m), a sucrose verification standard (1 mg-C L⁻¹) and verified calibration blanks (low TOC water); calibration checks were performed throughout the campaign. Note: we define ppm_m as the grams of compound per million grams of solution. The TOC was operated to perform real-time analysis with ± 15 % accuracy, 2 % precision, 4 s analysis time, 3.25 min response and a dynamic analytical range of 0.20 ppb_m to 10 ppm_m. The TOC analyzer sampled the liquid flows at a flow rate of 1.1 mL min⁻¹ through an approximately 1 m long, 0.0254 cm ID, 0.1588 cm OD (0.01 inch ID, 1/16 inch OD) PTFE tubing; excess flow from the MC or PILS was sent to waste or collected in vials and stored frozen. Due to the TOC instrument response time (3.25 min) and the residence time in tubing (about 45 sec), our system had an approximately 4 min system response time to changes in WSOC in indoor air.

Each sampling day began with either a 20 min calibrant check and/or 20 min clean water flush of the TOC analyzer before collection of dynamic blanks and sampling of indoor air. A 20 min calibrant check was performed 1 to 2 times a week using a 2.0 ppm_m KHP solution followed by a 20 min de-ionized water flush. The measured calibrant concentration was 1.915 ppm_m ± 0.076 ppm_m over the duration of the study. MC and PILS dynamic blanks (n=3) were then measured alternately in 6 min intervals for a total of 36 min for blank-correction of WSOC in indoor air. Next, WSOC_g and WSOC_p in indoor air were continuously collected, switching every 6 min between the MC and PILS, respectively, for the duration of the day. The last 2 min of data before each valve switch was used. Measurements of indoor air typically began at least 36 min prior to experimental perturbations in the house to obtain average (n=3) unperturbed WSOC_g and WSOC_p concentrations (house background). On days with early perturbation start times, dynamic blank collection was skipped (2 to 3 times a week) to ensure enough time to collect indoor background measurements before perturbations began. A single day was selected for extended collection of the KHP standard, clean water, and dynamic blanks (Figure S6). In addition, airborne WSOC was measured indoors and then outdoors for 2 h each on a single day (April 11^th) toward the end of the campaign (Section S4, Figure S7).

The building-specific WSOC_g emission rate (Ef) without the introduction of purposeful sources was calculated using Equation 1: (1) Ef=ACR∗V∗(Cin−Cout) where V is the volume of the main house, Cin is the mean WSOC_g house background concentration, and Cout is the outdoor WSOC_g concentration measured on April 11 (mean: 5 ± 1 μg-C m⁻³; see Figure S7).

The following additional measurements were performed in the NZERTF and are described in detail in Section S5. Gas-phase mixing ratios of C2 to C4 water-soluble aldehydes, ketones, and alcohols were measured using a Vocus proton-transfer-reaction (PTR-MS) mass spectrometer.⁷⁴ A fast-scanning gas chromatograph⁷⁵ was periodically interfaced to the inlet of the PTR-MS which provided chromatographic separation for each sampled ion and enabled correction for interferences. The gas-phase mixing ratios of C1 to C5 water-soluble carboxylic acids (i.e., formic, acetic, propanoic, butanoic, and pentanoic acids) were measured using a Time-of-Flight Chemical Ionization Mass Spectrometer (ToF-CIMS) equipped with iodide (I⁻) reagent ions.^76,77 Mixing ratio uncertainties for formic, acetic, propanoic, and pentanoic acids, ethanol and acetaldehyde are estimated to be within 20 %; they are likely factor of two for acetone, butanone, and hydroxyacetone. Formaldehyde and formic acid mixing ratios were measured on the 1^st floor and outside the NZERTF using a quantum cascade tunable infrared direct absorption spectrometer (QC-TILDAS; LOD of 0.1 ppb_v). O₃ mixing ratios were measured using an ozone analyzer (LOD: 0.5 ppb_v) on the 1^st floor and outside the NZERTF.

The vast majority of WSOC in the house background was in the gas phase (Section S6, Table S2). On average, the house background WSOC_g concentration was 36 times higher than WSOC_p. In other words, only 3 % of total airborne WSOC was in the particle phase (F_p). The mean WSOC_g and WSOC_p in the house background were 215 μg-C m⁻³ ± 29 μg-C m⁻³ and 6 μg-C m⁻³ ± 3 μg-C m⁻³, respectively. Using an average organic molecular weight to carbon weight for gaseous water soluble organic matter (WSOM_g) of 2.1, estimated as described below, and an organic molecular weight to carbon mass for particle-phase water-soluble organic matter (WSOM_p) of 2, previously reported for outdoor aerosol,^44,78 the house background concentrations of WSOM_g and WSOM_p were 452 μg m⁻³ and 12 μg m⁻³, respectively. The 2.1 value is the average molecular weight-to-carbon mass ratio for the compounds identified in this study and for elemental formulas of species detected via positive mode high-resolution quadrupole time-of-flight mass spectrometry with an electrospray ion source (ESI-QTOF-MS) in indoor mist chamber sampling by Duncan et al.² (Table 2 in Duncan et al.)².

WSOC_g in the house background was substantial (Table S2), suggesting the importance of contributions from house reservoirs such as building materials to indoor air WSOC concentrations. The indoor-outdoor ratio (I/O) of WSOC_g measured on April 11^th was approximately 19 (outdoor WSOC_g concentration: 5 μg-C m⁻³ ± 1 μg-C m⁻³; Figure S7); indoor WSOC_p was consistently equal to or greater than the outdoor concentration (WSOC_p I/O ≥ 1 using the LOD as an upper-bound estimate for outdoor WSOC_p concentrations below LOD). The reported I/O ratio could be underestimated because it was measured on a day when the house was more open than usual (i.e., frequently opened or propped open front door) resulting in lower indoor air concentrations of airborne WSOC; indoor WSOC_p concentrations were near or below LOD (Figure S7). The fact that a dramatic reduction in the WSOC_g concentration was also observed with window opening, discussed later, also supports the conclusion that there is a substantial indoor source of WSOC_g.

The building-specific emission rate of WSOC_g was roughly 31 mg-C h⁻¹. Normalized to the area associated with painted walls, painted ceilings, and wooden floors in the main house, the surface area-specific E_f was approximately 18 μg-C h⁻¹ m⁻². Average whole-house emission factors for various VOCs which included water-soluble organic acids, alcohols, aldehydes, ketones and glycols were previously measured over a 15 month sampling period 1 year after the NZERTF was constructed (2013).⁵⁹ When summed, the average emission factors from January to July 2014 equated to an average WSOC_g emission rate of 148 mg-C h⁻¹ (using only measured compounds with K_H > 10 M atm⁻¹);⁵⁹ note this study did not directly measure total WSOC and so the sum of measured water-soluble species provides a lower bound for total WSOCg emissions. Our E_f suggests at least a 79 % reduction in (total) WSOC_g emissions in the 8 years that followed; more substantial reduction likely occurred in the first few years after construction.⁷⁹

The sum of measured water-soluble species accounted for 51 % to 72 % of the WSOC_g house background mass (Figure 2; Table S3). The remaining approximately 28 % to 49 % is expected to include organic compounds with K_H values from 10¹ M atm⁻¹ to greater than 10³ M atm⁻¹, which we verified we collected with efficiencies of 41 % to greater than 75 %. The most abundant species from each compound class were ethanol (46 μg-C m⁻³ to 82 μg-C m⁻³), acetic acid (12 μg-C m⁻³ to 24 μg-C m⁻³), acetone (11 μg-C m⁻³ to 16 μg-C m⁻³), and acetaldehyde (7 μg-C m⁻³ to 9 μg-C m⁻³), which have been measured in a variety of indoor air studies and have known building- and human-derived sources of acetic acid^{2,6,29,32,33,80,81}, acetone^{2,30,32,33,80-84}, acetaldehyde^{30,33,81,83-86,23,82,84-87}). Notably, they could be emitted from building materials such as wood-based products,^86-89 paints,^33,90,91 and finishes (e.g., adhesives).^59,92,93 Oxidation chemistry is also a potential source.^34,94-96

Figures 3-4 and Table 1 show the effect of several perturbations on WSOC_g and WSOC_p in the house (NZERTF). Some perturbations increased (e.g., cooking with vinegar, wood smoke additions, CASA cocktail addition, ozone injection, and closing windows) or decreased (e.g., surface cleaning and window opening) WSOC, but others did not result in a measurable change in airborne WSOC (e.g., pesticide product additions, inorganic acid/base additions, portable air cleaner operation and selected cooking activities that did not include vinegar).

Surface cleaning inside the wood smoke contaminated NZERTF on April 7^th caused a reduction in WSOC_g concentrations (Figure 3). First, a small peak of WSOC_p (8 μg-C m⁻³) was observed as researchers were in the house preparing to clean, presumably because of resuspension of particles from researchers’ movement.⁹⁷ Then, following vacuuming, dusting, mopping and surface wiping, WSOC_g concentrations decayed at a rate of 0.42 h⁻¹, decreasing by 39 % after surfaces were cleaned, which may be due to removal of WSOC reservoirs on wood smoke-contaminated surfaces. Li et al. observed concentration reductions of individual water-soluble organic gases, specifically formic acid and formaldehyde, during the same time period inside the NZERTF and previously drew the same conclusion.⁷⁷ Air cleaners used on a separate day after smoke injections (March 24) did not result in an observable change in WSOC (see Table 1 and Section S7, Figure S8).

The “Window Opening and Closing” perturbation (Figure 3) provided evidence consistent with emissions and air concentrations being influenced by surface reservoirs. WSOC_g concentrations were diluted upon window opening and reduced by 81 % from 147 μg-C m⁻³ to 28 μg-C m⁻³ before rebounding after windows were closed, increasing from 28 μg-C m⁻³ to 125 μg-C m⁻³. The dramatic decrease with window opening and subsequent rebound provided additional evidence that WSOC in building materials/reservoirs are responsible for the high house background. WSOC_g decayed at a rate of 2.8 h⁻¹ post window opening due to dilution with outdoor air and an increased air change rate (outdoor ACR = 21.2 h⁻¹; see Section S8, Figure S9 & Figure S10). The decrease in air concentrations also likely increased emissions from surfaces reservoirs, causing WSOC_g concentrations to decay slower than they would if the emission rates were constant. The estimated E_f of WSOC_g during the steady-state period of window opening was 301 mg-C h⁻¹, which is much greater than the steady-state E_f associated with the typical house background (31 mg-C h⁻¹) and before windows were opened (21 mg-C h⁻¹). Approximately 90 minutes after closing windows, the WSOC_g concentration had nearly returned to pre-perturbation levels. The characteristic rate for WSOC_g rebound from window closing was 0.8 h⁻¹, which is faster than the ventilation rate (outdoor ACR). Gas exchange with surface reservoirs is likely responsible for this rapid rise in the indoor air concentration of WSOC_g.^16,19,40 Effects on WSOC_p concentrations were too small to measure but concentrations appear to slightly decrease upon window opening and increase after window closing. It is important to note that the measurement of such a high outdoor ventilation rate in the fully open house (i.e., during the window opening perturbation; Section S1) is expected to be somewhat uncertain due to increased variations in interzonal mixing, but even an error of 50 % would still produce a substantially increased E_f during the period of window opening.

The change in WSOC from cooking activities depended on the food items cooked. Cooking bacon, peppers, and tater tots (pan fry or air fry) did not result in a measurable response in either gas or particle phase WSOC, likely due to low water-solubility of cooking emissions (e.g., cooking oils and long-chain fatty acids). However, cooking with vinegar did produce a large WSOC_g response due to acetic acid emissions. The first peak (approximately 450 μg-C m⁻³) in Figure 4a shows a large WSOC_g peak from boiling 2 tablespoons of balsamic vinegar to produce a balsamic glaze. The second peak (approximately 350 μg C m⁻³) shows the signal from cooking a similar glaze with 2 tablespoons of balsamic vinegar, but with vegetables included in the dish. The smaller peak is likely due to the other food ingredients absorbing some of the acetic acid leading to less vinegar evaporating from the glaze. Although an increase in indoor PM_2.5 was observed during the second cooking event (with vegetables; up to 10 μg m⁻³ to 15 μg m⁻³), WSOC_p did not increase.

The addition of the CASA cocktail (Figure 4b) and wood smoke (Figure 4c) significantly increased indoor WSOC concentrations over background levels, but pesticide addition did not result in a measurable change in WSOC (Table 1). The pesticide product contained water insoluble or sparingly soluble chemicals that were not expected to be captured in our system (see SI Section S1). Injection of the CASA cocktail, which contained many compounds that are relatively water soluble, resulted in a 100 μg C m⁻³ to 390 μg C m⁻³ increase in WSOC_g (Table 1; Figure 4b). wood smoke injections increased WSOC_g and WSOC_p by 25 μg C m⁻³ to 220 μg C m⁻³ and 16 μg C m⁻³ to 100 μg C m⁻³, respectively (Table 1; Figure 4c). WSOC_p concentrations were highest for wood smoke, with gas to particle concentration ratios of approximately 1 to 2. The WSOC system was run for only 1 of the 2 high RH (approximately 74 % RH) smoke injections, precluding investigation of an RH effect on airborne WSOC from wood smoke injections. Li et al. reported the bulk organic chemical composition of the CASA campaign wood smoke aerosols was 95 % organic, 3 % nitrate, 1 % sulfate, and 1 % ammonium, and that gaseous wood smoke emissions included formic acid, acetic acid, and formaldehyde.⁷⁷ Based on Li et al. organic aerosol data⁷⁷ and an OM:OC conversion factor of 2.0 derived from fireplace wood combustion,^78,98 about 32 % to 67 % of wood smoke organic aerosol mass (OM) is water soluble.

The addition of inorganic species (i.e., ozone, NH₃, or CO₂; Section S9) had moderate to no impact on WSOC levels measured by our system. Although the MC chamber is well-suited for collecting oxidation products,⁶³ ozone additions did not consistently produce a measurable increase in WSOC_p or WSOC_g. A change in WSOC in our system was observed for 3 out of the 9 ozone additions that took place during times of WSOC sampling, perhaps due to the high loading of WSOC_g in the house air to begin with (house background). In some cases spikes in WSOC were observed when ozone was added shortly after wood smoke injections (Table 1, Figure 4c); the mixing ratio of O₃ in the main house was typically about 10 ppb_v when not intentionally perturbed. Additions of basic NH₃ and acidic CO₂ did not induce a measurable change in WSOC, but measurements of gas-phase total inorganic carbon (TIC) in the system became more sensitive to indoor CO₂ after NH₃ injection, presumably due to pH elevation of MC water as discussed in Section S9 (see Figure S11).

WSOC_g peaks decayed at rates ranging from 0.4 h⁻¹ to 4.0 h⁻¹ across the perturbations (Table 2). These rates are 1.7 to 16.7 times faster than the ventilation rate, suggesting the presence of surface reservoirs that amplified the loss rates of WSOC_g in air. When surface reservoirs and gas-phase concentrations approach steady-state, outdoor ventilation and exfiltration become the dominant removal mechanisms. Decay rates were influenced by internal mixing (1.3 h⁻¹), surface removal rates, and ventilation (0.25 h⁻¹), slowing with time. For example, decay of the CASA cocktail injection was rapid to begin with (initial decay rate: 2.6 h⁻¹ ± 0.02 h⁻¹), perhaps associated with some mixing with the rest of the house, but also uptake by surfaces (see Figure S12). The initial peak decay rate held for about 15 minutes before the decay rate reduced (later decay rate: 1.7 ± 0.02 h⁻¹), indicating that surface reservoirs were beginning to shift away from being a net sink, but had not yet become a net source (see Figure S12). Long-term decay (i.e., longer than 2 h to 3 h) of WSOC when concentrations were just elevated over normal conditions and associated variability were not captured in our system. After internal mixing is complete and there is neither net surface uptake nor removal, decay rates reduce to the ventilation rate, consistent with tracer gas (SF₆) decay.

Wood smoke injections introduced water soluble organic gases in indoor air that were not detected in the house background. Based on the mass closure on peak WSOC_g concentrations (54.1 μg-C m⁻³) during wood smoke injections, 75 % (by mass) of WSOC_g in the wood smoke was uncharacterized and 25 % was composed of gases also detected in the house background (Figure 5). The most abundant of the characterized fractions was formaldehyde (6.5 μg-C m⁻³) and acetaldehyde (2.6 μg-C m⁻³), which comprised 17 % of peak WSOC_g during the perturbation.

Decay rates that were larger than the ventilation rate, followed by concentration rebound when windows and doors were closed after periods of elevated outdoor ventilation, demonstrate the substantial role that surface reservoirs play in WSOC_g concentration dynamics (i.e., the building surfaces themselves, irrespective of house activities). Previous studies point to differences in the behavior of VOCs compared to semi-volatile organic compounds (SVOCs) in response to window opening and closing; comparison with these past studies aids our understanding of the behavior of WSOC_g. Although adequate ventilation reduces airborne concentrations, it does not reduce or generally influence building material emission rates of nonpolar VOCs because slow bulk diffusion within deep reservoirs in the material limits the emission rate.^21-23 SVOC emission rates are typically enhanced when the concentration gradient between the material and air increases (i.e., by increasing ventilation), suggesting emission is not limited by diffusion in the material (i.e., SVOCs are readily available from near surface reservoirs). ^24,25 This is because SVOCs are more uniformly distributed in their host materials and evaporate more slowly at the surface than VOCs.^24,25 We observed that WSOC_g emissions are enhanced at elevated ventilation, similar to the emission behavior of SVOCs and individual polar VOCs (e.g., formaldehyde)^22,85,104. This behavior suggests that WSOC (or oxygenated VOCs) are readily available from near surface reservoirs and emissions are not limited by diffusion in the materials. The indoor WSOC_g concentration reduced by 81 % from 147 μg-C m⁻³ to 28 μg-C m⁻³ during the period of elevated outdoor ventilation associated with open windows which was 5.6 times higher than outdoors (I/O=5.6). If emission rates were instead constant and independent of the ventilation rate, the I/O ratio would have approached 1.0 and the WSOC_g concentration would have been reduced by 95 % to 7 μg-C m⁻³.

Although this study was performed inside a test house, data was discussed in the context of related indoor measurements from literature that were often performed in real, occupied homes. The homes across various indoor air studies are not identical buildings and will have considerable variations in conditions (e.g., ventilation rate, dimensions, contents, materials, outdoor air composition, and human activities). Major differences between the NZERTF and typical homes include a lack of substantial occupancy, furniture, exposure history (e.g., years of cooking emissions), the use of low-VOC emitting building materials in construction, and building envelope air leakage rates at the lower end of the range for typical homes. Therefore, comparisons to the measurements taken inside the NZERTF test house are used to provide insights into how the building (i.e., walls, flooring, and ceilings) in the absence of typical features of occupied residences may contribute to and control airborne WSOC, and how perturbations induce spikes of WSOC above the house background. The high house background prevented measurements of relatively small changes in WSOC_g. Also, it should be noted that a QFF was used to remove particles upstream of the MC sampler. Organic gases can adsorb to QFFs, reducing measured WSOC_g concentrations until gases come into equilibrium between the QFF and the gas phase. Thus, reported WSOC_g concentrations should be considered lower bound values.