Compartmentalization and ventilation system impacts on air and contaminant transport for multifamily buildings

Abstract

Provision of acceptable indoor air quality in multifamily buildings (MFBs) depends on the interior air flows that impact dilution of contaminants, cross-contamination between units and building energy use. The airtightness of interior partitions and design of ventilation systems in MFBs determine the flows across building partitions. These flows change the total ventilation rate for the building and individual units, and impact the mixing of air and contaminants between apartment units or with common spaces. This study examines the changes in air flow and contaminant transport in MFBs using combined CONTAM/EnergyPlus models. Key parameters were systematically varied, including climate, apartment airtightness, and mechanical ventilation system type. Simulations were performed for mid-rise buildings with and without an enclosed common corridor, and a 20-story high-rise building. Contaminants simulated in the analysis were PM_2.5, formaldehyde, water vapor, and CO₂. Key results of this work are that current airtightness requirements in ASHRAE 62.2 sufficiently limit transport of key contaminants, independent of the type of ventilation system across all three building typologies, and significantly reduce energy use in colder climates. The results of this work are intended to assist codes and standards bodies in setting appropriate airtightness limits and ventilation system design guidelines for MFBs.

HIGHLIGHTS

• Tightening units below ASHRAE 62.2’s compartmentalization limit provided diminishing returns for controlling inter-unit contaminant transfer

• Increased compartmentalization provided better control of air flows and contaminant concentrations, resulting in less extreme peak concentrations

• Mechanical ventilation system type had a marginal impact on the level of inter-unit contaminant transfer.

• Compartmentalization significantly reduced space conditioning energy loads in colder climates

Keywords:

• Indoor contaminant concentrations
• multifamily buildings
• compartmentalization
• airtightness
• CONTAM
• EnergyPlus
• ventilation system performance

1. Introduction

Multifamily buildings have internal airflows that transport air, contaminants, and heat within the building, both from dwelling-to-dwelling and between dwellings and other common spaces. The ASHRAE Handbook of Fundamentals (ASHRAE, 2021) discusses how these flows result from internal leakage pathways and are driven by pressure differences due to wind, stack effects, and mechanical systems. The effects of internal stack-driven flows on heating loads in multi-story buildings are discussed in detail in (Lozinsky & Touchie, 2020; Yoon et al., 2019). A recent review by Lozinsky and Touchie (2020) showed that uncontrolled inter-zonal airflow can allow contaminant transport between dwelling units, decrease ventilation system efficiency, and increase overall building energy use.

To reduce these negative impacts, various recommendations are emerging to limit allowable internal leakage (also referred to as compartmentalization). Many states in the US have adopted International Energy Conservation Code (IECC) requirements in new construction that include air leakage limits for both single and multifamily buildings. The IECC metric is given in Air Changes Per Hour at 50 Pa (ACH₅₀) to the exterior and at least one study (Bohac et al., 2020) has developed a conversion factor for converting between metrics, using a multiplier of 0.13 for converting ACH₅₀ to cfm₅₀/ft² (or 0.65 to convert to L/s₅₀/m²). ACH₅₀ is also used in many international building codes and regulations, but some countries use other metrics, e.g. France uses m³/h/m² at 4 Pa. More detailed summaries of airtightness requirements in different countries can be found in (Leprince et al., 2017; Poza-Casado et al., 2020; RDH Building Engineering, 2013). The ASHRAE 62.2 Standard (ANSI/ASHRAE, 2022) specifies a maximum dwelling unit air leakage rate of 1.0 L/s₅₀/m² (0.2 cfm₅₀/ft²), based on an analysis of over 2,000 test results in new buildings from the US EPA/DOE ENERGY STAR Multifamily Highrise Program showing that this was a readily achievable tightness level. California’s building performance regulations includes an optional compliance path at a level of 1.5 L/s₅₀/m² (0.3 cfm50/ft²), which mirrors guidance from the US ENERGY STAR Multifamily New Construction National Program Requirements (EnergyStar, 2023).

Field measurements of compartmentalization and leakage rates in multifamily buildings confirm that newly constructed buildings can consistently meet the performance standards described above. A recent study of 25 new low-rise multifamily buildings (Bohac et al., 2020) found a median dwelling unit air leakage rate of 0.85 L/s₅₀/m² (0.17 cfm₅₀/ft^2). Another recent study (Lozinsky & Touchie, 2021) found that 12 mostly high-rise new buildings had average unit tightness of 0.49 L/s/m² (0.10 cfm/ft²). Studies of older buildings have found slightly higher leakage levels. A sample of six units constructed between the 1980s and early 2000s (Finch, 2007) had an average leakage of 1.44 L/s₅₀/m² (0.28 cfm₅₀/ft²), with the leakiest unit having an air leakage rate of 3.41 L/s₅₀/m² (0.67 cfm₅₀/ft²). Another study, consisting of six units constructed in the 1970s measured dwelling unit air leakage rates between 1.15 L/s₅₀/m² (0.23 cfm₅₀/ft²) and 1.59 L/s₅₀/m² (0.31 cfm₅₀/ft²) (Gulay et al., 1993).

The connection between Indoor Air Quality (IAQ) metrics and interior leakage are not well established. Some field studies have reported metrics of IAQ in multifamily buildings (Bohac et al., 2007, 2011; Enermodal Engineering Ltd, 2002; Gulay et al., 1993; King et al., 2010; Kraev et al., 2009; Wilson et al., 2020), but most do not provide a comprehensive view of indoor carbon dioxide (CO₂), volatile organic compounds (VOCs) (including formaldehyde), carbon monoxide (CO), bioaerosols, and Environmental Tobacco Smoke (ETS) constituents, and very few measured interior airtightness or inter-zonal air flow along with these IAQ conditions. Only three of the studies completed measurements before and after interior air sealing and/or ventilation system retrofits (Bohac et al., 2007, 2011; Wilson et al., 2020), and unfortunately, except for Bohac et al. (2007, 2011), none of the studies measured interior airtightness or inter-zonal air flow. This lack of consistent and comprehensive measurement makes it difficult to draw comparisons between contaminant concentrations and dwelling unit compartmentalization. From the three studies that completed pre/post retrofit measurements, increasing dwelling unit exhaust air flow rates significantly reduced apartment CO₂ concentrations, but had no significant impact on CO or VOC concentrations (Wilson et al., 2020). Bohac et al. (2011) found that air sealing reduced inter-dwelling unit air flows by a median of 29%, based on tracer gas testing. Nicotine concentrations in non-smoking dwelling units were also lower following interior air sealing and ventilation system retrofits. However, pre-air sealing nicotine concentrations in non-smoking dwelling units ranged from 0.0 (< Limit of Detection (LOD)) to 0.4 μg/m³, so the reductions were minimal. It is worth noting that most of the field studies mentioned above are based on short-term measurements over hours or days, and are not necessarily representative of longer-term average building conditions. The external validity of the field studies is also limited, as all of these studies only assessed building performance in a single climate zone.

To support the limited field data, simulation studies have demonstrated that ventilation system design and/or interior compartmentalization can reduce inter-zonal air flows, and that this compartmentalization provides energy-savings benefits. (Carlsson et al., 2019) found that switching from a pressurised corridor ventilation system to in-unit Heat Recovery Ventilators (HRV) and tightening dwelling unit entry doors reduced building infiltration rates. Dwelling unit heating loads were also reduced by 39% (in a Vancouver, BC climate). (Markley et al., 2014) found that reducing whole dwelling air leakage from 2.0 L/s/m² to 1.0 L/s/m² (0.40 cfm₅₀/ft² to 0.20 cfm₅₀/ft²) reduced building heating loads by 6-8%, for various California climates. Markley et al. also found that inter-zonal flow decreased with increased compartmentalization. Neither study looked at contaminant transport. (Underhill et al., 2020) completed a parametric simulation analysis using coupled CONTAM (CONTAM, 2002) and EnergyPlus (EnergyPlus, 1999) simulations to assess the IAQ and energy implications of building envelope and ventilation system retrofits. They showed improved airtightness combined with mechanical ventilation reduced indoor mass concentrations of particulate matter with a diameter less than 2.5 μm (PM_2.5) by 15-25%. (Mckeen & Liao, 2022) and (Mao et al., 2015) used CONTAM simulations to model inter-zonal contaminant transport in high-rise residential buildings during winter conditions. Both studies used CO₂ as a proxy for airborne contaminants and found that building airtightness characteristics impact inter-zonal air flows; however, both studies assumed that inter-zonal airflow only occurred via interior doors and that interior partitions (i.e. floors, ceilings, and walls between dwellings) were completely airtight, which likely resulted in underestimation of inter-zonal contaminant transport. (Fabian et al., 2016) modelled PM_2.5 associated with Environmental Tobacco Smoke (ETS) in a townhouse complex using CONTAM. Consistent with other studies, the authors found that inter-zonal contaminant transport decreased with interior air sealing interventions.

While the term compartmentalization refers to internal leakage only, field measurement methods and metrics to assess compartmentalization commonly include both internal and external leakage combined together. In this paper we use the term ‘compartmentalization’ to mean all the leakage for a dwelling unit, consistent with current measurement metrics. The key metric used to assess compartmentalization is to normalise the dwelling unit leakage at 50 Pa (0.2 in. water) (L/s₅₀ or cfm₅₀)) by dividing by the six-sided surface area of the dwelling to obtain an airflow per unit area (L/s₅₀/m² or cfm₅₀/ft²). The leakage at 50 Pa is measured using fan pressurization techniques. These are commonly based on procedures developed for single zone buildings (e.g. ASTM E779 (ASTM International, 2018) or ISO 9972 (ISO, 2015)) although some test methods have specific instructions for testing multifamily buildings (ANSI/RESNET/ICC, 2018). Some US air leakage testing requirements use a 75 Pa (0.3 in. water) reference pressure (e.g. ASHRAE Standard 189.1 (ASHRAE, 2020; U.S. Army Corps of Engineers, 2012)).

Even if the leakage is known, converting this to air flows that can impact contaminant transport and exposure, or lead to changes in energy use requires that the pressures driving these flows are also known. Previous studies have investigated air flows and contaminant transport using both field measurements and simulations. Field studies have shown a large variability in internal air flows, with the fraction of flows from other dwelling units ranging from 2 to 25% of total air flow (Bohac et al., 2007; Francisco & Palmiter, 1994; Harrje et al., 1988; Moffatt et al., 1998; Ricketts, 2014; Wilson et al., 2020; Wray, 2002). The large range in fractional flows was due to differences in airtightness, weather, building type and height, and operation of ventilation systems. Transport of contaminants via these airflows can also be reduced by filtration and deposition mechanisms. For example, a recent field study found that transport of PM_2.5 was unmeasurably low (Modera et al., 2023) and another study found that changes in transport of nicotine (as a proxy for ETS) was much lower than the changes in air leakage (Bohac et al., 2011).

In this paper, we expand on the previous work described above to examine how contaminant concentrations and airflows depend on building typology, weather conditions, the degree of compartmentalization, and ventilation system configuration. This study combined overall contaminant concentration analyses based on typical contaminant source strengths in every dwelling unit, as well as the release of contaminants in target dwelling units, which allowed for tracking of contaminants from individual dwelling units as they spread through the building. Some preliminary results for the four-story, mid-rise building have been presented previously (Walker et al., 2022). The study also included specific parametric analyses that are topics of interest for standard-setting bodies (such as ASHRAE 62.2). This work was directed towards answering several key questions:

• Are current compartmentalization requirements in codes and standards adequate?

• Can we recommend a level of compartmentalization that minimises internal air flows to the point where there are diminishing returns for IAQ?

• At that level of compartmentalization, does it matter what ventilation system is used?

• Are different ventilation systems more or less sensitive to compartmentalization?

• What are the space conditioning energy impacts of compartmentalization?

2. Methods

The energy and IAQ effects of compartmentalization were investigated by simulating energy use, ventilation air flows and contaminant transport in multifamily dwellings over a range of airtightness levels. We used a co-simulation technique over a full year of weather conditions. The IAQ analysis assumed that all dwellings were occupied and were impacted by contaminants of both indoor and outdoor origin. A key issue is that we are assessing airtightness and compartmentalization in ASHRAE 62.2 that only deals with chronic exposure. Therefore, our simulations and analysis of the IAQ results focused on long-term chronic exposures (annual averages) rather than short-term fluctuations or occupancy changes.

2.1. Prototype buildings

Building geometry, building envelope, and heating, ventilation, and air conditioning (HVAC) equipment were developed based on the Pacific Northwest National Laboratory (PNNL) mid-rise multifamily prototype model used in energy code analysis (Department of Energy (DOE), n.d.). The PNNL mid-rise multifamily prototype is a 4 storey building, with eight dwelling units per floor, a common corridor, and stairwell and elevator shafts on each end of the corridor, as shown in Figure 1. In addition to this typology (hereafter referred to as ‘mid-rise with a common corridor’), two other typologies were simulated: (1) mid-rise without a common corridor (mid-rise walk up); and (2) high-rise (20 story) with a common corridor. Simulations for the ‘mid-rise walk-up’ typology removed dwelling units 5—8, the common corridor, and the stairwell and elevator shaft zones. Simulations for the ‘high-rise with a common corridor’ typology added 16 additional floors. All other model parameters and inputs remained the same between the three typologies. Each dwelling unit has a floor area of 88.25 m² (950 ft²) and is assumed to house four occupants (two adults and two children). Simulations assumed that the dwelling units were always occupied, with CO₂ and moisture emissions, as well as thermal loads estimated from biological activity. Each dwelling unit was modeled as a single zone, ignoring any interior partitions.

Figure 1. Layout of a prototypical building floor for the mid-rise and high-rise common corridor simulations (Typologies (1) and (3), respectively). The mid-rise and high-rise buildings consisted of 4 and 20 floors respectively, each with identical floor layouts.

The HVAC equipment specified in the PNNL prototype models were used, with each dwelling unit and corridor zone serviced with a gas furnace with 80 Annual Fuel Utilization Efficiency (AFUE) and direct expansion (DX) cooling coils with a rated COP of 3.6. Heating and cooling thermostat set-points were: 21.1 °C (70 °F) for heating and 24.0 °C (75.2 °F) for cooling. Notable departures from the PNNL prototype models include simulating each dwelling unit independently (i.e. avoiding the use of zone multipliers) and the addition of elevator and stairwell zones. In addition, the ventilation equipment and air flow calculations in the PNNL prototype models were abandoned in favor of the nuanced approach leveraging CONTAM co-simulation described below.

2.2. Simulation framework

The buildings were simulated using EnergyPlus v9.1. The airflow calculations, and most transport estimates, were performed by CONTAM v3.4, using its Functional Mockup Unit (FMU) co-simulation capability. Moisture transport was simulated in EnergyPlus, in order to take advantage of its advanced models for surface interactions. During co-simulation, at each timestep EnergyPlus passed CONTAM the zone temperatures and humidity ratios, ventilation system airflows, weather data, generation rates for contaminant sources, and other model parameters. CONTAM returned the infiltration and zone-to-zone airflows, along with the concentrations of the contaminants of interest. The simulations were run at a 3-min time-step, but outputs were recorded at only one-month intervals. All outputs are reported either as mean values or as sums over the month. The short time step was required to allow for adequate representation of short time-scale events (e.g. cooking and the operation of intermittent exhaust fans) and for future analyses using real-time ventilation controls that are not part of this paper. The short time step also minimised any issues due to the time-step lag introduced between EnergyPlus and CONTAM by co-simulation. Refer to Dols et al. (2016) and (Justo Alonso et al., 2022) for a detailed description of the coupling process.

In total, five simulation parameters were varied: 1) climate zone; 2) ventilation system type; 3) dwelling unit leakage (compartmentalization); 4) ambient particle levels, and 5) particle filtration, with a total of 1,100 combinations.

2.2.1. Climate zone

We considered five EnergyPlus climate zones (CZs) to cover the range of conditions experienced in the continental United States: CZ 2 A (Hot humid, Tampa, FL), CZ 2B (Hot dry, Tucson, AZ), CZ 3 C (Warm marine, San Diego, CA), CZ 4 A (Mixed humid, New York, NY) and CZ 7 (Very Cold/International Falls, MN). The hourly TMY3 weather data (Wilcox & Marion, 2008) were linearly interpolated by EnergyPlus within each hour to accommodate the 3-min time steps used in the simulation.

2.2.2. Ventilation system

We considered 11 whole dwelling dilution ventilation system types, summarised in Table 1. Dwelling unit mechanical fan flow rates met the minimum calculated requirement of 27.5 L/s (58 cfm) based on ASHRAE 62.2-2019 (ANSI/ASHRAE, 2022) 19). The trickle vent cases are notable for having small, medium, and large fan flow rates, representing 100, 150 and 200% of the code minimum flow rate, respectively. These options have been suggested as potential compliance pathways for future versions of the ASHRAE standard. The corridor on each floor is supplied with outside air at a minimum rate of 19.2 L/s (41 cfm) (based on ASHRAE 62.1 requirements of 0.3 L/s/m² (0.06 cfm/ft²) of floor area). The sole exception is the Unit Exhaust with Corridor Supply case, where outside air is intentionally delivered to each unit via the corridor based on the sum of the supply ventilation flows for each dwelling unit (220 L/s (466 cfm)). The Balanced HRV and Balanced energy recovery ventilator (ERV) cases include sensible (70%) heat recovery, and the Balanced ERV also includes latent (60%) heat recovery of moisture.

Table 1. Whole dwelling dilution ventilation system types.

Ventilation Type	Dwelling Exhaust Flow L/s (cfm)	Dwelling Supply Flow L/s (cfm)	Corridor Supply Flow L/s (cfm)
Unit Balanced HRV	27.5 (58)	27.5 (58)	19.2 (41)
Unit Balanced ERV	27.5 (58)	27.5 (58)	19.2 (41)
Unit Exhaust, Corridor Supply	27.5 (58)	0	220 (466)
Unit Exhaust, 5 Pa Trickle Vent, Small	27.5 (58)	0	19.2 (41)
Unit Exhaust, 5 Pa Trickle Vent, Medium	41.25 (87)	0	19.2 (41)
Unit Exhaust, 5 Pa Trickle Vent, Large	55.0 (116)	0	19.2 (41)
Unit Exhaust, 10 Pa Trickle Vent, Small	27.5 (58)	0	19.2 (41)
Unit Exhaust, 10 Pa Trickle Vent, Medium	41.25 (87)	0	19.2 (41)
Unit Exhaust, 10 Pa Trickle Vent, Large	55.0 (116)	0	19.2 (41)
Unit Supply	0	27.5 (58)	19.2 (41)
None	0	0	0

In this paper, our analysis and figures include only one trickle vent case (‘Unit Exhaust with 5 Pa Trickle Vent, Small’ relabeled as ‘Unit Exhaust, Trickle Vent’), because we only want to compare ventilation types providing the same design airflow. Throughout, the ‘Unit Exhaust, Trickle Vent’ can be considered representative of the other ‘Trickle Vent’ cases shown in Table 1. Similarly, the HRV air flow results are used in this paper to represent both ERV and HRV ventilation types, due to their similar impacts on airflow and contaminant transport.

Each dwelling unit was assumed to be equipped with intermittent kitchen (50 L/s (106 cfm)) and bathroom (25 L/s (53 cfm)) exhaust fans. Each unit also included laundry exhaust (37.5 L/s (79 cfm)). These intermittent local exhaust flows in all dwelling units were controlled on an identical, fixed daily schedule (see Table 2). For the kitchen and bath exhausts, this coincides with the times of contaminant emissions and sensible heat gains from cooking and bathing.

Table 2. Schedule of activities and local exhaust fan operation.

Time		Activities	Kitchen Fan L/s (cfm)	Bathroom Fan L/s (cfm)	Laundry Fan L/s (cfm)
Start	End
7:00	7:30	Showering	0	25 (53)	0
7:30	8:00	Cooking and Showering	50 (106)	25 (53)	0
11:45	12:15	Cooking	50 (106)	0	0
18:00	18:30	Cooking	50 (106)	0	0
21:00	21:30	Dishwasher	0	0	0
21:30	22:00	Laundry and Dishwasher	0	0	37.5 (79)

2.2.3. Compartmentalization

We considered five compartmentalization levels: 1. Typical (5.1 L/s₅₀/m² (1.0 cfm₅₀/ft²)); 2. Current Practice (1.5 L/s₅₀/m² (0.3 cfm₅₀/ft²)); 3. Moderate Target for better performance (1.0 L/s₅₀/m² (0.2 cfm₅₀/ft²)); 4. Tight (0.5 L/s₅₀/m² (0.1 cfm₅₀/ft²)); and 5. Super Tight (0.25 L/s₅₀/m² (0.05 cfm₅₀/ft²)). We distributed leakage areas to the envelope surfaces based on the results of field studies in low-rise multifamily buildings (Ricketts, 2014; Bohac et al., 2007, 2020). The total dwelling leakage area derived from the compartmentalization level was distributed as follows: 2.5% to each party wall, 10% to each floor or ceiling surface, 45% to the corridor wall and 30% to exterior wall surfaces. The split between interior and exterior leakage is broadly similar to results of other studies summarised in Lozinsky and Touchie (2020) and for high rise buildings in Lozinsky and Touchie (2023). Several types of dwellings units required special leakage area distributions, including ground floor units, corner units and top floor units. In corner units, which have only one party wall and two exterior walls, the 2.5% leakage was added to the typical 30% exterior wall leakage. For ground-floor dwelling units, the building was assumed to have a slab-on-grade foundation, and all floor leaks were eliminated. In top floor units, the ceiling surface leakage to outside was treated as 10% of the total leakage area. Each envelope leakage element was treated using the power law formulation in CONTAM, with a discharge coefficient of 1.0, pressure exponent of 0.67, and a reference pressure of 4 Pa.

The model also included discrete flow paths that were not included in the dwelling unit compartmentalization values described above. Elevator door leakage was estimated based on the results of several studies (Ricketts, 2014; Tian et al., 2020; Yoon et al., 2019) to be 300 cm² (46.5 in²) (although some studies (Jo et al., 2007; Tamura & Shaw, 1976) give values closer to 400 cm² (62 in²)). The stairwell door leakage was 200 cm² (0.22 ft²). Dwelling unit entry door undercuts were assumed to be 13 cm² (2.0 in²) (Tian et al., 2020) for all ventilation types except for Unit Exhaust, Corridor Supply, which assumed dwelling unit entry door undercuts of 210 cm² (32.6 in²) (Moffatt et al., 1998). Trickle vent system types included in each dwelling unit an exterior wall leak at 1.0 m (3.3 ft) above floor height acting as make-up air for exhaust fans. Trickle vents were sized based on guidance from European manufacturers to have 5 or 10 Pa (0.02 or 0.04 in. water) design pressures at the dwelling unit exhaust fan flow. The discrete flow paths were treated as orifices using the power law formulation in CONTAM, with a discharge coefficient of 0.6, pressure exponent of 0.5, and a Reynolds number of 30.

2.2.4. Filtration and particle removal

Two particle filtration scenarios were considered: 1) no PM_2.5 filtration; and 2) Minimum Efficiency Reporting Value (MERV) 13 with 90% PM_2.5 removal efficiency on all mechanical supply air flows. Filtration was included on dwelling unit supply ventilation air flows, corridor supply ventilation air flows, and dwelling unit recirculated air flows from the heating and cooling system. Particles were also removed as they passed through interior and exterior envelope leakage elements, with a removal efficiency of 50%. This removal rate is based on field tests of outdoor particle removal as they pass through building envelopes (Singer et al., 2016). The same removal efficiency was applied to interior partitions. Field studies (Bohac et al., 2011; Modera et al., 2023) have observed that very few particles are transported to other units in multifamily buildings supporting the need to include particle removal for this air flow path. Particles were also removed by deposition at a rate of at 0.6/h.

2.2.5. Ambient particles

Typical and worst-case scenarios were considered for PM_2.5. 665 US EPA monitoring sites across the continental US were reviewed. We selected specific locations representing the 50^th percentile for a typical scenario (Sussex, Delaware, Site #1002, mean of 8.1 µg/m³ (5⋅10⁻¹⁰lb/ft³) and the 99^th percentile for the worst-case scenario (Los Angeles, California, Site #4008, mean value of 14.8 µg/m³ (9⋅10⁻¹⁰lb/ft³)). For each of these sites, typical diurnal hourly patterns for weekdays and weekends for each month of the year (24 diurnal patterns in total) were calculated and used to assemble an entire year of hourly ambient particle data for each site. In this paper, we only present results based on the Typical ambient particles, because worst-case ambient particles had no observable impact on inter-unit transport.

2.2.6. Contaminant emissions

Four contaminants were considered: water vapor, CO₂, formaldehyde (CH₂O), and PM_2.5. The formaldehyde emission rate was calculated at each time step from the zone’s air temperature, relative humidity, and ventilation rate, based on a model fitted to measured field data in occupied single-family homes (Zhao et al., 2022). Formaldehyde emission rates in common corridors, stairwells, and elevator shafts were assumed to be zero. The outdoor formaldehyde concentration was fixed at 2 ppb.

PM_2.5 generation rates were based on the broad range of particle emissions associated with cooking and occupant activities. PM_2.5 equivalents were derived from the measured PM_2.5 values from a recent field study in newly constructed California homes (Chan et al., 2020). A Random Forest machine learning algorithm was used to differentiate the indoor concentrations as contributed from three sources: the outdoors, indoor cooking, and ‘other’ indoor particle sources. Indoor loss of PM_2.5 due to deposition and ventilation was estimated by fitting a regression model to the decaying indoor concentration after cooking. The resulting average emission rate was 0.0416 mg/s (329⋅10⁻⁶lb/h) for cooking events and 0.00007 mg/s (0.55⋅10⁻⁶lb/h) per occupant for other background emissions generated by occupants. A range hood capture efficiency of 50% was assumed, which reduced the cooking emissions to 0.0208 mg/s (165⋅10⁻⁶lb/h).

CO₂ emissions were taken from occupant respiration rates (Emmerich et al., 2005) with average CO₂ generation rates for adults of 10 mg/s (79⋅10⁻³lb/h) when awake and 6.5 mg/s (51⋅10⁻³lb/h) when asleep, and for children of 6.5 mg/s (51⋅10⁻³lb/h) when awake and 4 mg/s (31⋅10⁻³lb/h) when asleep. The occupants spent 8 h sleeping and 16 h awake each day on a fixed schedule. CO₂ emissions from cooking were set at 400 mg/s (3.17 lb/h) and the outdoor CO₂ concentration was fixed at 400 ppm.

Water moisture generation rates for each dwelling unit were aligned to estimates of moisture generation provided by (ANSI/ASHRAE Standard 160, 2016), which are similar to those reported in NISTIR-7212 (Emmerich et al., 2005) and NISTIR-6162 (Persily, 1998). The water vapor emissions were from: cooking (280 mg/s (2.2 lb/h)), showering (660 mg/s (52 lb/h)) and dishwashing (130 mg/s (1.0 lb/h)), as well as emissions from occupant respiration of 15 mg/s (0.1 lb/h) (awake)/9 mg/s (0.06 lb/h) (asleep) for adults and 10 mg/s (0.07 lb/h) (awake)/6 mg/s (0.04 lb/h) (asleep) for children. For both cooking and bathing we assumed that local exhaust ventilation captured half of the moisture released during the activity resulting in reduced emission rates of 140 mg/s (1.1 lb/h) for cooking and 330 mg/s (26 lb/h) for showering. Lastly, we estimated background moisture generation to be 20 mg/s throughout the house based on ASHRAE Standard 160. We used the EnergyPlus Effective Moisture Penetration Depth (EMPD) module to estimate moisture transport and storage in the simulations.

We simulated two types of indoor contaminants: (1) global contaminants, and (2) shadow contaminants. Both contaminant types assumed the same emission rates and schedules described above. Global contaminants were emitted in every dwelling unit and were meant to characterise baseline indoor contaminant concentrations from typical occupant activities. Global contaminants were not unit-specific, that is, we did not track the origin or movement of the global contaminants. Shadow contaminants were emitted in Unit 2 on three levels in each building (levels 1, 3 and 4 in the mid-rise buildings and Levels 1, 11 and 20 in the high-rise buildings). These contaminants were labelled with a unique identifier so that they could be tracked independent of the global contaminants and were used to characterise the impact of compartmentalization on inter-unit contaminant transport.

3. Results

The following section presents the results of the simulations and the relative impacts of compartmentalization, ventilation system type, and climate on building performance. Section 3.1 focuses on dwelling unit-level air flows, including air flow sources and magnitudes. Section 3.2 summarises the contaminant concentration results, considering both global contaminants (those emitted in all dwelling units) and shadow contaminants (those emitted in specific source zones). Finally, Section 3.3 summarises HVAC energy use consumption trends.

3.1. Airflows

In a CONTAM simulation, the air flow in or out of a zone at a given time step is determined by the pressure differential across the zone boundary and the leakage characteristics of that boundary. For each unit, dwelling unit airflows have been tracked based on the source of the air: from unit (red line) represents air flow through unit-to-unit walls and floor/ceiling assemblies; from outside (blue line) represents air flow through the building envelope, and from corridor (yellow line) represents air flow through the unit-to-corridor wall. The ‘from unit’ flows only account for air flow directly from an adjacent unit. This value does not account for air flow from non-adjacent units, either via other units or via common areas. The figure illustrates how the airflows vary with time: there are changes depending on weather (stack and wind pressures) and mechanical ventilation operation - the regular increases in flow from outside are due to intermittent local exhaust fan operation.

The relative contribution from each air source varies systematically, depending on unit location within the building. ‘From unit’ flows are consistent throughout the three-day period, at approximately 40 L/s (85 cfm) for units on Levels 2—4. No ‘from unit’ flows were noted for the Level 1 unit, indicating that ‘from unit’ flows are dominated by vertical unit-to-unit air flow, with little-to-no horizontal air flow, even with very leaky partitions. There is a clear height dependency for the ‘from corridor’ flows: the average ‘from corridor’ flow rates are lower in the Level 1 unit and increase with each level. This is consistent with previous studies that have measured corridor-to-unit differential pressures induced by corridor supply ventilation systems (Air Solutions Inc., 2005; Lozinsky et al., 2023; Maxwell et al., 2014; Ricketts, 2014). Corridor-to-unit differential pressures and, by extension, air flow rates, are consistently lower for units below the neutral pressure plane (NPP), compared to units at or above the NPP. Below the NPP, the corridor-to-unit differential pressure induced by the corridor supply can be insufficient to overcome the stack-induced pressures driving infiltration, resulting in low (or negative) corridor-to-unit differential pressures and air flows. ‘From outside’ flows exhibit an inverse relationship, with ‘from outside’ flows highest in the Level 1 unit, and lowest for the Level 4 unit. Assuming no other forces driving air flow, under winter conditions, stack pressures will drive infiltration (air entering the building from outside) below the NPP and exfiltration (air exiting the building to outside) above the NPP. Except for periods of unit exhaust fan operation, ‘from outside’ flows are less temporally variable than the ‘from corridor’ flows. Unit exhaust operation increases ‘from outside’ flow rates by 5—50 L/s (11—106 cfm), depending on the unit level and the exhaust fan flow rate.

The impact of compartmentalization on unit-level air flows is demonstrated in Figure 2, which illustrates these same ‘from’ air flows for Unit 2 on Level 4 in the mid-rise common corridor prototype, for both the ‘Typical’ and ‘Super Tight’ leakage cases. The difference in results between the two tightness levels shows how tighter construction makes air flows much more consistent, resulting in effectively zero flow from other units. Air flows from all three sources (unit, outside, and corridor) in the ‘Super Tight’ case remain constant throughout the three-day simulation period, except during periods of exhaust fan operation, which cause a simultaneous increase in ‘from outdoor’ flows and decrease in ‘from corridor’ and ‘from unit’ flows.

Figure 2. Building-level annual average unit air flow rates, by source of incoming air, separated by leakage (L/s₅₀/m²) and ventilation system type.

Dwelling unit IAQ is a function of not only the source(s) of unit air flows, but also their relative magnitudes. Figure 3 shows the building-level annual average of dwelling unit ‘from’ air flows, disaggregated by ventilation type and dwelling unit leakage level. Dwelling unit total air flows have been broken down based on the source of the air flow: from adjacent units (red), from the corridor (blue), and from the exterior (green). The mean total flow rates are lowest for the ‘No Unit Ventilation’ case and are similar across the four unit-level mechanical ventilation cases. ‘From unit’ flows representing cross-contamination between directly adjacent dwelling units is remarkably consistent across all ventilation types and appears to vary predominantly with leakage. The Unit Exhaust, Corridor Supply system has much more air from the corridor than the other three systems that all had very similar contributions from other units, the corridor and outside. This is an intentional outcome of the ventilation system design. Higher leakage rates resulted in higher total air flow, with all three air sources increasing with leakage.

Figure 3. Annual average ‘from unit’ air flows in the Highrise Common Corridor prototype, assuming Unit Exhaust, Corridor Supply ventilation (CZ 7, Typical leakage (5.1 L₅₀/s/m² (1.0 cfm₅₀/ft²))).

With an understanding of all the flow sources in hand, we now focus on air entering a unit directly from other adjacent units. Figure 4 shows an example of the annual average ‘from unit’ flows for each individual unit in the high-rise common corridor prototype in CZ 7, for a ‘Typical’ leakage level and Unit Exhaust, Corridor Supply ventilation. The results for individual units are consistent with the ‘from unit’ flows for the mid-rise common corridor prototype, shown in Figure 5. Level 1 ‘from unit’ flows were negligible, indicating minimal horizontal unit-to-unit transfer. Annual average ‘from unit’ flows were relatively consistent on Levels 2—20, ranging from 17 to 21 L/s (36 − 44 cfm). ‘From unit’ flows to the corridor zones were largest near the bottom of the building and decreased with each floor, again consistent with Figure 5. Unit Exhaust, Corridor Supply ventilation systems rely on corridors being positively pressurised relative to the units in order to deliver make-up air to the units and to prevent unit-to-unit air flow and contaminant transport via the corridors. Positive ‘from unit’ flows into the corridor zones indicate that the Corridor Supply ventilation system is not functioning as intended in this leaky building. Wind and stack pressures and the resulting air flows for this leakage case are strong enough to overcome the intended design of the mechanical ventilation system. Notably, this example was chosen to highlight these issues. It represents the leakiest and tallest building sited in the coldest climate region, where stack pressures are greatest. Previous field and simulation studies have shown similar results where corridor supply systems did not maintain consistent positive corridor pressurization, facilitating unit-to-unit air flow and contaminant transport (Air Solutions Inc., 2005; Lozinsky et al., 2023; Maxwell et al., 2014; Ricketts, 2014). ‘From unit’ flows, both between units and between units and corridors, generally decreased with increased compartmentalization and milder climate zones.

Figure 4. Unit-level air flows for Unit 2 on each level of the Midrise Common Corridor prototype, assuming Unit Exhaust, Corridor Supply ventilation, for three days in January (CZ 7, Typical leakage (5.1 L₅₀/s/m² (1.0 cfm₅₀/ft²))).

Figure 5. Comparison of ‘from unit’, ‘from outside’, ‘from corridor’, and total ‘from’ flows for Unit 2 on Level 4 in the Midrise Common Corridor prototype, assuming Unit Exhaust, Corridor Supply ventilation, for three days in January (CZ 7, Typical (5.1 L₅₀/s/m² (1.0cfm₅₀/ft²)) and Super Tight (0.25 L₅₀/s/m² (0.05 cfm₅₀/ft²)) leakage).

In order to ensure adequate ventilation provision and IAQ in all units in a building, we examine the worst-case, represented in this analysis using the maximum annual average unit-to-unit air flow (‘from unit’ air flow) observed for a given unit from each simulation case. From the example building shown in Figure 4, the worst-case unit was unit seven on level three (21.01 L/s). For maximum ‘from unit’ flows, the worst-case unit is not very different from the average unit in the building, because of the uniformity in the ‘from unit’ flows seen throughout (see Figure 4). This is not the case for other metrics explored later in this paper (e.g. shadow contaminant concentrations). Figure 6 shows the distributions of these maximum unit-to-unit flows, aggregated by leakage, climate zone, ventilation type, and building prototype categories. Maximum ‘from unit’ flows are consistent between the ventilation types, though the median for the Unit Supply cases is marginally higher than the other ventilation cases. This is consistent with the findings from a recent simulation study by Modera et al., which found that there was no significant difference in unit air flows between Supply and Exhaust systems (Modera et al., 2023). Maximum ‘From unit’ flows are most sensitive to climate zone and air leakage, with the median maximum ‘from unit’ flow for the ‘Typical’ leakage level (5.1 L₅₀/s/m² (1.0 cfm₅₀/ft²) approximately 12 times that for the ‘Super Tight’ (0.30 L₅₀/s/m² (0.05 cfm₅₀/ft²) leakage level. The importance of leakage and climate zone was corroborated through an analysis of variance (ANOVA) test, that indicated statistically significant differences between leakage and climate zone categories. Building prototype appeared to have a marginal (though statistically significant) effect on ‘from unit’ flows, with the median maximum ‘from unit’ flows varying by less than 1 L/s (2 cfm) between the three prototypes.

Figure 6. Distributions of building-level maximum ‘from unit’ flow rates from adjacent units, across simulation parameters.

In order to characterise any interactions between the climate zone and leakage effects on ‘from units’ flows, Figure 7 shows the maximum ‘from units’ flows averaged by leakage and climate zone combined. We observe that the effects of climate zone and leakage are indeed interactive, and that variability in maximum ‘from unit’ flows driven by climate factors is minimised in very tight buildings (<1 L/s (2 cfm)) and increases substantially with greater leakage (ranging from 9 L/s to 20 L/s (19 − 42 cfm)). The variability by climate for a fixed leakage class is typically smaller than the variability by leakage for a fixed climate region, suggesting that leakage is the more important factor here. For example, in the Moderate 1.0 L₅₀/s/m² (0.20 cfm₅₀/ft²) leakage class, the range over all climate zones is from 1.6 to 3.9 L/s. In contrast, in the CZ 3 cases, the range over all leakage classes is from 0.7 to 11.6 L/s. Notably, if we exclude the leakiest class of buildings, the impacts of leakage and climate zone become roughly equal. In the leakiest class of buildings, these ‘from unit’ flows represent up to 25% of all the air flow entering the unit from all sources (outside, corridor and other units), whereas in the most airtight cases, the ‘from unit’ flows represent only roughly 5% of the total air flow.

Figure 7. Mean worst-case ‘from unit’ flows, separated by leakage and climate zone, for all simulation cases.

As noted in Section 1, ASHRAE 62.2 specifies a maximum unit-level air leakage rate of 1.0 L₅₀/s/m² (0.2 cfm₅₀/ft²) (ANSI/ASHRAE, 2022). Tightening below the current ASHRAE 62.2 compartmentalization target provides diminishing returns for controlling inter-unit air flows, particularly for the warm climate zones. Even for CZ7, the mean worst-case From-Unit flows are all below 5 L/s.

3.2. Contaminant concentrations

Table 3 summarises the annual average global contaminant concentrations and their variability for each simulation parameter. For example, the ‘Unit Exhaust, Corridor Supply’ row is the average for that ventilation system type across all tightness levels, climates and building typologies. The variability was calculated by squaring the unit concentrations at each time step and recording the sum of these squares. The sum of each concentration for the simulation was recorded and squared. This allowed the calculation of variance. Taking the square root of the variance resulted in a standard deviation for each contaminant for each month. Annual variance was computed as the average of the 12 monthly values.

Table 3. Annual average global contaminant concentrations and standard deviations, by parameter.

Parameter	Annual Average Concentrations			Concentration Root Mean Square Errors (RMSE)			Concentration Standard Deviation Across Simulations
	CH2O (μg/m^3)	CO2 (ppm)	PM2.5 (μg/m^3)	CH2O (μg/m^3)	CO2 (ppm)	PM2.5 (μg/m^3)	CH2O (μg/m^3)	CO2 (ppm)	PM2.5 (μg/m^3)
Unit Exhaust, Corridor Supply	22.0	1009	15.5	4.2	193.2	20.8	0.9	8.4	2.6
Unit Balanced HRV	21.6	983	15.0	4.2	188.6	20.4	0.9	8.5	2.7
Unit Supply	24.5	1104	16.4	4.2	221.6	22.2	0.9	9.2	3.0
Unit Exhaust, Trickle Vent	23.3	1046	16.0	4.5	195.8	20.9	1.0	6.9	2.7
5.1	20.3	871	13.6	4.9	191.6	19.5	1.2	17.2	2.4
1.5	23.1	1039	15.7	4.3	200.2	21.1	0.8	14.0	2.7
1.0	23.5	1068	16.1	4.2	201.1	21.4	0.8	14.2	2.8
0.5	23.7	1094	16.5	4.1	202.5	21.6	0.7	14.1	2.8
0.3	23.8	1107	16.7	4.0	203.7	21.7	0.7	13.8	2.8
Midrise Common Corridor	23.2	1051	15.8	4.2	200.9	21.2	0.9	14.0	2.8
Midrise Walk Up	22.7	1032	15.9	4.4	199.3	21.0	0.9	17.7	2.9
Highrise Common Corridor	22.8	1025	15.4	4.2	199.2	20.9	0.9	13.9	2.8
2A	34.0	1038	14.9	5.2	198.7	20.1	0.9	13.6	3.5
2B	16.0	1096	15.7	4.7	209.0	21.1	0.2	13.3	3.2
3C	28.4	1049	16.1	4.2	199.5	21.4	0.4	13.5	2.5
4A	20.5	985	15.7	4.4	194.3	21.1	0.4	15.7	2.3
7	15.6	1009	16.0	2.8	197.6	21.6	0.2	16.1	2.2

We assessed the impacts of each simulation parameter using the percent difference between the highest average concentration and the lowest average concentration for each contaminant and parameter. Variability was by far lowest according to building prototype (2-3%), followed by ventilation type (9-12%), climate zone (8-10%, except formaldehyde at 54%) and leakage (15-21%). Formaldehyde results based on climate zone are notable for being highly variable (54%). This resulted from formaldehyde emission rates that were a function of humidity and temperature. Locations with hotter and more humid conditions (CZ2 and CZ3C) had higher emissions and average concentrations (28-34 ug/m³), whereas cold and dry locations (CZ7) had the lowest emissions and concentrations (16 ug/m³). All three contaminants showed lower average concentrations as leakage increased, due to increased rates of outside air exchange. This distinction was strong only for the leakiest buildings, whereas average concentrations were much less variable (3-6%) amongst all of the lower leakage rates (including 0.3, 0.5, 1.0 and 1.5 L/s₅₀/m²).

The No Unit Ventilation cases were excluded from Table 3 and the discussion of variability above, because they showed much higher average concentrations. When compared with the average of the mean values for ventilated cases, the No Unit Ventilation cases had average Formaldehyde, CO₂ and PM_2.5 concentrations that were 179%, 278% and 120% higher, respectively (40.9 μg/m³, 2,883 ppm, and 18.8 μg/m³, respectively). The average CO₂ concentration in the No Unit Ventilation cases were higher than current indoor air quality guidelines (ASHRAE, 2022; Health Canada, 2021). The average Formaldehyde concentrations for both No Unit Ventilation and Unit Ventilation cases were lower than the 8-h exposure limit listed in Health Canada’s Residential Indoor Air Quality Guidelines (50 μg/m³) (Health Canada, 2006), but were considerably higher than California’s chronic reference exposure limit of 9 μg/m³ (California Office of Environmental Health Hazard Assessment, 2014). These results clearly illustrate the value of mechanical ventilation systems in controlling indoor contaminant concentrations. The discussion of variability above suggests that the specific ventilation equipment used is much less important. While variability was fairly low, Unit Supply Ventilation equipment consistently showed the highest average concentrations, while the Unit Balanced HRV equipment consistently showed the lowest average concentrations.

The standard deviation results were fairly consistent over all cases, with the exception that the No Unit Ventilation cases had substantially more variability than the mechanically vented cases (by about 50% for Formaldehyde and about a factor of two in CO₂) and a definite trend that tighter construction had less variability and more consistency (fewer extremes).

Simulations were run with and without MERV-13 particle filtration, as described in Section 2.2.4. As expected, the use of MERV-13 filters resulted in lower mean PM_2.5 concentrations, compared to the no filtration cases (12.5 µg/m³ vs. 20.1 µg/m³). Because PM_2.5 transport between units was already low, due to losses from deposition and filtration by envelope and interior partition components, the additional impact of filtration on PM_2.5 transport was minor. More results for the distribution of mean zone global contaminant concentrations, across the simulation parameters may be found in the Supplemental Information. The impact of filtration on global PM_2.5 concentrations depended on climate zone, due to the different heating and cooling airflow and runtimes needed to meet the space conditioning loads in these scenarios. The greatest distinction was seen in cooling-dominated CZ 2 A (10.3 vs. 19.5 µg/m³). This difference shrank in the coldest climates (e.g. CZ7, 13.6 vs. 19.2 µg/m³). Increasing unit airtightness beyond the current ASHRAE 62.2 compartmentalization requirement increased annual average global zone concentrations by less than 5% (refer to Table 3 and Figures S1, S3–S5 in the Supplementary Information).

To further investigate unit-to-unit transport, propagation of contaminants released in an individual unit was investigated in this study using shadow contaminant releases. Shadow contaminants were emitted in Unit 2 on Levels 1, 3, and 4 in the mid-rise prototypes and Levels 1, 11, and 20 in the high-rise prototype, and were tagged with unique identifiers, allowing us to track their concentrations separate from the global contaminants emitted in all units (refer to Section 2.2.6 for more information). Figure 8 shows a heat map representation of an example building (CZ7, Leaky, Unit Exhaust, Corridor Supply) where the annual average fraction of middle shadow CO₂ is shown for each unit. This metric was calculated by dividing the annual average concentration of the middle shadow CO₂ by the annual average global CO₂ for each dwelling unit, illustrating the average proportion of dwelling unit contaminant that originated from a designated emission source (in this case, Unit 2, Level 11). The source zone is evident, with 38% of zonal CO₂ originating in the zone itself. The worst-case non-source zone in the building is the unit directly above the source zone (unit 2 on floor 12), where the fraction of CO₂ originating from the source unit below is 7%. Aside from the worst-case zone, further transmission to other adjacent zones rapidly diminishes to the point of being unobservable. These results suggest that for any given unit anywhere in the building, there are likely only a few other units making substantial (i.e. >1%) contributions to the zone concentration. These will commonly be the two or three zones directly beneath the zone of interest, along with the two zones on either side.

Figure 8. Heat map plot of the shadow middle CO₂ fraction in the Highrise Common Corridor prototype, assuming Unit Exhaust, Corridor Supply ventilation (CZ 7, Typical leakage (5.1 L₅₀/s/m² (1.0 cfm₅₀/ft²))).

Figure 9 illustrates distributions of bottom floor shadow CO₂ fractions for the worst-case unit in each building (excluding the shadow source unit). CO₂ fraction was calculated for each dwelling unit using the approach outlined for Figure 8, with the bottom floor shadow CO₂ used as the contaminant of interest (emitted in Unit 2, Level 1). The worst-case unit for each building had the highest annual average fraction of shadow CO₂, and in all buildings, this was the unit directly above the source unit. CO₂ fractions in worst-case units range from 0—20%. Like the air flow results, reducing leakage significantly reduces the fraction of CO₂ transported between units by about a factor of ten. Similarly, colder climate zones increase the unit-to-unit transport with more than a factor of two increase from the mildest to coldest climate zone. The differences between mechanical ventilation system types are very small (less than 1% of shadow CO₂) and all the system types showed low unit to unit transport of shadow CO₂. The results for formaldehyde and PM_2.5 sourced in other units show very similar results (see Supplemental Information for specific results). Transport of PM_2.5 was even lower than the other contaminants, due to other loss mechanisms, including deposition and filtration of air passing through interior envelope surfaces between zones. As with the Inter-Unit air flows, increasing compartmentalization beyond the ASHRAE 62.2 requirement had diminishing returns for controlling contaminant fractions. While the maximum non-source zone fraction for all three contaminants did decrease with increased compartmentalization, the reduction was marginal (Figures 9, S8 and S9 in the Supplemental Information).

Figure 9. Maximum annual average non-source CO₂ fraction (%) for each simulation case.

The fractional values shown in Figure 8 assume contaminant transport from a single source zone. These results represent the lower limit of inter-unit contaminant transport, as they do not account for cumulative effects from multiple source zones. Lozinsky simulated CO₂ transport using the same low-rise and high-rise typologies of this study, for a range of building airtightness configurations located in CZ5 (Lozinsky, 2023). Each contaminant in each dwelling unit was tagged with a unique unit identifier, which allowed for the differentiation between source zone and non-source zone contaminants, assuming multiple source zones. Assuming a single source zone (consistent with the shadow contaminant analysis in this paper, and the results shown in Figure 9), the maximum non-source zone CO₂ fractions in Lozinsky (2023) were a reasonable predictor for the building-level averages when all units acted as source zones. Assuming all units acted as source zones, building level shadow fractions for CO₂ ranged from 1—19%, which varied from the maximum non-source zone results from the single source zone analysis by no more than 4 percentage points. In contrast, the maximum non-source zone results from the single source zone analysis were poor predictors for the building-level maximums assuming multiple source zones—particularly for the ‘leaky’ cases. The building-level maximum fractions ranged from 2—31% and varied from the maximum non-source zone fractions by up to 22 percentage points.

Consistent with the air flow results discussed in Section 3.1 and illustrated in Figure 6, most contaminant transport occurs vertically between units, with some horizontal unit-to-unit and unit-to-corridor transport. Figure 10 shows a three-day time series plot of shadow CO₂ released in Unit 2 on Level 1, and the corresponding concentrations in the adjacent zones. While concentrations in the source zone reach a peak of approximately 700 ppm during the monitored period, the peak concentrations in the unit above do not exceed 200 ppm, while the Level 1 corridor and horizontal adjacent units don’t exceed 40 ppm. Compartmentalization increased contaminant concentrations in the source zone, but effectively eliminated inter-unit contaminant transport, as shown in Figure 11. This is consistent with previous simulation studies that have shown that compartmentalization increases global zone concentrations (by reducing dilution caused by air exchange with adjacent zones) but decreases zone concentrations of contaminants originating in other zones (Lozinsky, 2023; Mao et al., 2015; Modera et al., 2023).

Figure 10. Time-series plot of shadow CO₂ emitted in Unit 2 on Level 1 and resulting concentrations in adjacent zones for three days in January (Mid-Rise Common Corridor prototype, Unit Exhaust, Corridor Supply ventilation (CZ7, ‘Typical’ leakage (5.1 L₅₀/s/m² (1.0 cfm₅₀/ft²))).

Figure 11. Time-series plot of shadow CO₂ that compares ‘Typical’ (5.1 L₅₀/s/m² (1.0 cfm₅₀/ft²)) vs. ‘Super Tight’ (0.30 L₅₀/s/m² (0.05 cfm₅₀/ft²)) buildings. Shadow CO₂ was emitted from Unit 2 on Level 1. Resulting fractions are shown for the source zone and for Unit 2 on Level 2 for three days in January (Mid-Rise Common Corridor prototype, Unit Exhaust, Corridor Supply ventilation, CZ7).

Regardless of the source zone location, the maximum non-source zone concentration was always in the dwelling unit directly above the source zone; however, source zone location had a noticeable impact on contaminant transport potential. Assuming a source zone below the NPP, inter-zonal contaminant transport occurs via one of two routes: (1) directly between units, either vertically or horizontally or (2) indirectly, via the corridors and stairwell and elevator shafts. Even under the worst-case scenario the amount of contaminant transported indirectly from a single source zone via the common areas was negligible; the shadow CO₂ concentrations in the 20^th floor units of the high-rise prototype were less than 2% of the source zone concentration. For source zones at or above the NPP, most inter-zonal contaminant transport occurred via unit-to-unit transport. While there was some contaminant accumulation in the common area zones, this was not the primary transport route between units. Shadow contaminants released near the top of the building had minimal impact on adjacent units. Contaminant properties also affected transport potential. On average, the maximum non-source zone contaminant concentration was approximately 4.5% that of the source zone concentration for CO₂, 4% for formaldehyde, and 1% for PM_2.5 (assuming a source zone near the bottom or middle of the building). The lower contaminant fraction for PM_2.5 is due to deposition and penetration loss mechanisms, which reduced its inter-zonal transport potential. Modera et al. measured inter-unit CO₂ and PM transport in multi-family buildings (Modera et al., 2023). At approximately 0.80 L₅₀/s/m² (0.16 cfm₅₀/ft²), direct unit-to-unit transport of CO₂ was very small (between 0—3%) for both Balanced and Exhaust ventilation systems, while PM transport was below the LOD (< 1%). These finding align well with the results from our simulations. Note that these maxima are for non-mechanically ventilated cases and with mechanical ventilation and compartmentalization the concentrations are much lower (by factors of 5-10).

In order to consider some uncertainty owing to variability in activities, we varied the schedules of local exhaust fans, cooking and bathing over the course of a day and whether they differed between dwellings. As mentioned above, we considered a small range of schedule changes, where the operation of fans might shift 30 min between dwellings. These simulations were performed for CZ 7 (i.e. the coldest) in both the Typical (5.1 L₅₀/s/m² (1.0 cfm₅₀/ft²)) and Super Tight (0.3 L₅₀/s/m² (0.05 cfm₅₀/ft²)) cases for the mid-rise common corridor building across six ventilation types. Across these parameters, the maximum annual average ‘from unit’ flows observed were similar between the fully aligned and not aligned schedules (median of 0.8 vs. 1.4 L/s for the Super Tight cases, and median of 19.6 vs. 19.8 L/s for the Typical cases). The maximum non-source zone for shadow CO₂ was slightly more variable, as non-simultaneous schedules increased the shadow CO₂ from 8 to 13 ppm. Both are extremely low values that are physically inconsequential. These results indicate that assuming simultaneous activities introduces little biases into the results.

Similarly, we assessed the sensitivity of our results to our assumption that all windows were closed in the simulated multi-family buildings. We performed additional simulations with all windows open annually on the top and bottom floors of the building. These simulations were performed for CZ 7 for the Typical (5.1 L₅₀/s/m² (1.0 cfm₅₀/ft²)) and Super Tight (0.3 L₅₀/s/m² (0.05 cfm₅₀/ft²)) cases, for the mid-rise common corridor building across six ventilation types. For individual units with open windows, the outdoor air exchange was increased, though these changes remained small in the Super Tight units (30.8 to 31.6 L/s) and were greater in the Typical units (47.2 to 56.9 L/s). The presence of open windows on the top and bottom floors tended to modestly increase the flow from other dwelling units in the building (from 0.8 to 1.9 L/s in the Super Tight buildings and from 19.6 to 22.7 L/s in the Typical buildings) and to modestly reduce global CO₂ concentrations in the dwelling units (from 1,079 to 1,059 ppm in the Super Tight buildings and from 822 to 773 ppm in the Typical buildings). Relatively speaking, the flow from other units more than doubled in the very tight cases, but the absolute flows remained very small. Consistent with these results, in the most airtight cases, we observed a relatively large but globally minor increase in the maximum non-source zone concentration of CO₂ emitted from apartment 2 in the 1^st level. This was a notable change, but the overall concentration of the shadow contaminant remained very low (increasing from only 0.7% to 2.3% of global CO₂ in the windows closed vs. top and bottom open cases, respectively).

3.3. Energy use

Whole building HVAC energy use was normalised by the floor area to obtain an Energy Use Intensity (EUI in kWh/m²-year). The distributions of the resulting EUI are shown according to each simulation parameter in Figure 12. Unlike the air flow and contaminants explored above, the whole building HVAC EUI were driven most strongly by climate zone, followed by ventilation system type, leakage and building prototype. Because other building load elements were the same for each case, the changes in EUI are due to ventilation-related energy use only. This explains why the non-mechanically ventilated cases have the lowest EUI - the reduced ventilation air flows result in lower energy use. The impact of non-ventilation-related energy use is illustrated by the greater EUI for the mid-rise walk up, which has proportionally greater surface area compared with the enclosed volume. Averaged over all climates, the impact of air tightening did not scale directly with leakage. The fivefold reduction from 5.1 L₅₀/s/m² (1.0 cfm₅₀/ft²) to 1.0 L₅₀/s/m² (0.2 cfm₅₀/ft²) reduced EUI by about 25%. The lowest tightness levels did not have a significant impact on EUI compared to the 1.0 L₅₀/s/m² (0.2 cfm₅₀/ft²) results, indicating that most of the energy benefit of tighter construction can be obtained at this leakage level with significantly diminishing returns for the two tightest cases.

Figure 12. Distributions of total HVAC energy use intensity (kWh/m²), across simulation parameters.

Disaggregating the EUI by leakage and climate (as shown in Figure 13) shows that there is little to no energy benefit from tighter construction outside of areas with substantial heating loads (CZ4 and 7, in this work). In colder climates there are substantial advantages. Unit Supply and Unit Exhaust, Trickle Vent system configurations had the lowest EUI, while Unit HRV and Unit Exhaust, Corridor Supply systems had higher EUI. The higher energy use attributed to HRV systems and to systems with corridor supply air result from two distinct features. First, are differences in fan power requirements and differences in total air flow. Second, is the loss of free cooling associated with outside air exchange in warmer climates. The HRV acts to pre-heat incoming cool air from outside and reduces the benefit of ventilation cooling, and this offsets the advantage of pre-heating air in cold weather. For example, in CZ2A the average EUIs for HRV cases was 13% higher than the average EUI for trickle vent cases, which have the same mechanical ventilation air flow rates as HRV, except for dwelling unit supply. In cooling-dominated climate zones, the modest savings in heating EUI provided by the heat recovery are overwhelmed by the increased fan energy and cooling loads required to pre-condition the dwelling unit supply air. In contrast, EUIs for HRV cases in CZ7 were, on average, 19% lower than the trickle vent cases, despite having higher dwelling unit air flow rates. These results have implications for mechanical equipment selection in multifamily buildings, where the optimum system is climate dependent.

Figure 13. Mean total HVAC energy use intensity (kWh/m²) aggregated by leakage and climate zone.

4. Conclusion

This study coupled CONTAM and EnergyPlus models to assess the energy, air flow, and contaminant transport impacts of compartmentalization in mid-rise and high-rise multi-family buildings, for a range of building airtightness, climate, and ventilation system conditions, with the goal of answering five key questions:

• Are current compartmentalization requirements in codes and standards adequate?

• Can we recommend a level of compartmentalization that minimises internal air flows to the point where there are diminishing returns for IAQ?

• At that level of compartmentalization, does it matter what ventilation system is used?

• Are different ventilation systems more or less sensitive to compartmentalization?

• What are the space conditioning energy impacts of compartmentalization?

This study found the current compartmentalization leakage requirement in ASHRAE 62.2 (1.0 L₅₀/s/m² (0.2 cfm₅₀/ft²)) already leads to very low air and contaminant transport between units in both mid-rise (4 story) and high-rise (20 story) buildings. Tightening units beyond this level had diminishing returns with respect to limiting inter-unit air flows and contaminant transport. Cross contamination increased for compartmentalization levels above the current ASHRAE 62.2 limit (i.e. more leakage in each unit), as well as with increasingly cold climate conditions, due to increased air flows driven by stack effect. Air and contaminant transport in the building was largely vertical, with some unit-to-unit transport via corridor and stairwell/elevator shaft connections; however, indirect unit-to-unit contaminant transport via common areas typically represented a very small proportion of total zone contaminant concentrations. Tighter construction allowed for better control of air flows and contaminants resulting in less extreme peak values for energy use and contaminant concentrations. Air flow and contaminant transport trends were largely consistent between the mid-rise and high-rise cases.

The type of mechanical ventilation system had only very marginal impacts on cross-contamination in the building. No Unit Ventilation cases resulted in excessively high unit-level contaminant concentrations, beyond current best practices recommendations, supporting the need for mechanical ventilation with airtight construction. Increased compartmentalization resulted in more consistent air flow patterns within the building, indicating that compartmentalization helps mechanical ventilation systems to operate closer to their design intent. From an energy perspective, Unit Balanced HRV systems were found to be most effective for cold climates and less effective for moderate climates where ventilation cooling is a significant energy saving strategy. Compartmentalization had a significant impact on space conditioning energy loads in colder climates, with space conditioning energy loads increasing with air leakage rate, but a negligible impact on space conditioning energy loads in moderate climates.

There are several limitations to this study that need to be acknowledged. Although contaminant transport was low from a health and general IAQ perspective, odors can be perceived at very low concentrations and our general results about satisfactory IAQ may be more difficult to attain when considering odors. Our results do not suggest that there is a leakage and ventilation option that could allow smoking, for example. Another limitation is that it was assumed that all systems were correctly sized and installed and that dwelling units, ventilation systems, leakage areas, and leakage distributions were the same for all units. In real construction, such uniformity cannot be guaranteed and may lead to less ideal results. Similarly, the simulations did not include extremes of air flows, such as very powerful high flow kitchen exhausts, which may result in greater inter-unit air flows.

Acknowledgements

The authors would like to thank our colleagues Marianne Touchie (University of Toronto), Irene Poza (University of Valladolid) and Spencer Dutton (LBNL) for their guidance and support.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the Assistant Secretary for Energy Efficiency and Renewable Energy, Building Technologies Office, of the U.S. Department of Energy under Contract No. DE-AC02- 05CH11231 and the Natural Sciences and Engineering Research Council of Canada under PGSD3-547141-2020.

Notes on contributors

Iain S. Walker

Iain S. Walker, Mechanical Staff Scientist/Engineer and Leader of the Residential Building Systems Group at Lawrence Berkeley National Laboratory. His current work focuses on residential ventilation and home decarbonization.

Brennan D. Less

Brennan D. Less, Technology Researcher at Lawrence Berkeley National Laboratory. Research interests include state-of-the-art field measurements and simulations of smart ventilation systems, hygrothermal performance of sealed and insulated attics, indoor air quality, deep energy retrofits, zero net energy homes, and the health-energy efficiency nexus

Cara H. Lozinsky

Cara H. Lozinsky, PhD graduate from the University of Toronto and current Post-Doctoral Researcher at Lawrence Berkeley National Laboratory. Her research interests include building airtightness, mechanical ventilation system efficiency, and indoor air quality in residential buildings.

David Lorenzetti

David Lorenzetti, Staff Software Developer at Lawrence Berkeley National Laboratory. His work focuses on methods for solving whole-building airflow and pollutant transport problems, and applications of building airflow models to issues of life-safety in buildings.

Nuria Casquero-Modrego

Nuria Casquero-Modrego, Post-Doctoral Researcher at Lawrence Berkeley National Laboratory. Her research focuses on the energy and health impacts of residential decarbonization.

Michael D. Sohn

Michael D. Sohn, Mechanical Staff Scientist/Engineer at Lawrence Berkeley National Laboratory. His contributions were in the contaminant transport modeling sections of this paper.