Improving Indoor Air Quality with Green Walls: An Experimental Study

Indoor air quality (IAQ) is critical to health, especially in indoor environments such as schools and offices [1]. Since people spend up to 90% [2] of their time indoors, it is essential to monitor and improve IAQ to prevent adverse health effects [3, 4]. There is a well-documented correlation between air quality and thermohydrometric comfort, with the impact on both physical and psychological well-being [5]. Stagnant or odorous air impairs well-being [6], causing discomfort, headaches, and irritation, and reduces perceived comfort even when temperature and humidity are optimal [7]. Carbon dioxide (CO₂) is a frequently employed tracer gas in the assessment of air exchange and the accumulation of pollutants in indoor environments [8]. Controlled mechanical ventilation (CMV) remains the standard for IAQ control, but it is energy-intensive and often does not respond to changes in occupancy or pollutant loads [9]. In energy-efficient buildings, reduced natural ventilation increases the risk of indoor pollution accumulation [10]. The impacts on health and the economy are significant, with estimated social costs of up to €20 billion per year [11]. In addition, CMV systems face rising costs due to energy prices, complex maintenance, and the need for advanced technologies [12], which is driving the search for sustainable alternatives. To improve indoor environmental quality, integrating vertical vegetated structures—Green Walls (GWs)—offers a sustainable, nature-based solution. Their vertical layout maximizes vegetated area with minimal floor space and adds both pollutant reduction and psychological benefits [13]. GWs include green facades, which use climbing plants, and living walls (LWs), which consist of pre-grown plants on panels with artificial irrigation and nutrient systems [14]. LWs, though effective, are maintenance-intensive, especially indoors where light, rain, and ventilation are limited, requiring artificial lighting, drainage, and controlled irrigation. In this context, some studies have focused specifically on the ability of LW to remove CO₂ in controlled environments, evaluating the role of factors such as lighting, and plant type. Dominici et al. [15] showed that optimized lighting (200 $\mu \mathrm{mol} \cdot \mathrm{m}^{-2} \cdot \mathrm{~s}^{-1}$ at 15°) significantly enhances CO₂ uptake in passive green walls, achieving an equivalent air exchange rate of 0.21 $\mathrm{h}^{-1}$ in a 40 m³ office. Under typical indoor lighting (≤ 50 $\mu \mathrm{mol} \cdot \mathrm{m}^{-2} \cdot \mathrm{~s}^{-1}$), performance drops sharply to just 0.03 $\mathrm{h}^{-1}$. Danielski et al. [16] studied two Swedish classrooms equipped with 310 plants and found a 10% CO₂ reduction during school hours (~ 70 ppm), a 0.8℃ rise in indoor temperature with greater thermal stability, and up to 18% annual heating energy savings with reduced ventilation. The study also reported improved student well-being, including reduced stress.

Despite numerous studies, quantitative data usable for experimental validation or simulation remains limited. To address this gap, we conducted tests in a full-scale climate chamber equipped with two vertical LW, comparing CO₂ decay under two conditions: with and without artificial LED lighting (LED ON, and OFF). Concentration was measured over time with point sensors and fitted using a first-order exponential decay model. A Random Search Method (RSM) was used to estimate decay constants for both cases. Unlike previous small-scale or theoretical studies, this work provides high-resolution data in a thermally controlled environment, enabling quantification of both instantaneous and cumulative photosynthetic effects. The integration of the two fitted curves allows estimation of the net CO₂ removal due to lighting, showing a 30% reduction in half-life and an additional 6 g of CO₂ absorbed in 6 hours. These findings support the synergy between lighting and vegetation for IAQ control.

The present work is organized as follows: Section 2 Materials and Methods provides an overview of the climatic chamber used for testing and the measurement methodology used. The results of the tests carried out are given in Section 3. Finally, the conclusions and future developments of the work are presented in Section 4.

This section presents and discusses the results of the tests carried out in the present work. The experimental procedure was conducted under controlled temperature conditions of 24℃ (± 0.5℃).

3.1 Experimental decay curves

To assess the reproducibility of the experimental setup, three independent tests under LED OFF conditions were analyzed. As shown in Figure 5, the decay curves obtained in LED OFF mode are similar. The mean standard deviation $\sigma$ among the three tests was 9.28 ppm: this confirms the high degree of repeatability and consistency of the measurements across different trials. The maximum decay time was about 800 min, the minimum 740 min. Table 1 presents the CO₂ decay profiles recorded during the tests. The initial reading, taken at the commencement of the experiment, indicated an approximate concentration of 3020 ppm. Over the course of the 800-minute test period, there was a consistent decrease in concentration, reaching a final average of 502 ppm. Looking at Figure 5 and Table 1, the decay exhibited across all three replicates is characterized by a smooth and progressive decline, with a more rapid decrease during the initial 180 minutes, followed by a gradual stabilization. It is noteworthy that a high degree of consistency is observed between the three trials. At each point, the measured values are closely aligned, with only minor deviations.

Table 1. Evolution of CO₂ concentration (ppm) over time (min) under experimental conditions (24℃ ± 0.5℃) with LED illumination off (LED OFF)

5.jpg

Figure 5. Experimental decay curves for the condition LED ON/OFF (24℃ ± 0.5℃)

This phenomenon is reflected in the average column, which closely follows each individual dataset. The low spread observed across the three series is indicative of the high reproducibility of the experimental setup.

Figure 5 also presents the CO₂ concentration decay profiles under LED ON conditions for three independent tests. The curves are color-coded as follows: orange (Test 1), green (Test 2), and magenta (Test 3).

It is evident that all three curves demonstrate the characteristic exponential decay behavior.

Despite the general similarity in shape, the decay rates vary between the three tests. It is noteworthy that Test 3 (magenta) exhibited the most rapid reduction, achieving values below 600 ppm in less than 300 minutes. In contrast, Test 1 (orange) displays a slower decay, while Test 2 (green) presents an intermediate trend. These differences can be attributed to the increasing leaf area of the plants across the successive tests. As the plants grew, their photosynthetic capacity improved, leading to higher rates of CO₂ absorption. This phenomenon is especially pronounced in the steeper decline observed in Test 3. Table 2 provides numerical confirmation of these observations. At each point, the CO₂ concentrations recorded in Test 3 are lower than those in Tests 1 and 2. For instance, following 180 minutes of operation, the concentration in Test 3 is recorded as 657 ppm, in comparison to 1006 ppm and 893 ppm in Tests 1 and 2, respectively. The mean values derived from the three tests demonstrate this trend, exhibiting a progressive decline that corresponds to the physiological development of plants.

A comparison between the CO₂ decay profiles under LED OFF and LED ON conditions reveals substantial differences in the dynamics of concentration reduction. In all cases, the presence of illumination leads to a more rapid decrease in CO₂ levels, consistent with the activation of photosynthetic mechanisms.

The findings of the individual tests demonstrate that the rate of decay is significantly increased under the influence of light. For instance, following 180 minutes of operation, the CO₂ concentrations in the LED ON tests ranged from 657 to 1006 ppm, while the LED OFF tests exhibited higher concentrations, ranging from 1128 to 1150 ppm (Tables 1, 2). This discrepancy becomes increasingly evident over time. The divergence between the two conditions manifests at the outset and becomes more pronounced as time progresses. Following a period of 60 minutes, the mean concentration of the substance under the condition of LED ON was found to be 1788 ppm, whereas under the condition of LED OFF it remained at 2034 ppm, a discrepancy of 249 ppm (-12%). This discrepancy is found to widen progressively: after 180 minutes, the LED ON condition shows an average of 852.01 ppm, compared to 1139.24 ppm in the LED OFF scenario, resulting in a delta of 287.2 ppm (-25%).

As demonstrated in Tables 1 and 2, the difference in mean CO₂ concentration stabilizes at 300 minutes, with an average of 667.54 ppm when lighting is operational, and 776.08 ppm in the absence of lighting. This indicates a persistent difference of 108.5 ppm. This finding suggests that the most significant separation occurs within the initial 3 - 4 hours, coinciding with the period when the plants' photosynthetic activity is at its peak relative to the background decay. The implementation of artificial illumination not only speeds up CO₂ removal but also reduces the residual concentration level towards which the system tends to converge. This outcome is consistent with the anticipated effect of elevated CO₂ uptake by the plants, a consequence of the stimulation of photosynthetic activity by the LED lighting.

Table 2. Evolution of CO₂ concentration (ppm) over time (min) under experimental conditions (24℃ ± 0.5℃) with LED illumination on (LED ON)

3.2 Model fitting of CO₂ decay

This analysis aims to model the decay of CO₂ concentration over time within a confined environment under the two conditions (LED ON, and OFF).

The behavior of indoor CO₂ concentration in the absence of significant new emissions can be described by an exponential decay model [9]. The simplest form [10] of this model is:

$C(t)=C_0 \cdot e^{-k \cdot t}$    (1)

where, C(t) is concentration (ppm) at time t (min), $C_0$ is initial concentration (ppm) and k is decay coefficient (min^-1). However, the curves in Figure 5 show a minimum concentration limit value (approximately 500 ppm), so the model in Eq. (1) can be evaluated [10]:

$C(t)=C_{\infty}+\left(C_0-C_{\infty}\right) \cdot e^{-k \cdot t}$    (2)

where, $C_{\infty}$ is the asymptotic concentration.

The analysis was conducted using the Random Search Method (RSM) [18], which allows parameter estimation and minimizing the average absolute relative deviation (AARD) between the experimental values and the model. The initial value was kept fixed at the mean value at time $t_0$.

AARD(C(t)) [19] is expressed as:

${AARD}(C(t))=\frac{100}{N} \sum_{i=1}^N\left|\frac{C(t)_{{calc }}-C(t)_{exp }}{C(t)_{exp }}\right|$    (3)

where, N is the number of experimental points, $C(t)_{{calc }}$ and $C(t)_{exp }$ are the calculated and the experimental concentration value respectively.

Fitting was conducted separately for each condition imposing lower and upper bounds to ensure numerical stability.

The following equations are derived from the analysis:

$C_{O N}(t)=474.6+2534.8 \cdot e^{-0.011 \cdot t}$    (4)

$C_{O F F}(t)=511.2+2509.5 \cdot e^{-0.007 \cdot t}$    (5)

The equations demonstrate an AARD (%) of 1.2% for Eq. (4) and 2.4% for Eq. (5). It is evident that both models demonstrate a high degree of accuracy. The estimated values of the decay constant k highlight a clear difference between the two experimental conditions. Under LED ON, the decay rate was $k=0.011 \mathrm{~min}^{-1}$, while under LED OFF it was $k=0.007 \mathrm{~min}^{-1}$. This corresponds to a relative increase of approximately 45% in the presence of illumination, indicating that CO₂ is removed more rapidly when photosynthetic activity is stimulated by light.

To better interpret this result, the half-life time $\tau$, defined as the time required for the CO₂ concentration to reduce to half of its initial excess, is computed as [20]:

$\tau=\frac{\ln (2)}{k}$    (6)

Accordingly, the half-life under LED OFF conditions increases to 90 minutes, while under LED ON conditions is approximately 63 minutes (-30%). This substantial reduction in $t_{1 / 2}$ confirms that the presence of artificial lighting significantly accelerates the decay of CO₂ concentrations. To facilitate comparison with mechanical ventilation, the decay constants were converted into equivalent air change rates (ACHₑ) using $A C H_e=k \cdot 60$, assuming a well-mixed volume [10]. The resulting values are 0.46 h^-1 (LED OFF) and 0.66 h^-1 (LED ON), indicating a marked increase (43%) in effective air renewal due to photosynthetic activity. The system responds more efficiently to CO₂ accumulation, leading to faster improvements in IAQ when illumination is provided. As illustrated in Figure 6, the decay curves are depicted as a function of time, under LED ON, and OFF conditions. The solid curve denotes the mean values derived from the experimental data points, while the dashed curve signifies the calculated curves. To quantitatively assess the cumulative impact of artificial lighting on CO₂ removal, the area between the two model curves was computed.

This area, shown in gray in Figure 6, represents the integral overtime difference between the two scenarios:

$A=\int_0^{360}\left(C_{O F F}(t)-C_{O N}(t)\right) d t$    (7)

The integration was performed over the interval from 0 to 360 minutes and was computed numerically using the trapezoidal rule [21], which approximates the area under a curve by summing up the areas of trapezoids formed between adjacent points of the fitted model curves. The result is approximately 86’515 ppm·min, which quantifies the excess CO₂ that accumulates in the absence of illumination: It reflects the cumulative exposure to higher CO₂ levels under LED OFF conditions. Although not directly a mass, this integral is proportional to the total amount of CO₂ retained, and—when multiplied by the room volume and an appropriate conversion factor—allows for estimation of the effective mass removed. In this case, the difference corresponds to an estimated 6.06 grams of CO₂ removed more efficiently when lights were on.

6.jpg

Figure 6. Comparison between experimental and fitted CO₂ decay curves under LED ON and LED OFF conditions. The shaded area highlights the integrated concentration difference between the two scenarios, calculated over the first 360 minutes

We would like to thank the company Pellegrini Giardini, and in particular Dr. A. Pellegrini and P. Curini, for providing the species of the living wall along with some of experimental systems used for testing, and for their valuable support during the whole experimental campaign. The authors would also like to thank F. Tartaglini for his support during the mounting of the experimental set-up.