Opportunities Arising from COVID-19 Risk Management to Improve Ultrafine Particles Exposure: Case Study in a University Setting

Abstract

Particulate matter (PM) is recognized as a leading health risk factor worldwide, causing adverse effects for people in living and working environments. During the COVID-19 pandemic, it was shown that ultrafine particles (UFP) and PM concentrations, may have played an important role in the transmission of SARS-CoV-2. This study aims to investigate whether the mechanical ventilation system installed as a COVID-19 mitigation measure in a university dining hall can be effectively and sustainably used to improve indoor UFP exposure levels, integrated with a continuous low-cost sensor monitoring system. Measurements of particle number concentration (PNC), average diameter (D_avg), and Lung Deposited Surface Area (LDSA) were performed over three working days divided into ten homogeneous daily time slots (from 12:00 am to 11:59 pm) using high-frequency (1 Hz) real-time devices. PM and other indoor pollutants (CO₂ and TVOC) were monitored using low-cost handheld sensors. Indoor PNC (Dp < 700 nm) increased and showed great variability related to dining activities, reaching a maximum average PNC level of 30,000 part/cm³ (st. dev. 16,900). D_avg (Dp < 300 nm) increased during lunch and dinner times, from 22 nm at night to 48 nm during post-dinner recovery activities. Plasma-based filter technology reduced average PNC (Dp < 700 nm) by up to three times, effectively mitigating UFP concentrations in indoor environments, especially during dining hall access periods. It could be successfully adopted also after the pandemic emergency, as a sustainable health and safety control measure to improve UFPs exposure levels.

1. Introduction

Particulate matter (PM) has been recognized as one of the leading environmental health risk factors worldwide, contributing to adverse health effects in life and work environments [1,2]. The negative impact of PM exposure was significant in terms of costs and quality-adjusted life years and it accelerated as the exposure time increased, emphasizing the importance of early interventions [3].

Even if it is known that exposure to atmospheric PM causes severe health problems in humans and significant damage to their respiratory and cardiovascular systems [4], research published is not yet exhaustive regarding the role of air pollution on the geographic spread of the COVID-19 disease both regionally and globally [5]. During the pandemic, it was shown that environmental factors, including PM, may play a significant role in the differential distribution and transmission of SARS-CoV-2 [6,7]. Other evidence from the literature highlighted the important contribution of chronic exposure to PM on the spread and lethality of COVID-19. In particular, PM_2.5 and nitrogen dioxide (NO₂) appear to be more closely correlated with COVID-19 than PM₁₀ [8]. The review by Zhu et al. [9] provided insights into the detrimental effects of PM on various human health problems including respiratory, circulatory, nervous, and immune systems along with their possible mechanisms of toxicity. Furthermore, the potential effects of short- and long-term exposure to atmospheric pollution on COVID-19 risk and fatality rates were well described in the analysis of the first epidemic wave in Northern Italy [10] and in the Catalan Tarragona Province in Spain [11]. In the United States, Zhou et al. [12] found strong evidence that wildfires amplified the effect of short-term exposure to PM_2.5 on COVID-19 cases and deaths, although there was substantial heterogeneity across counties. The study conducted in the Lombardia region (Northern Italy) concluded that the interaction between atmospheric PM and NO₂ led to a significant increase in health events linked to the pandemic, with an estimated number of hundreds of infections, additional hospitalizations and intensive care admissions compared to non-urban areas [13].

Other studies reported the important role of indoor air quality in SARS-CoV-2 transmission, highlighting that virus particles may have similar dynamics to other suspended ultrafine particles (UFP, e.g., arising from road traffic, heating or other sources) [14,15,16]. Furthermore, the association between indoor carbon dioxide (CO₂) levels and virus spreading has been explored also before the COVID-19 pandemic [17]. In a recent study, Peng and Jimenez [18] described CO₂ levels corresponding to a given absolute SARS-CoV-2 infection risk varying by two orders of magnitude for different environments and activities. Other studies report a defined threshold of potential risk of infected droplets/aerosol inhalation fixed at 700 ppm of CO₂ in indoor environments [5]. Finally, the improvement of indoor air quality by using ventilation systems has been highlighted as a COVID-19 risk mitigation method [19] as well as social distancing, face masks and hand hygiene and cleaning/disinfection of surfaces [20,21,22,23]. Furthermore, indoor air quality monitoring gained high relevance also to better know the mechanisms of interaction with COVID-19 spreading [24].

Air quality monitoring is conducted using sensors and instrumentation, which face many challenges. Accurate detection of indoor air pollutants and PM requires the selection of appropriate sensors and systems. The utilization of a high number of key performance indicators may increase the reliability of air quality assessment outcomes and associated mitigation strategies [25]. The indoor environment is heterogeneous, with significant variability within space and between different microenvironments and locations [26]. Sensor placement, occupancy, and activity reports, as well as measurements in different microenvironments and locations, can contribute to understanding this variability. Outdoor pollutants can enter the indoor space via the building envelope; however, the measurement of external pollution and environmental conditions and recording details on the building fabric and ventilation conditions can help the apportionment of external contributions [27].

In this framework, the present study would answer the research question of whether mechanical ventilation systems installed as a COVID-19 mitigation measure can be effectively and sustainably used after the pandemic to improve UFP exposure levels, integrated with a continuous low-cost sensor monitoring system.

In this view, the study investigates the role of a mechanical ventilation system equipped with plasma-based filter technology installed inside the dining hall of a university site. An experimental campaign was conducted in May 2021 to mainly assess the impact of such a filtration system on UFP concentration levels, by high frequency (1 Hz) real-time measurements related to the activities carried out in the canteen. PM and other indoor pollutants, such as total volatile organic compounds (TVOC) and CO₂, have been simultaneously monitored by low-cost handheld sensors to evaluate overall indoor air quality for the users (workers and students). Finally, considerations about the affordability of low-cost sensors used to monitor indoor PM₁ by comparison with high-frequency real-time measurement data of airborne particle concentration are made to evaluate the sustainability of their future applications as continuous UFP and indoor air quality monitoring systems.

2. Materials and Methods

2.1. State of COVID-19 Pandemic Spread at the Time of the Experimental Campaign

Scientific evidence about the transmission ways of the SARS-CoV-2 virus constantly evolved during the pandemic emergency, as shown in the results of studies conducted at the time of the present investigation campaign (first half of May 2021). In 2021, three different pandemic waves were identified in Italy due to the subsequent co-circulation of SARS-CoV-2 variants Beta (B.1.351), Gamma (P.1), Delta (B.1.617) and Omicron (B.1.1.529). On 27th of December 2020, Italy started the vaccination campaign but limitations to inter-regional mobility were enforced since the beginning of January 2021 to contain the second pandemic wave. The re-opening of inter-regional mobility and other lifting measures started on April 26th 2021 and a progressive restart of all activities with safety protocols was obtained, starting from May 19th [28].

The state of knowledge reported that the virus spreads mainly through human-to-human contact, when the infected individual releases respiratory fluids through acts of breathing, speaking, coughing and sneezing or through contact with contaminated surfaces on which the droplets (>5 μm) have settled. The wide dimensional spectrum of respiratory particles (droplets and aerosol) that are transported differently by the airflow may change in size and composition according to the microclimatic conditions of the ambient air (e.g., temperature, relative humidity, etc.) [23]. Evidence that emerged indicated that in addition to the transmission via large droplets and fomites, SARS-CoV-2 was also transmitted via inhalation of smaller aerosol particles [29,30,31,32]. An extensive study regarding the airborne transmission of respiratory viruses was published recently by Wang et al. [33] concluding that the aerosol transmission pathway needs to be reevaluated for all respiratory infectious diseases. In December 2021, the World Health Organization updated the FAQs on how COVID-19 spreads between people, affirming that “another person can then contract the virus when infectious particles that pass through the air are inhaled at short range (this is often called short-range aerosol or short-range airborne transmission) or if infectious particles come into direct contact with the eyes, nose, or mouth (droplet transmission). The virus can also spread in poorly ventilated and/or crowded indoor settings, where people tend to spend longer periods of time. This is because aerosols can remain suspended in the air or travel farther than conversational distance (this is often called long-range aerosol or long-range airborne transmission)”.

No face coverings in indoor environments with poor ventilation and high levels of occupancy clearly increase the risk of transmission [34]. A study of fluid dynamics of SARS-CoV-2 reports that smaller airborne particles may also travel beyond the social distancing limit of 1–2 m [35]. Other numerical simulations show the methods of diffusions of particles in the case of two persons seated at the same table face-to-face with and without face masks [36]. Such evidence suggests the potential risk of COVID-19 infection in the typical setting of eating and drinking in indoor establishments confirmed by many outbreaks reported in food and beverage service activities [37,38]. In this respect, a canteen setting in a university/school may be compared to a restaurant or bar, in which people eat food and drink beverages without face masks.

In general, the “Education” sector reported concerns about high rates of infection from SARS-CoV-2 [39], and a high incidence of COVID-19 cases among workers was reported in different countries [28,40,41,42]. In Italy, during the first period of the pandemic, dining services in schools and universities (as well as in other working activities) were suspended. Governmental protocols for reopening activities provided specific measures for common spaces including canteens, related to: (i) restricting the amount of access, (ii) organizing specific ways for entry and exit, (iii) ensuring 1 m of distance between users, (iv) limiting the time of permanence, (v) providing for regular sanitization, and (vi) guaranteeing continuous ventilation of indoor spaces. In 2022, the Italian Ministry of Health published technical guidelines for ventilation systems to be installed in indoor spaces of universities and schools [43].

2.2. Site Description

This study was conducted within the university canteen of the Scuola Normale Superiore (SNS) of Pisa (Italy) in the period 10–13 May 2021. The places were frequented by students, PhD, postdocs, teachers and staff personnel of SNS for meals on two different shifts (lunch and dinner). The kitchen and canteen service workers have access to the site even before and after the two shifts for preparation and recovering. The service staff also take care of the cleaning and tidying of the rooms before and after each shift, using cleaning products and electrical equipment such as vacuum cleaners and floor cleaners. The dining hall (Figure 1) is located on the mezzanine floor of a building located in the city center (Figure S2), in a restricted traffic area. In the adjacent rooms and on the upper floors there are administrative and management offices. A specific area is used for kitchens where meals are prepared. The food distribution area consists of a counter with exhibitors to which canteen service staff have access on one side and users on the other side, via a self-service path. The dining area is divided into two main zones where the tables are positioned: dining area #1 (Din#1) with windows on the south side, and dining area #2 (Din#2) with windows on the south and east sides of the building. Between the two areas, there is recovery area #3 (Rec#3) dedicated to the storage of the trays used at the end of the meal.

Before the COVID-19 pandemic, the university canteen had about 280–300 people access simultaneously per daily shift, with an average stay time estimated by the managers of 30 min per person. Access to the canteen is via a system of turnstiles that open upon passage through a card reader supplied to authorized users. For the management of COVID-19 risk within its buildings, the SNS administration put in place safety protocols for the mitigation of the risk of contagion for students, its staff and canteen service workers. In addition to the general actions to ensure social distancing, hand hygiene, use of surgical (or superior) masks for staff and students, testing, tracking and promotion of vaccination, specific organizational and structural measures were implemented in the canteen areas, as it represents a high-risk setting due to the number of people who have access and to the activity of consuming meals that takes place there, also generally linked to moments of conversation and socialization without face coverings.

Such specific measures included:

• Identification of routes for entry and exit in order to reduce the possibility of gathering;
• Restriction of access to the canteen (occupation was reduced to a maximum of 120 people simultaneously);
• Instructions to users to stay only for the time necessary to consume the meal (no later than 30 min);
• Installation of Plexiglas® barriers between the face-to-face seats;
• Sanitization and cleaning of all surfaces after each shift.

In the areas used for the consumption of meals, the air changes were obtained through a mechanical ventilation system consisting of convectors, equipped with high-efficiency filtering plasma-based technology to reduce the fine fraction of the particulate before it accumulates in the filters of the convector and therefore increase the life of the filters themselves. Such technology is based on an electrostatic filter that collects airborne microparticles by means of electric fields created by applying a high voltage between electrodes. Particles ionized by the electric field are attracted to the surfaces of the collecting electrode. The filter has two stages of operation: particles are first charged by the discharge electrode (positive) by means of the corona effect, and then the charged particles are attracted to the collection surfaces of the plate electrodes of the second stage [44]. Manufacturers declared that the fan coil range can be fitted with the electronic filter, featuring a class D rating (80–90% average efficiency) according to UNI 11254 [45], with similar performances to the initial ones of a traditional mechanical filter featuring a class F9 rating (weighted average efficiency < 95% of 0.4 μm particles) according to EN 779:2012 [46] (updated by EN ISO 16890-1:2016 [47]).

2.3. Measurement Strategy

Measurements were performed at the dining hall in three working days divided into ten homogeneous daily time slots, according to the schedule reported in

Table 1. Daily time periods of measurements and related activities in the dining all.

No.	Period	Time Slot	Occupancy	Ventilation	Activities
T1	Night	0 am–6 am	None	Doors/windows closed; forced ventilation ON	None
T2	Morning	6 am–10 am	None	Doors/windows closed; forced ventilation ON	None
T3	Pre-Lunch	10 am–12 pm	Canteen personnel max 10 pers.	Doors/windows occasionally open; forced ventilation ON	Cleaning/washing and cooking preparation activities
T4	Lunch	12 pm–2:30 pm	Canteen personnel and students max 120 pers.	Doors/windows frequently open; forced ventilation ON	Lunch activities
T5	Post-Lunch	2:30 pm–4 pm	Canteen personnel max 10 pers.	Doors/windows occasionally open; forced ventilation ON	Cleaning/washing and recovery activities
T6	Afternoon	4 pm–5:30 pm	None	Doors/windows closed; forced ventilation ON	None
T7	Pre-Dinner	5:30 pm–7 pm	Canteen personnel max 10 pers.	Doors/windows occasionally open; forced ventilation ON	Cleaning/washing and cooking preparation activities
T8	Dinner	7 pm–9:30 pm	Canteen personnel and students max 120 pers.	Doors/windows frequently open; forced ventilation ON	Dinner activities
T9	Post-Dinner	9:30 pm–11 pm	Canteen personnel max 10 pers.	Doors/windows occasionally open; forced ventilation ON	Cleaning/washing and recovery activities
T10	Night	11 pm–0 am	None	Doors/windows closed; forced ventilation ON	None

. On day 1, electrostatic filters were activated on all the fan coils; on day 2, the filter option was turned off starting from 12 pm. Measurements continued during the night of day 2 until the morning of day 3 at 9 am.

The high-frequency real-time instruments involved in the measurement campaign were:

• Condensation Particle Counter (CPC mod. 3007, TSI Inc., Shoreview, MN, USA) to measure in real-time the PNC (part/cm3) from 10 nm to 1000 nm, with 1 s time resolution (1 Hz) and accuracy ±20% (total flow 0.7 L/min; detection limits 1 to 100,000 part/cm³).
• DiSCmini (DM mod. TESTO SE & Co. KGaA, Lenzkirch, Germany), handheld instrument for the measurement of personal PNC in the range 10–700 nm, average diameter (D_avg) of diffusion charging and Lung Deposited Surface Area (LDSA) in the range 10–300 nm, based on the model published by the International Commission on Radiological Protection [48], with a lower 1 s time resolution. Tygon^TM sampling lines 1.5 m in length were used to minimize the particle losses [49]. Three different DMs (DM-UF3, DM-UF4, DM-UF5) were used for simultaneous measurements.

The low-cost sensors included the following devices (further specifications are reported in Table S1):

• AirWits PM (AW-PM, GeNano Ltd., Espoo, Finland) to measure PM₁, PM_2.5 and PM₁₀, temperature (°C) and air humidity (%) every 30 min.; detection limits 0–1000 μg/m³, accuracy ±10 μg/m³ (between 0–100 μg/m³) or ±10% (between 100–1000 μg/m³).
• AirWits CO₂ (AW-CO, GeNano Ltd., Espoo, Finland) multipurpose meter, equipped with high-quality sensors to measure CO₂, temperature (°C) and air humidity (%) every 30 min; detection limits 0–5000 ppm, accuracy 50 ppm or three percentages from reading 0.2/2%. Two AW-CO devices were used during the campaign, denominated AW-CO1 and AW-CO2.
• AirWits IAQ (AW-VOC, GeNano Ltd., Espoo, Finland) to measure the amount of TVOCs, indoor air temperature (°C) and air humidity (%) every 30 min; detection limits 0–60,000 ppb; accuracy 15%. Two AW-VOC devices were used, denominated AW-VOC1 and AW-VOC2.

Instruments were placed as shown in Figure 2. The following measurements were taken near the center of dining area #1: PNC, D_avg and LDSA with DM-UF5; CO₂ with AW-CO1; PM1.0, PM2.5 and PM10 with AW-PM; TVOC with AW-VOC1; air temperature and relative humidity. The following measurements were taken near the center of dining area #2: PNC, D_avg and LDSA with DM-UF4; CO₂ with AW-CO2; TVOCs with AW-VOC2; air temperature and relative humidity. The following measurements were taken near the center of recovery area #3: PNC, D_avg and LDSA with DM-UF3. Outdoor PNC was measured with CPC (sampling point #4) on the south side of the building. It is worth noting that in dining area #1, fan coils recirculated indoor air, while in dining area #2 and recovery area #3, some fan coils were allowed to insert air from the outdoor environment (pale blue arrows in Figure 2). Progressive anonymous data counting of users’ access during the lunch and dinner shifts was allowed by the administrative office of the school. It has been supposed that each user remains inside the dining hall for 30 min, according to the COVID-19 safety protocol established by the University (see Section 2.2).

High-frequency instrument (CPC and DMs) comparisons were performed after and before the measurement campaign in order to align output values, according to a standard methodology [50,51]. Data values of instruments corrected after the comparison are indicated with an asterisk in the following text. DM data were calculated according to Fierz et al. [52].

3. Results and Discussion

3.1. Time Series and Average Data Related to the Activities Inside the Dining Hall

Figure 3 shows real-time high-resolution (1 Hz) measurement data of UFPs and micrometric particles release parameters (PNC, D_avg and LDSA) for two days simultaneously conducted in three indoor sampling points (#1, #2 and #3) and one outdoor (#4). The cumulative number of users progressively accessing the canteen for lunch and dinner is also shown on the secondary axis (black line). The mean and standard deviation values of the same parameters per each daily time period (as identified in Table 1) are reported in

Table A1. Day 1 average PNC, D_avg and LDSA per each time period (T1–T10) and in each sampling point (dining area #1, dining area #2, recovery area #3 and outdoor #4).

Day 1		PNC (part/cm3)				Davg (nm)			LDSA (μm/cm2)
		Din#1	Din#2	Rec#3	Out#4	Din#1	Din#2	Rec#3	Din#1	Din#2	Rec#3
T1	avg
12 am–6 am	st.dev.	n.a.	n.a.	n.a.	n.a.	n.a.	n.a.	n.a.	n.a.	n.a.	n.a.
T2	avg
6 am–10 am	st.dev.	n.a.	n.a.	n.a.	n.a.	n.a.	n.a.	n.a.	n.a.	n.a.	n.a.
T3	avg	3186	21,135	25,589	5753	59.67	33.72	36.21	4.87	41.14	47.97
10 am–12 pm	st.dev.	409	15,439	20,143	1755	9,26	7.77	8.18	0.67	22.10	24.06
T4	avg	7456	8359	12,233	12,112	32.69	25.03	28.72	9.12	11.34	19.72
12 pm–2:30 pm	st.dev.	3170	3173	5008	4363	7.84	4.17	3.98	4.07	5.76	6.91
T5	avg	10,751	6347	11,433	14,724	29.85	27.42	30.79	13.87	8.65	20.32
2:30 pm–4 pm	st.dev.	4620	2356	4332	4420	3.10	2.05	2.60	6.13	4.36	6.73
T6	avg	3649	3246	5233	11,396	38.49	37.05	29.66	3.97	3.82	8.29
4 pm–5:30 pm	st.dev.	325	152	545	3445	1.87	2.24	1.90	0.56	0.14	0.64
T7	avg	7677	12,332	26,778	8886	34.22	31.27	31.56	11.37	23.16	52.92
5:30 pm–7 pm	st.dev.	4167	9134	28,537	4743	5.13	5.90	7.83	9.28	20.83	61.31
T8	avg	24,706	25,093	30,060	8855	35.23	38.53	37.31	41.47	75.11	70.80
7 pm–9:30 pm	st.dev.	18,388	21,345	16,881	3735	7.06	12.10	7.89	29.32	63.06	40.74
T9	avg	11,981	22,791	20,328	15,908	47.23	38.01	45.46	30.66	64.74	61.38
9:30 pm–11 pm	st.dev.	4671	9241	5506	6628	4.31	4.86	3.52	13.31	27.83	19.65
T10	avg	5475	7331	8017	10,111	42.52	30.31	38.79	9.43	12.43	18.73
11 pm–12 am	st.dev.	907	1720	1745	791	1.39	1.17	1.35	2.43	4.29	5.15

and

Table A2. Day 2 average PNC, D_avg and LDSA per each time period (T1–T10) and in each sampling point (dining area #1, dining area #2, recovery area #3 and outdoor #4).

Day 2		PNC (part/cm3)				Davg (nm)			LDSA (mm/cm2)
		Din#1	Din#2	Rec#3	Out#4	Din#1	Din#2	Rec#3	Din#1	Din#2	Rec#3
T1	avg	8897	3642	9184	9035	24.53	22.59	22.00	8.95	2.55	11.21
12 am–6 am	st.dev.	1435	448	1410	3750	1.55	3.36	1.53	1.95	0.44	2.11
T2	avg	5806	2978	5650	5093	24.52	24.73	21.81	5.14	1.79	6.57
6 am–10 am	st.dev.	859	431	871	1397	2.23	3.96	2.28	1.60	0.86	1.76
T3	avg	10,230	11,047	11,607	7206	32.08	29.17	29.44	13.59	17.32	18.40
10 am–12 pm	st.dev.	8810	13,705	11,651	6499	3.75	3.94	3.35	12.71	23.65	17.90
T4	avg	24,142	21,060	21,033	12,796	33.23	31.92	31.32	48.02	53.09	44.78
12 pm–2:30 pm	st.dev.	9520	14,377	9883	12,189	5.80	6.34	5.69	21.27	28.35	20.03
T5	avg	23,808	17,497	22,202	8242	40.34	43.40	39.88	53.69	57.84	56.60
2:30 pm–4 pm	st.dev.	9657	6522	8878	1889	1.95	1.82	1.61	22.82	25.39	22.76
T6	avg	8779	6653	8105	6285	34.66	33.31	33.24	14.12	12.44	15.89
4 pm–5:30 pm	st.dev.	1221	1086	1007	3400	2.65	2.18	2.79	3.83	3.98	3.64
T7	avg	8450	11,325	9727	4948	39.29	35.40	37.83	16.74	29.90	24.64
5:30 pm–7 pm	st.dev.	5488	10,969	8384	2980	5.02	4.70	5.64	17.56	40.16	27.96
T8	avg	21,358	20,565	23,774	8521	41.74	34.20	37.65	49.13	46.68	53.44
7 pm–9:30 pm	st.dev.	9785	10,324	14,177	5460	6.55	5.75	5.77	25.28	21.22	25.10
T9	avg	11,683	10,654	11,756	5323	48.31	40.82	45.72	30.89	29.66	34.01
9:30 pm–11 pm	st.dev.	3978	4177	4180	922	4.10	0.85	2.23	13.90	14.33	12.00
T10	avg	5990	5649	5599	6046	45.78	39.92	44.13	11.88	12.15	14.31
11 pm–12 am	st.dev.	357	286	483	861	1.24	0.71	1.17	0.98	0.94	1.38

.

The PNC increase at all three indoor points (Figure 3a,b) appears strictly related to the users’ access and the dining activities during lunch/dinner time periods (T4 and T8) on both days. Before and after the meals (T3, T5, T7 and T9 periods), preparation and recovery activities were carried out by the canteen personnel (with up to 10 people inside the rooms). Such activities include cooking operations that generate UFPs, the use of cleaning products with possible VOC and secondary organic aerosol generation and the use of vaporization systems that generate water nanoparticles: this exposure scenario for canteen personnel is evident in particular before the lunch and dinner times, with PNC peaks in recovery area #3 higher on day 1 than on day 2. The highest values of indoor airborne submicronic particles associated with the activities inside the canteen are confirmed by average values of PNC during dining times (T4 and T8) compared to the other periods (Table A1). During the night and early morning (T1 and T10 periods) when no people are inside the canteen, PNC indoor values tend to approach the outdoor ones.

Outdoor PNC shows a similar general trend over the two days, with the average value increasing during the day and decreasing in the nighttime. On day 1, an unexpected release of particles until 40,000 part/cm³ is clearly identified after dinner time. On day 2, high concentration peaks until 140,000 part/cm³ before and during lunch time with greater variability than day 1 are reported.

A D_avg increase (Figure 3c,d) has been highlighted during the lunch and dinner times, higher on day 2 than on day 1. This growing trend is confirmed during the post lunch and dinner recovery activities. Otherwise, D_avg decreases during the preparation activities before lunch and dinner in which smaller particles may be generated by the indoor activities.

As expected, LDSA trends (Figure 3e,f) are correlated to the PNC in each sampling point and time period.

The analysis of PNC time series and average values in different daily time periods (T1–T10) provides evidence that human activities lead to a significant increase in indoor submicrometric particle concentrations.

The comparison between PNC and the related D_avg (Dp < 300 nm) values, which shows an increase beyond 30 nm during the activities inside the dining hall and reaches the minimum value of 22 nm (st.dev. 1.53 nm) in T1, shows the major contribution made by the size fractions in the Aitken mode (typically particle size between 30 nm and 100 nm) and accumulation mode (particles size range 110 nm–1 μm) compared to the nucleation one (particle size < 30 nm). By definition, the nucleation mode includes UFP newly formed through source condensation processes or atmospheric chemical reactions. Aitken mode particles are formed from gas-to-particle conversion at ambient temperature, as well as from the condensation of hot vapors during combustion processes. Particles with an aerodynamic diameter greater than 30 nm tend to have longer atmospheric lifetimes, allowing them to grow into the accumulation mode [53]. According to the literature, anthropogenic indoor sources of UFP may include cooking and heating activities related to meal services [54], electrical appliances [55] and emissions from building and furnishings materials [56]. Furthermore, the data analysis confirms that the presence of a human can represent a relevant source of particles [57]. Morawska et al. [58] also described the role of individuals in terms of lifting the particles deposited on the ground in the indoor environment and the transport of particles from the outside by the individuals themselves (i.e., by clothes, shoes, etc.). In other studies, the presence of people in indoor environments has been already associated with particle emissions from clothes abrasion [59] and bioaerosol release (i.e., skin fragments) [60].

In addition to human-related sources, indoor PNC may be dependent on the outdoor concentration and the penetration of the particles from the outside, not only due to the particle transport made by the people but also to the infiltration through the windows and doors (and their possible leakages) and through the mechanical ventilation system, which withdraws the outdoor air. For this reason, UFPs of outdoor air pollutants (e.g., vehicular traffic emissions) may enter and remain suspended in indoor environments [61]. This behavior is confirmed by the analysis of Pearson’s correlations between outdoor and indoor PNC in each sampling point (

Table A3. Pearson product moment correlation coefficient R between indoor and outdoor PNC on day 1, day 2 and day 3, per each sampling point and time slot.

	Day 1			Day 2			Day 3
	#1	#2	#3	#1	#2	#3	#1	#2	#3
T1	-	-	-	−0.0955	0.3388	−0.0212	0.7079	0.5263	0.4402
T2	-	-	-	−0.2379	−0.1295	0.1904	-	-	-
T3	0.1523	−0.1604	−0.2393	−0.0009	0.0081	−0.0314	-	-	-
T4	0.7583	0.5952	0.6043	0.4494	0.2438	0.3794	-	-	-
T5	0.4042	0.6501	0.5907	0.6509	0.6670	0.6793	-	-	-
T6	0.0725	−0.1095	0.1106	0.4881	0.5428	0.4658	-	-	-
T7	−0.5402	−0.1111	−0.1876	−0.1431	−0.1135	−0.1149	-	-	-
T8	0.6076	−0.0363	0.4222	0.0190	0.0857	0.0738	-	-	-
T9	−0.5115	−0.5948	−0.7251	−0.0873	−0.0914	−0.0979	-	-	-
T10	0.4881	0.5861	0.5235	0.6232	0.7511	0.7027	-	-	-

). The correlation is positive in the time periods with human activities inside the dining hall (T4, T5 and T8) and during the night (T10). The correlation is moderate and high for dining area #1, which is near outdoor sampling point #4 and has the same orientation; moderate and weak correlations are reported for dining area #2 and recovery area #3, which are farther than dining area #1 from outdoor sampling point #4. It is worth noting that the ventilation system also inlets outdoor air inside the indoor space by fan coils located in dining area #2, which means the outdoor PNC may influence indoor values in that sampling point.

3.2. Two Days Comparison and Plasma-Based Filtration Technology Impact on UFP Concentrations

Box plots in Figure 4 highlight the differences among PNC, D_avg and LDSA over two days during each lunch and dinner activity (T4 and T8 according to Table 1) at three different indoor sampling points. In the box plots, the lowest edge of the box indicates the 25th percentile, the line inside the box marks the median and the top of the box indicates the 75th percentile. Whiskers above and below indicate the lowest and highest values.

PNC values for lunchtime (T4) on day 1 results are significantly lower (Wilcoxon test p-value < 0.001) than on day 2 in all three indoor sampling points. D_avg average and median values are slightly lower on day 1 than on day 2. The same differences are not confirmed for dinner time (T8). In addition, these findings could be associated with influences from outside in which the higher values and variability of outdoor PNC on day 2 during lunchtime are not repeated at dinner time.

Furthermore, some differences between the daily trend of indoor PNC on day 1 and day 2 could be connected to the effects of the activation of the plasma-based technology filters on the fan coils. On day 1, electrostatic filters were activated, while on day 2, the filter option was turned off from 12 pm until the morning of day 3.

Figure 5 reports the values of γ which is the ratio between PNC (Dp < 700 nm) measured in each sampling point without the activation of the plasma-based technology filters (filters OFF) and the same parameters measured when the filters were used (filters ON):

γ = PNC _{30 min avg} (filters OFF)/PNC _{30 min avg} (filters ON)

(1)

PNC 1 Hz measurements were averaged over 30 min periods. The zero values in the scale correspond to the missing data.

In all the sampling points, γ had positive values during the dining periods (T4 12–14:30 and T8 19–21.30), reaching the maximum value of 3.2 in dining area #1 during lunchtime (T4), highlighting the contribution of plasma-based filtration technology to reduce PNC (Dp < 700 nm). This effect is also highlighted for dining area #2 (Figure 5b), which also reports γ values more than 1 during the nighttime. A less relevant contribution of the filtration system to reducing PNC is reported for recovery area #3.

3.3. Low-Cost Sensors PM, CO₂ and TVOCs Data Analysis

In Figure 6, the comparison between AW-PM low-cost sensor response and DM-UF5 high-resolution (1 Hz) PNC and D_avg data (30 min average) is reported. Instrument data are available only for day 2. Both devices were placed at the same sampling point (dining area #1). PNC was measured in the size range 1–700 nm and D_avg in the size range of 1–300 nm by DM-UF5. Although the low-cost sensor reported some missing data, the general trend alignment between PNC and PM₁ is respected, with both instruments highlighting two peaks in correspondence with the lunch and dinner activities (Figure 6a). Low D_avg values (30–40 nm) in Figure 6b correspond to higher PNC compared to PM₁, while high D_avg values (40–50 nm) reflect lower PNC compared to PM₁. This behavior is consistent with the major contribution of smaller particles given to the PNC as compared to the PM1. Otherwise, the contribution in terms of mass of the larger particle sizes is higher than the nanometric ones [62].

An added value of the present study is represented by the successful application of high-resolution real-time measurement techniques for nano and microparticles in occupational settings [63,64]. The findings confirm that the daily trends of PNC (Dp < 700 nm) and PM₁ are comparable, but the quantitative correlation is not respected depending on the size of the measured particles. In fact, as the size decreases, the contribution in terms of PNC becomes increasingly significant compared to the contribution in terms of mass concentration. Mass concentration, generally used for determining occupational exposure levels to harmful agents, is currently considered a metric less sensitive for nanoparticles compared to other metrics, such as surface area and PNC [65].

PM1 average concentration measured by the low-cost sensor in the same sampling location and time period (from 1:00 PM to 7:00 AM of the following day) with the activation of plasma-based filtration technology is reduced by a factor of 0.62 (

Table 2. Average PM values measured by the low-cost sensor from 1:00 PM to 7:00 AM in dining area #2, with (Filters ON) and without (Filters OFF) plasma-based technology activation.

Filters	PM1 (μg/m3)	PM2.5 (μg/m3)	PM10 (μg/m3)
ON	1.23	3.94	6.35
OFF	3.23	6.59	9.91

). PM_2.5 average concentrations with and without filter activation are, respectively, 3.94 μg/m³ and 6.59 μg/m³. The same reduction trend (with a low reduction factor of 0.36) allowed by the filters’ activation is shown for PM₁₀ average concentration. As reported in Table S3, the comparison between indoor PM measured values and outdoor data available from the nearest official station of the Regional Environmental Protection Agency (ARPAT) [66] for day 1 and day 2 shows an opposite relation: the daily average values of indoor PM_2.5 and PM₁₀ were lower on day 1 (filters ON) than on day 2 (filters OFF), although the respective outdoor daily average PM values were higher on day 1 than on day 2. This seems to further support the effectiveness of plasma-based filtration systems in reducing PM indoor concentration.

By properly evaluating these results, the local conditions of outdoor PNC on day 2 compared to day 1 should be considered, which also have an influence on the indoor PM values, as previously discussed.

PM_2.5 measurements obtained by low-cost sensors report average values lower than the limit values for the protection of human health to be attained by 1 January 2030 (not to be exceeded more than 18 times per calendar year) as reported by the recent EU Directive 2024/2881 on ambient air quality and cleaner air for Europe fixed at 25 μg/m³ [67]. The WHO Air Quality Guideline (AQG) states that annual average concentrations of PM_2.5 should not exceed 5 µg/m³, while 24 h average exposures should not exceed 15 µg/m³ more than 3–4 days per year [68]. In this view, the use of plasma-based filtration technology could make an effective contribution to reducing PM_2.5 below the recommended exposure limits.

A recent cohort study reports that long-term exposure to PM_2.5 could be associated with a higher risk of SARS-CoV-2 breakthrough infections and a reduced antibody response to vaccination [13]. In this view, the general reduction of PM_2.5 levels in the dining hall really contributed to the mitigation of COVID-19 risk.

Mean concentration values of CO₂ (ppm) and total volatile organic compounds (TVOCs, ppb) registered by low-cost sensors during each time period in two sampling positions (Din#1 and Din#2) inside the dining hall are reported in Figure 7. All available data per each time slot according to Table 1 have been averaged.

As expected, an increase in CO₂ levels (Figure 7a) corresponds to the presence of people inside the dining hall, with maximum average levels of 837 ppm and 636 ppm measured, respectively, in T4 and T8 (dining area #1). A specific correlation (Pearson R = 0.81) has been highlighted between the number of people inside the dining hall and the CO₂ levels, as reported in Figure A2. This correlation was obtained considering the daily average values of CO₂ (per 30 min intervals) and the estimated number of people calculated according to the time of access and supposing a maximum stay of half an hour inside the dining hall per person.

As a general recommendation, CO₂ concentration values permanently higher than 1000 ppm during the continuous occupation of indoor environments whose main physical and use characteristics are known indicate that external air exchanges and ventilation are insufficient and must be improved. This limit value was still considered adequate for the pandemic period [69]. Therefore, based on the findings of our study, the recommended average concentration limit of 1000 ppm was not overcome during the presence of each person inside the dining hall.

No significant correlation has been observed between CO₂ and PM₁ in indoor environments under the conditions of the present study (Figure S2).

TVOC indoor levels (Figure 7b) also arise during human activities, with the maximum values registered during the recovery activities (T5 and T9) after the dining periods. This behavior is more evident in dining area #1 compared to the other sampling point. The result is in line with evidence from the literature, which reports that both cleaning activities and occupants are sources of VOC generation in indoor environments [70].

3.4. Study Limitations

The major limitations of the study are related to the measurements carried out in only one session of three days, in the same seasonal climatic conditions, despite the fact the data collected every second (1 Hz) by high-resolution instruments allow significant statistics regarding the daily trends and the specific release events related to the monitored activities. Furthermore, the choice of sampling points was adapted to the needs of the dining hall in order to guarantee the use and normal activities of the occupants, also taking into account the COVID-19 procedure in force at the time of the experimental campaign. In addition, the study set-up had the advantage of making measurements in real exposure conditions, not simulated or performed ad hoc.

Another limitation could be associated with the effectiveness of real-time measurements, which may be influenced by differences between instruments used (i.e., CPC and DM for PNC and among three DMs with different calibration times) regarding principle of operation, size range, upper/lower limits of detection and operating conditions when the measurement was carried out. In order to overcome some of these possible drawbacks, an instruments comparison session was performed in the workplace before the sampling campaign to identify the correction factors and align the PNC, D_avg, and LDSA values measured by different devices. The CPC could be used as a reference instrument for PNC comparison according to Asbach et al. [71]. The spherical approximation of particles on which DMs are based may induce further limitations to D_avg and LDSA measured levels when the emissions are related to nanoparticles with different morphology. A comparison between three DMs was performed to choose the reference instrument according to Bellagamba et al. [51].

The major limitation related to the effectiveness of low-cost sensors is related to the low-resolution data available (1 measurement every 30 min), which was not enough for describing specific release events with short duration. Furthermore, some missing data (i.e., PM concentrations from 7 am to 1 pm on day 2) were found, probably due to the unstable connection of the sensors’ wireless system with the server and the related software for data collection. In addition, relative humidity and quantity of PM concentrations may have notable impacts on the low-cost sensor data [72]. Finally, there is little information available in the literature regarding the calibration tests of the low-cost sensors used in the present study, so the data collected may have uncertainty due to a lack of proper calibration procedures. Further intercomparison sessions by standardized methods should be recommended [73,74].

4. Conclusions

In the present study, the role of a mechanical ventilation system installed inside the dining hall of a university site and the effectiveness of the introduction of UFP filtration systems have been evaluated as a measure for improving air quality in indoor environments at specific risk from SARS-CoV-2 contamination (such as the dining hall) and in which other personal protective measures (as face masks) were not used during the consumption of meals.

An experimental campaign was conducted in May 2021 to mainly assess the impact of the filtration system on UFP concentration levels using high-frequency (1 Hz) real-time measurements related to the activities carried out in the canteen. Other indoor pollutants such as PM, TVOCs and CO₂ were monitored by low-cost sensors to evaluate the indoor air quality for the users (workers and students) as a possible indicator of COVID-19 risk mitigation.

In general, the measurements carried out allowed us to conclude that the opportunity that arose from COVID-19 mitigation measures to use plasma-based filtration systems showed a reduction in PNC and PM concentrations in the tested conditions of use, proving to be effective in improving airborne submicrometric particles exposure, in particular during the hours of users’ stay. This study demonstrates that a mechanical ventilation system equipped with plasma-based filter technology could also be effectively used after the pandemic at the university site to improve indoor air quality and UFP exposure levels as a general indoor air quality improvement measure for occupational safety and health. It will also have an impact in terms of health outcomes and costs associated with the ambient fine particle exposure that may increase the sustainability of the proposed solution.

This study confirms the importance of high-resolution monitoring techniques to evaluate UFP concentrations and dynamics in indoor environments.

Although with the described limitations, low-cost sensors showed moderate reliability compared to the high-resolution instruments in the monitoring of PM at low-resolution, in particular for describing macroscopic trends and significant gradients. They can also produce data useful to evaluate when other indicators of indoor air quality such as average CO₂ and TVOC concentrations may overcome the recommended limit values. Further studies to improve data collection in that direction are recommended.

In conclusion, based on the lessons learned by COVID-19 risk mitigation measures the improvement of low-cost sensors used to monitor indoor PM and other parameters relevant to the occupants’ health could be a promising sustainable application as a continuous indoor monitoring system to improve indoor air quality levels. In this view, future investigation and comparison with reference instruments are needed to overcome general biases of low-cost sensor measurements and improve the quality of their data.