5D Library › Architectural Acoustics, Reverberation Time and the Environmental Psychology of Built Space · 中文

Chapter 6 Carbon Dioxide, Ventilation and Decision-Making

What kind of indicator is carbon dioxide?

Indoor carbon dioxide (CO2) comes mainly from human breath. The outdoor concentration is now about 420 ppm (the 400 ppm in the author’s framework was the level of about a decade ago), and a full, poorly ventilated meeting room easily rises above 1,000 ppm. Ventilation engineering has long used indoor CO2 as an indicator of whether ventilation is adequate: people exhale CO2 together with water vapour, odours and a range of volatile organic compounds (together called “bioeffluents”), so high CO2 means that these are accumulating too. Figures such as 1,000 ppm are reference values for ventilation design, not health limits for CO2 itself.

Note that CO2 (carbon dioxide) and CO (carbon monoxide) are completely different gases. Carbon monoxide is produced by incomplete combustion and can kill at low concentrations; it needs a dedicated carbon monoxide alarm. The CO2 calculations in this chapter have nothing to do with it.

Derivation: mass balance for a meeting room

Calculation

Assume well-mixed indoor air of volume VV, with NN people each exhaling a volume GG of CO2 per second, an outdoor air supply QQ and an outdoor concentration CoutC_{out}. The mass balance of CO2 is

VdCdt=NG−Q(C−Cout).V\frac{dC}{dt} = NG – Q\,(C – C_{out}).

Writing the air change rate as ACH=Q/V\mathrm{ACH} = Q/V (per hour), the solution is

C(t)=Css−(Css−C0)e−ACH⋅t,Css=Cout+NGQ,C(t) = C_{ss} – (C_{ss} – C_0)\,e^{-\mathrm{ACH}\cdot t},\qquad C_{ss} = C_{out} + \frac{NG}{Q},

where CssC_{ss} is the steady-state concentration and 1/ACH1/\mathrm{ACH} the time constant of the approach to it. Keeping the steady state at or below a target C*C^* requires an outdoor air supply of G/(C*−Cout)G/(C^* – C_{out}) per person.

CO2 output depends on body mass, sex, age and activity. Persily and de Jonge re-estimated the CO2 generation rates of building occupants from metabolic physiology in 2017; in their tables, seated adults at 1.2–1.4 met (office work, meetings) produce about 0.0035–0.0056 litres per second per person, and adults sitting quietly (1.0 met) about 0.0029–0.0039 L/s (Persily and Jonge 2017). The program below takes 0.005 L/s as an illustrative value, at the upper end of the range for seated office work and therefore on the conservative side, and follows CO2 after six adults enter a 4.5 × 4.0 × 2.7 m meeting room.

Calculation

# Well-mixed room: V dC/dt = N G - Q (C - Cout). Six adults in a 4.5 x 4.0 x 2.7 m
# meeting room; G = 0.005 L/s per person (Persily & de Jonge 2017 tabulate about
# 0.0035-0.0056 L/s for seated adults at 1.2-1.4 met); outdoor 420 ppm; start at outdoor
# level. Q = ACH x V. Last line: the same room with G = 0.004 L/s (quiet sitting; Persily
# & de Jonge give about 0.0029-0.0039 L/s per adult at 1.0 met).
import math

V, N, G, C_OUT = 4.5 * 4.0 * 2.7, 6, 0.005e-3, 420.0           # m3, people, m3/s, ppm
SRC = N * G * 3600 * 1e6                                         # ppm x m3 per hour

def conc(ach, t_h, c0=C_OUT, src=SRC):
    css = C_OUT + src / (ach * V)
    return css - (css - c0) * math.exp(-ach * t_h), css

def minutes_to(ach, target, src=SRC):
    css = conc(ach, 0, src=src)[1]
    if css <= target:
        return None
    return -math.log((css - target) / (css - C_OUT)) / ach * 60

print(f"room volume {V:.1f} m3; CO2 source {N} x {G * 1e3:.3f} L/s"
      f" = {N * G * 3.6e6:.0f} L/h")
print(" ACH | outdoor air/person | steady state | at 60 min | minutes to 1000 | to 800")
for ach in (0.5, 1, 2, 3, 4, 6):
    q_pp = ach * V / 3.6 / N                                     # L/s per person
    c60, css = conc(ach, 1.0)
    m1, m8 = minutes_to(ach, 1000), minutes_to(ach, 800)
    f = lambda m: "never" if m is None else f"{m:5.0f}"
    print(f" {ach:3.1f} |     {q_pp:5.1f} L/s      |   {css:5.0f} ppm  |   {c60:5.0f}   |"
          f"      {f(m1):>5s}      | {f(m8):>5s}")
for target in (1000, 800):
    print(f"outdoor air needed to stay below {target} ppm:"
          f" {G * 1e3 / ((target - C_OUT) * 1e-6):.1f} L/s per person")
src4 = N * 0.004e-3 * 3600 * 1e6
print("G = 0.004 L/s each: minutes to 1000 ppm at ACH 0.5, 1, 2, 3 = "
      + ", ".join(f"{minutes_to(a, 1000, src4):.0f}" for a in (0.5, 1, 2, 3)))
# room volume 48.6 m3; CO2 source 6 x 0.005 L/s = 108 L/h
#  ACH | outdoor air/person | steady state | at 60 min | minutes to 1000 | to 800
#  0.5 |       1.1 L/s      |    4864 ppm  |    2169   |         17      |    11
#  1.0 |       2.2 L/s      |    2642 ppm  |    1825   |         18      |    11
#  2.0 |       4.5 L/s      |    1531 ppm  |    1381   |         22      |    13
#  3.0 |       6.8 L/s      |    1161 ppm  |    1124   |         31      |    14
#  4.0 |       9.0 L/s      |     976 ppm  |     965   |      never      |    17
#  6.0 |      13.5 L/s      |     790 ppm  |     789   |      never      | never
# outdoor air needed to stay below 1000 ppm: 8.6 L/s per person
# outdoor air needed to stay below 800 ppm: 13.2 L/s per person
# G = 0.004 L/s each: minutes to 1000 ppm at ACH 0.5, 1, 2, 3 = 21, 24, 32, 77
CO2 concentration after six adults enter a 48.6 m³ meeting room (well-mixed model, 0.005 L/s per person, outdoor 420 ppm); the solid line marks 1,000 ppm, the dotted line 800 ppm.

At one air change per hour (about 2.2 L/s of outdoor air per person, a common order of magnitude for a room with doors and windows shut), the steady state is about 2,640 ppm and 1,000 ppm is passed 18 minutes into the meeting; at three per hour the steady state is about 1,160 ppm, passed after 31 minutes; at four per hour the steady state is about 976 ppm and 1,000 ppm is never reached, but 800 ppm is exceeded after 17 minutes; at six per hour (about 13.5 L/s per person) the steady state is about 790 ppm. In this model, keeping the steady state below 1,000 ppm needs about 8.6 L/s of outdoor air per person, and below 800 ppm about 13.2 L/s. In a poorly ventilated room, most of a one-hour meeting is spent above 1,000 ppm. The last line is a sensitivity check: if each person produces only 0.004 L/s (closer to quiet sitting), 1,000 ppm is passed after about 24, 32 and 77 minutes at one, two and three air changes per hour; near the critical air change rate the result is very sensitive to the generation rate, so a real room should be measured.

The Harvard COGfx study and the experiment before it

Evidence

Satish and colleagues, 2012. Twenty-two participants (mostly university students), in groups of four, took the “Strategic Management Simulation” (SMS) test for 2.5 hours at each of about 600, 1,000 and 2,500 ppm CO2 on the same day, with ventilation and temperature held constant; the higher concentrations were produced by injecting pure CO2, and participants did not know the condition. Compared with 600 ppm, six of nine measures fell significantly at 1,000 ppm (by about 11–23 %); at 2,500 ppm seven fell significantly, with “initiative” and “basic strategy” falling to 7 % and 6 % of their earlier values. The authors themselves wrote that the effects at 2,500 ppm were so large that they “almost defy credibility” and needed confirmation (Satish et al. 2012).

Evidence

Allen and colleagues, 2016 (the COGfx study). Twenty-four professionals (architects, designers, programmers, engineers, marketers and managers) spent six full working days in the same controlled office laboratory, with participants and scorers blind to each day’s condition. The conditions were: conventional (about 20 cubic feet per minute of outdoor air per person, with added sources of volatile organic compounds), “Green” (the same ventilation, low VOCs), “Green+” (double the ventilation), and two days on which pure CO2 was added to Green+ to reach about 945 and about 1,400 ppm. Compared with the conventional condition, average SMS scores were 61 % higher under Green and 101 % higher under Green+; compared with Green+, they were about 15 % lower at about 945 ppm CO2 and about 50 % lower at about 1,400 ppm, with seven of nine domains declining. By regression, each 400 ppm increase in CO2 was associated with average scores about 21 % lower. The authors list as limitations: only 24 people, one site, conditions run on consecutive days (possible carry-over), uncontrolled confounding that cannot be excluded; the study’s funding included a gift from an industrial company (Allen et al. 2016).

Note

Correction to the author’s framework. First, the CO2 conditions of the COGfx study were about 945 and 1,400 ppm, not 1,000 and 1,400 ppm, and the outdoor baseline is now about 420 ppm. Second, the often quoted “61 % and 101 % higher scores” compare the Green and Green+ conditions with the conventional one, and the difference comes from VOCs and ventilation, not from CO2 alone. Third, “large falls in crisis response, information use and strategy” are, in the CO2 conditions, part of seven of nine domains declining; the paper’s text gives no percentage for each domain, and this book gives none.

Later studies: pure CO2 or bioeffluents?

Evidence

Zhang and colleagues, 2017. Twenty-five participants were exposed to each of five conditions for 255 minutes: in three, outdoor air was plentiful, bioeffluents were removed and CO2 was at a reference level of 500 ppm or raised with pure CO2 to 1,000 or 3,000 ppm; in the other two, outdoor air was restricted so that bioeffluents and CO2 rose together to 1,000 or 3,000 ppm. Adding pure CO2 had no statistically significant effect on cognitive performance; bioeffluents at the 3,000 ppm level reduced the speed of addition, lengthened the response time in a redirection task and reduced the number of correct links in a cue-utilisation test (Zhang et al. 2017).

Evidence

Review. Du and colleagues in 2020 reviewed 37 studies of indoor CO2 and cognitive function, noting large differences between studies in exposure, tasks and populations and clear methodological gaps, and called for research aimed at setting safe exposure limits for settings such as schools (Du et al. 2020).

The two sets of experiments do not conflict quite as sharply as it seems: the CO2 conditions of Satish and of Allen also used injected pure CO2, but Zhang’s design manipulated CO2 and bioeffluents separately and did not reproduce an effect of pure CO2. The difficulty in real rooms is that CO2 and other bioeffluents share one source (people) and one exit (ventilation), so in field data they are almost impossible to separate. The program below builds 400 synthetic rooms in which performance depends only on bioeffluents, and asks what someone who measures only CO2 would conclude.

Calculation

# 400 synthetic rooms with random occupancy and ventilation: CO2 and a generic
# "bioeffluent" index share one source (people) and one sink (ventilation). By
# construction, performance depends ONLY on the bioeffluent index. What does someone who
# measures only CO2 conclude?
import math, random

rng = random.Random(2016)
rows = []
for _ in range(400):
    n = rng.randint(2, 12)
    q = rng.uniform(3, 20) * n                                  # outdoor air L/s
    gen_co2 = rng.gauss(0.005, 0.0006) * n * 1e3                # mL/s
    co2 = 420 + 1e3 * gen_co2 / q * rng.lognormvariate(0, 0.08)  # ppm (mixing error)
    bio = n * rng.lognormvariate(0, 0.15) / q * 100             # arbitrary units
    score = 100 - 1.5 * bio + rng.gauss(0, 4)                   # depends on bio only
    rows.append((co2, bio, score))

def corr(a, b):
    ma, mb = sum(a) / len(a), sum(b) / len(b)
    sab = sum((x - ma) * (y - mb) for x, y in zip(a, b))
    return sab / math.sqrt(sum((x - ma) ** 2 for x in a) * sum((y - mb) ** 2 for y in b))

co2, bio, sc = zip(*rows)
mean = lambda v: sum(v) / len(v)
cov = lambda a, b: sum((x - mean(a)) * (y - mean(b)) for x, y in zip(a, b)) / len(a)
slope = cov(co2, sc) / cov(co2, co2)
print(f"correlation CO2 vs bioeffluent index: {corr(co2, bio):.2f}")
print(f"correlation CO2 vs score: {corr(co2, sc):.2f} (CO2 has no causal role here)")
print(f"naive slope: {slope * 400:+.1f} points per +400 ppm CO2"
      f" ({slope * 400 / mean(sc):+.0%} of the mean score)")
# two-predictor least squares on CO2 and the bioeffluent index (2 x 2 normal equations)
a, b, cc = cov(co2, co2), cov(co2, bio), cov(bio, bio)
y1, y2 = cov(co2, sc), cov(bio, sc)
det = a * cc - b * b
b_co2, b_bio = (cc * y1 - b * y2) / det, (a * y2 - b * y1) / det
print(f"adjusting for the bioeffluent index: {b_co2 * 400:+.1f} points per +400 ppm CO2")
print(f"                                    bioeffluent slope {b_bio:+.2f} (true -1.50)")
print("but the bioeffluent index is almost never measured in the field; a controlled")
print("experiment that adds pure CO2 at fixed ventilation is needed to separate the two")
# correlation CO2 vs bioeffluent index: 0.92
# correlation CO2 vs score: -0.85 (CO2 has no causal role here)
# naive slope: -10.5 points per +400 ppm CO2 (-13% of the mean score)
# adjusting for the bioeffluent index: +0.3 points per +400 ppm CO2
#                                     bioeffluent slope -1.52 (true -1.50)
# but the bioeffluent index is almost never measured in the field; a controlled
# experiment that adds pure CO2 at fixed ventilation is needed to separate the two

In these rooms CO2 correlates 0.92 with the bioeffluent index and −0.85 with performance, and a simple regression on CO2 says that “each 400 ppm costs 10.5 points (13 % of the mean score)”, even though by construction CO2 has no effect on performance at all. Only when bioeffluents are measured and adjusted for does the CO2 coefficient return to near zero (+0.3 points) and the bioeffluent coefficient to near its true value (−1.52 against −1.50). Bioeffluents are almost never measured in the field, so the two can be separated only by experiments that manipulate them separately, as Zhang and colleagues did.

Note

Implications for practice. Whether or not CO2 itself has a direct effect, “high CO2” reliably signals that a room’s outdoor air per person is inadequate, so bioeffluents, moisture and other indoor pollutants are accumulating too. The checklist in Chapter 8 therefore still uses CO2 as an indicator of ventilation and takes 800 ppm, more conservative than common reference values, as the target for meeting rooms; this is a design choice, not a health threshold.

Hypothesis · not established

Below 1,500 ppm, declines in complex decision-making come mainly from bioeffluents and volatile organic compounds rather than from CO2 itself (not established). A testable consequence: in a 2 × 2 factorial experiment manipulating CO2 (pure CO2 injected, 600 versus 1,400 ppm) and bioeffluents (outdoor air supply) separately, decision-test scores should vary only with the bioeffluent condition. If pure CO2 at 1,400 ppm consistently lowers scores while bioeffluents are kept low, and this is replicated in several laboratories, the hypothesis is refuted.

Chapter summary

Indoor CO2 is an indicator of ventilation, and 1,000 ppm is not a health limit. The well-mixed mass balance VdC/dt=NG−Q(C−Cout)V\,dC/dt = NG – Q(C – C_{out}) shows that six people in a 48.6 m³ room at one air change per hour pass 1,000 ppm in 18 minutes, and that a steady state below 800 ppm needs about 13 L/s of outdoor air per person. Satish and colleagues (2012) and Allen and colleagues (2016) reported marked falls in decision-test scores after injecting pure CO2 (about 945–2,500 ppm), but with small samples and calls for replication from the authors themselves; when Zhang and colleagues (2017) manipulated them separately, pure CO2 up to 3,000 ppm had no effect while bioeffluents did. The illustrative calculation shows that in field data CO2 and bioeffluents can hardly be separated. The next chapter returns to the ancients: how they wrote about wind, water, shadows and houses.

References

Allen, Joseph G., Piers MacNaughton, Usha Satish, Suresh Santanam, Jose Vallarino, and John D. Spengler. 2016. “Associations of Cognitive Function Scores with Carbon Dioxide, Ventilation, and Volatile Organic Compound Exposures in Office Workers: A Controlled Exposure Study of Green and Conventional Office Environments.” Environmental Health Perspectives 124 (6): 805–12. https://doi.org/10.1289/ehp.1510037.
Du, Bowen, Marlie C. Tandoc, Michael L. Mack, and Jeffrey A. Siegel. 2020. “Indoor CO2_2 Concentrations and Cognitive Function: A Critical Review.” Indoor Air 30 (6): 1067–82. https://doi.org/10.1111/ina.12706.
Persily, Andrew, and Lilian de Jonge. 2017. “Carbon Dioxide Generation Rates for Building Occupants.” Indoor Air 27 (5): 868–79. https://doi.org/10.1111/ina.12383.
Satish, Usha, Mark J. Mendell, Krishnamurthy Shekhar, et al. 2012. “Is CO2_2 an Indoor Pollutant? Direct Effects of Low-to-Moderate CO2_2 Concentrations on Human Decision-Making Performance.” Environmental Health Perspectives 120 (12): 1671–77. https://doi.org/10.1289/ehp.1104789.
Zhang, Xiaojing, Pawel Wargocki, Zhiwei Lian, and Camilla Thyregod. 2017. “Effects of Exposure to Carbon Dioxide and Bioeffluents on Perceived Air Quality, Self-Assessed Acute Health Symptoms, and Cognitive Performance.” Indoor Air 27 (1): 47–64. https://doi.org/10.1111/ina.12284.

Fan, Y. (2026). Architectural Acoustics, Reverberation Time and the Environmental Psychology of Built Space: From Sabine's Equation and Fractal Visual Dimensions to Cognitive Performance, Read Beside Classical Chinese Spatial Design [建筑声学、混响时间与空间环境心理学的物理模型] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23782090 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).