
==== Front
Sci Rep
Sci Rep
Scientific Reports
2045-2322
Nature Publishing Group UK London

39232024
69731
10.1038/s41598-024-69731-7
Article
Comparison of cooking emissions mitigation between automated and manually operated air quality interventions in one-bedroom apartments
Pantelic Jovan jovan.pantelic@delos.com

1
Tang Mengjia 23
Byun Kunjun 1
Knobloch Yaakov 4
Son Young Joo 1
1 Well Living Lab, Rochester, MN 55902 USA
2 https://ror.org/00hj54h04 grid.89336.37 0000 0004 1936 9924 Department of Civil, Architectural, and Environmental Engineering, University of Texas at Austin, Austin, TX 78712 USA
3 https://ror.org/01qz5mb56 grid.135519.a 0000 0004 0446 2659 Buildings and Transportation Science Division, Oak Ridge National Laboratory, Oak Ridge, TN 37830 USA
4 https://ror.org/017zqws13 grid.17635.36 0000 0004 1936 8657 Department of Mechanical Engineering, University of Minnesota, Minneapolis, MN 55455 USA
4 9 2024
4 9 2024
2024
14 2063014 5 2024
8 8 2024
© The Author(s) 2024
2024
https://creativecommons.org/licenses/by-nc-nd/4.0/ Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
We implemented a crossover study design exposing 15 participants to two indoor air quality conditions in the Well Living Lab. The first condition, the Standard Control Condition, resembled the ventilation and air supply of a typical home in the USA with a manually operated stove hood. The second condition, Advanced Control, had an automated: (i) stove hood, (ii) two portable air cleaners (PAC), and (iii) bathroom exhaust. The PM2.5 sensors were placed in the kitchen, living room, bedroom, and bathroom. Once the sensor detected a PM2.5 level of 15 μg/m3 or higher, an air quality intervention (stove hood, PAC or bathroom exhaust) in that space was activated and turned off when the corresponding PM2.5 sensor had three consecutive readings below 6 μg/m3. Advanced Control in the overall apartment reduced PM2.5 concentration by 40% compared to the Standard Control. The PM2.5 concentration difference between Advanced and Standard Control was ~ 20% in the kitchen. This can be attributed to using the stove hood manually in 66.5% of cooking PM2.5 emission events for 323.6 h compared to 88 h stove hood used in automated mode alongside 61.9 h and 33.7 h of PAC use in living room and bedroom, respectively.

Subject terms

Mechanical engineering
Environmental impact
Delos Living LLCPanasonic North Americaissue-copyright-statement© Springer Nature Limited 2024
==== Body
pmcIntroduction

Studies have revealed that within residential settings, cooking emissions account for up to 73% of particle surface area concentrations in the indoor environment1, although in some cases ambient air can Cooking activities are linked explicitly to elevated short-term peak exposures compared to other periods within homes2 usually associated with cooking and post-cooking periods3. Moreover, when considering overall exposure, analyses of time activity patterns have demonstrated that people experience the highest exposure levels while at home during cooking4. The findings from these studies suggest that cooking is the most significant residential source of indoor pollution1,5–8. Exposure to cooking emissions has an adverse effect on respiratory health9, lung function10, cardiovascular health11, the human brain’s activities12, cognitive performance13 and is associated with elevated cancer risk14.

The emission rates during cooking are a function of cooking temperature, cooking method, type of food15,16, the heating source, burner size, cooking pan, cooking oil, additives, source surface area, ventilation, and position of the cooking pan on the stove are additional influencing parameters17. A study comparing emission rates of various cooking styles concluded that deep frying emits more particles than boiling, steaming, stir-frying, and pan-frying7,18. He et al.19 showed that frying could produce 745 μg/m3 peak particle levels. Besides PM emissions, gas phase pollutants are simultaneously emitted during cooking20,21.

Considering potentially high rates of cooking particle emissions, associated exposure and health risks, it is critically important to understand how to mitigate them. Mitigation strategies like stove hoods, opening kitchen windows, and portable air cleaners are often used alone or in combination. Opening kitchen windows during cooking can be an effective mitigation strategy reducing particle count up to 70%22.

The use of portable air cleaners (PAC) can effectively removes cooking-emitted PM2.518,22–24, but the location of the PAC to the particle source is a vital parameter25. Variation of apartment layouts and relative position of stove to the PACs showed that PAC can reduce exposure from 30 to 90%, suggesting wide distribution of effectiveness23. Studying impact of PAC location Another vital parameter is operating behavior. A study in China revealed that 81.4% of the PAC owners did not use them at all, among those who used it intermittently average operation time was between 1 and 4 h on the days the PACs were used, without causing the improvement of the indoor air quality26. PAC operating behavior study showed that occupants perceived “cooling” and “freshening” effects strongly correlating operation with the outdoor temperatures and thermal comfort27 rather than indoor air quality perception. Field comparison in Seattle of manual and automatically operated PAC showed that the automatic mode of operation reduced indoor concentration between 28 and 31% while manual operation resulted in PM2.5 concentration reduction between 4 and 19%27. Previous studies suggest that with the recent development of PAC with integrated sensing and possibility to operate in automatic mode, PAC are effective mitigation strategy against cooking emissions. Previous studies focused only on automated PAC and PAC did not investigate multiple automated interventions, leaving this as a knowledge gap.

The stove hood or range hoods are the most effective methods for control of cooking emissions24,28–30. The efficiency of the stove hood primarily depends on the cooking temperature, the exhaust airflow rate31, but can be affected by crossflows, so combining a stove hood with the open window has to account for the potential detriment of capture efficiency compared to a fully controlled makeup air supply32. For optimal performance, the balance has to be found between the concurrent air supply and exhaust by the hood33. To be effective, stove hoods must be used during cooking. Several studies reported survey based estimation of manual stove hood use. Most of the studies reported stove hood use between 28 and 34%34–36. Liu and Wallace37 reported lower use between 13% during winter and 10% in summer. Zhao et al.38, based on measured particle levels in homes and a machine-learning cooking detection algorithm, reported that stove hoods were used for 36% of cooking PM2.5 emission events in houses and 28% in apartments. Previous studies suggest stove hood is used in ~ 30% of cooking events and although very effective stays largely unutilized. Lawrence Berkeley National Lab (LBNL) report39 suggested that the most important barrier to the utilization of stove hoods is “the common misconception, based on a sensory perception of risk, that kitchen exhaust when cooking is unnecessary.” As a solution to this and other barriers, LBNL suggests public health efforts to raise awareness among the public and the building industry of the need to install and routinely use kitchen ventilation and development of methods for performance testing and setting of performance targets of high-performance products39.

With the recent advancement of Internet-of-Things and low-cost sensing technology40, the potential to overcome low manual stove hood utilization is to automate the operation. Pantelic et al. (2023)41 developed and experimentally tested cooking emission mitigation algorithms based on PM2.5 sensing and showed the potential to reduce PM2.5 concentrations for 90% compared to cooking without a stove hood.

Across research studies, surveys were the most commonly used method to provide insights into the frequency of stove hood use. Survey as a research methodology has inherent bias considering the general question and the uncertainties associated with human perception42. The study by Zhao et al.38 used a different method and quantified the frequency of stove hood use by approximating cooking PM2.5 emission events with machine learning detection based on particle counts. The current study is the first one that directly measures the use of both the stove and the stove hood and represents further progress in developing methods for detecting cooking PM2.5 emission events. The study by Pantelic et al.41 is currently the only research on the effectiveness of automated stove hoods that compared the cooking emission control effectiveness of automated stove hood with overhead air supply using one type of emission in a controlled setup. However, there remains a research gap in comprehending the relative effectiveness of automated air quality interventions when compared to manually operated methods. In this study, we aim to address this gap by evaluating the performance of manual and automated stove hoods, PACs, and bathroom exhaust fans in reducing PM2.5 concentrations within a one-bedroom apartment.

Methods

Study design

We recruited 15 participants (7 female, 7 male, and 1 other), collected and analyzed data. Details about the recruitment process, inclusion and exclusion criteria can be found in the Mayo Clinic IRB #20-007908 or in the Pantelic et al., 202343. All methods were performed in accordance with the relevant guidelines and regulations. The Mayo Clinic Institutional Review Board (IRB) approved all study procedures (Mayo Clinic IRB #: 20-007908). All participants provided written informed consent before the initiation of any study procedures. Study participants lived in the Well Living Lab’s (WLL, Rochester, MN, USA) residential modules configured as one-bedroom apartments (Fig. 1—floor plan, Suppl Fig. A1—photos of the laboratory). Participants lived in WLL residential modules, one at a time, for 4 weeks. Before starting the study, the participants were surveyed on indoor air quality (IAQ) features in their homes and habits related to IAQ. During the study, participants followed a defined set of activities in the residential module, including cooking breakfast and dinner. While cooking breakfast and dinner, participants followed provided recipes (Table A1 in the Suppl Appendix). Between participants, cooking emissions varied ~ 20%. For example, during burger cooking average peak emission was 560 μg/m3 with standard deviation of 117 μg/m3. The study participants followed a scheduled residential activity without disrupting the participant's daily living (e.g., going to work). The schedule for each participant consisted of a sequence of identical activities, but the time was tailored for each participant to accommodate their working hours. This schedule was designed to simulate exposure to residential activities (e.g., cooking and cleaning) and included periodic health assessments performed throughout these residential activities. More details about the participant's activity can be found in Pantelic et al., 202343. Besides scheduled activities in the module, participants utilized it as they would their home.Figure 1 Residential module layout.

We used a crossover study design with two conditions: (i) Standard Control and Advanced (ii) Control Conditions. Participants were allowed to manually turn on the stove hood during the periods with the Standard Control Condition. During the Advanced Control Condition, participants did not have manual control, and the stove hood, two PACs, and bathroom exhaust were activated automatically. The two control conditions are further explained in the following section.

Survey

Once participants enrolled in the study and were introduced to the WLL residential modules and instruments they are going to use in the study during their four-week stay, they received a link to the survey about IAQ-relevant features in their own homes, including the Heating, Ventilation and Air Conditioning (HVAC) system, habits related to cooking, ventilation, and general behavior related to the air quality interventions and indoor air quality management. Survey questions are provided in the Suppl Appendix.

Well Living Lab residential modules

The two residential modules had identical layouts (Fig. 1), with a floor area of 32.6 m2, including a kitchen, living room, bedroom, and bathroom. The ceilings were 2.6 m in height. A stove hood was installed above the stove. Windows in the modules were inoperable. Supply air diffusers were in each room; return grilles were located in the living room and bedroom; an exhaust grill was in the bathroom (Fig. 1).

Ventilation was provided through a central HVAC system with a Variable Air Volume (VAV) box in each module. In each module, the supply airflow and room temperature were controlled. Occupants could set their preferred air temperature (21 to 25 °C) on the thermostat installed in the kitchen, but each module received a fixed 140 m3/h combination of outdoor and recirculated air filtered by MERV14 filters, equivalent to an air exchange rate of 1.2 h−1. The outdoor air supply was 40% of the total flow rate or ~ 0.5 h−1. The airflow supplied to the residential modules was measured with the thermal diffusion airflow meter (ELECTRA-flo 5 Series Thermal Airflow Measurement System) connected to the Building Management System. The outdoor-to-recirculated ratio was kept constant throughout the study. Details about the air distribution can be found in Liu et al., 202224. In the same study, Liu et al., 202224 examined the impact of stove hoods, PAC, bathroom exhaust, and different combinations of these interventions on airflow patterns in the WLL residential modules.

The HVAC system was operated identically in the Standard Control and Advanced Control Conditions. The details of the two control conditions are as follows:Standard Control Condition consisted of an overhead air supply combining the recirculated and outdoor air. Four overhead four-way diffusers were placed in every zone, as depicted in Fig. 1, while return air grilles were in the bedroom and living room with the dedicated exhaust fan in the bathroom (Fig. 1). Standard Control was designed to maintain temperature set point by modulating air supply temperature. Study participants could set the air temperature they preferred on the thermostat placed in the kitchen. Only one thermostat controlled all the zones. The total supply flow rate was fixed at 170 m3/h and air supply was balanced between the supply diffusers. The outdoor air supply was set to match or be better than the ventilation requirements by ASHRAE 62.2-2017. Supply airflow was fixed to avoid differences in the dilution of cooking emissions. Besides air supply, Standard control consisted of a manually operated stove hood and bathroom exhaust fan. The stove hood had an on/off switch on the front part and the option of 4 speeds of the integrated exhaust fan. Participants could choose when to use the stove hood and at which fan speed they wanted. The bathroom had a manually controlled exhaust fan activated using a switch outside the bathroom.

Advanced Control Condition consisted of the air supply identical to the Standard Control Condition and decentralized air quality interventions controlled based on the PM2.5 levels in different zones. Identical to Standard Control Conditions, 170 m3/h of air (a mixture of outdoor and recirculated air) was supplied through four four-way diffusers depicted in Fig. 1. Decentralized air quality intervention consisted of a stove hood placed in the kitchen above the stove controlled based on the PM2.5 readings from the sensor installed at the side of the stove hood’s exhaust channel. Manual override of the stove hood was not enabled, and participants could not manually impact the operation. One PAC was placed in the living room and was controlled with the PM2.5 readings from the sensor installed in the living room above the couch (Fig. 1). The second PAC was placed in the bedroom and controlled with the readings from the PM2.5 sensor above the bed (Fig. 1). The PAC had no manual override, so the participants could not impact the operation. The bathroom exhaust was operated based on the PM2.5 reading from the sensor placed on the bathroom wall. The manual override was disabled, and the bathroom exhaust can only be operated automatically. All interventions were operated using open loop control algorithm. All air quality interventions are algorithmically controlled to power on when a single PM2.5 reading of the corresponding PM2.5 sensor is greater than 15 µg/m3 and to power off when three consecutive PM2.5 readings of the corresponding sensor are less than 6 µg/m3. For example, PAC in the bedroom is turned on or off only based on the PM2.5 sensor reading in the bedroom. The same principle applies to other air quality interventions as well.

Measurement instruments

IoT PM2.5 sensor

An IoT PM sensor (PurpleAir PA-II, PurpleAir, Draper, UT) was installed in each module room as shown in Fig. 1. All four sensors were mounted to a vertical surface at 1.2 m. The IoT PM sensor measured particle mass concentration in 6 different particle size ranges (0.3, 0.5, 1, 2.5, 5, and 10 μm), with a 0 to 1000 μg/m3 measurement range. Counting efficiency is 50% at 0.3 μm and 98% for particles with an aerodynamic diameter larger than 0.5 μm. The devices alternate measurements between two sensors every 5 s, and an average is taken over 120 s and sent to the data lake. After calibration, the measurement errors were ± 6 μg/m3 for measurements under 100 μg/m3 (details in the next section) and ± 10 μg/m3 for measurements in the 100 to 500 μg/m3 range. The sensor reported data every 2 min, transmitting it wirelessly to a cloud storage platform during the study period.

Calibration of purple air sensors

The Purple Air Sensors were calibrated using a recirculating wind tunnel that follows the ASHRAE 52.2-2017 Standard (Fig. A2 in the Suppl Appendix). The wind tunnel had a flow rate of 118 l/s (0.3 m/s average wind speed). An aerosol uniformity test was conducted following ASHRAE 52.2 standard before sensor calibration. A uniformity test was performed to ensure proper aerosol cross-section distribution at the sampling locations. The maximum Coefficient of Variation (CV) was 11.18%, which falls within ASHRAE 52.2 requirements. The Purple Air sensors were aligned in the tunnel with the inlet facing upstream of the airflow. They were mounted on lab stands and connected approximately 5 cm away from the tunnel's walls and 2.5 cm away from another sensor to prevent the boundary layer from affecting flow into the sensor. An isokinetic probe was attached to an Aerodynamic Particle Sizer (TSI Inc., Model 3321) and a Scanning Mobility Particle Sizer (TSI Inc., Model 3034) at 1 m in front of the sensors. Combining the APS data below 2.5 μm and SMPS data gave the PM2.5 concentration from 0.01 μm up to 2.5 μm. The Blaustein Atomizing Module (BLAM) generated KCl particles from a 3–5 wt% solution. The flow rate of the BLAM (2–5 l/min) was adjusted to change the concentration of PM2.5 in the tunnel. This produced an average size distribution of 0.7 μm for the following nominal mass concentration trials: 25, 60, 110, 180 μg/m3. The output of the BLAM was consistent, enabling the averaging process described in the following section.Figure 2 Data completeness: (a) PM2.5 sensors, (b) stove hood data. White squares indicate no participant in the study on that day.

A 0 μg/m3 trial (zero calibration conditions), where no aerosol was being generated, was done to confirm that the PM2.5 sensors, APS, and SMPS both read zero. The concentration was held in the tunnel for the duration of the test, which was 30 min. The measurement frequency of the Purple Air sensor was two minutes, the APS measured one sample per minute, and the SMPS required three minutes per sample. The APS samples were averaged over two minutes, and the SMPS time data, closest to the average time, was used. The APS and SMPS results were combined, averaged, and plotted against each Purple Air sensor result. The intercept was set at zero, and the resulting curve-fitted slope provided the correction factor (CF). This process was repeated for all Purple Air sensors and both channels of each PurpleAir Sensor. An example of regression is in Fig. A3 in the Suppl Appendix.Figure 3 Hourly average concentrations for the Standard and Advanced Control Conditions in the WLL apartment. Lines represent the hourly average PM2.5 concentration for all participants, and bands represent the 95% confidence interval of the hourly average.

During data processing, the readings of the two channels of PurpleAir sensors were divided by the corresponding correction factor, and their average was taken as the calibrated reading of the sensor. Values higher than the measurement range were replaced with 1000 μg/m3.

Dedicated circuit monitor

The stove was plugged into a Sense (Sense Labs Inc., Cambridge, Massachusetts) circuit monitor that monitored the stove's on/off status. The monitor reads the electrical current over 1 million times each second. Sense Labs Inc. has engineered a sophisticated circuit monitoring device that leverages Machine Learning (ML) to detect and differentiate the electrical signal patterns of various household appliances, including toasters, refrigerators, ovens, stoves, and incandescent light bulbs, among others. However, for the scope of our study, we did not employ the ML functionality of the Sense Labs circuit monitor. Instead, we connected the monitor to the power line serving only the stove. Consequently, any observed fluctuations in current were directly associated with the usage of the stove.

Stove hood

The Panasonic WhisperHood IAQ™ 30” Wall Mount Chimney Range Hood FV-36RCQL1, with dimensions of 900 mm × 486 mm × 600 mm (L × W × H), was installed above the stove in both residential modules before the study started. The stove hood can operate at 221 m3/h flow to 747.5 m3/h.

The status of the stove hoods (ON or OFF) was monitored in two ways, using a smart plug to read electrical energy use and monitoring and recording the actual IoT-connected stove hood speed of operation. These two sources of stove hood operations were used as cross-control methods to have better accuracy in measuring stove hood status/operation.

Portable air cleaners

Two BlueAir 480i portable air cleaners were installed in the living room and bedroom of the modules, as shown in Fig. 1. The fan can operate under three airflow speeds (204 m3/h, 340 m3/h, and 595 m3/h), with a clean air delivery rate (CADR) of around 476–510 m3/h when operated at the highest speed. The BlueAir 480i consumes 15–90 W. Given the BlueAir 480i’s CADR and the manufacturer’s recommendation for floor area coverage, two BlueAir 480i were installed, one in each of the two residential modules. During the experiments, devices were operated at medium speed with a CADR of 290 m3/h when turned on during the Advanced Control period.

Bathroom exhaust fan

The Panasonic FV-0511VK2 Exhaust Fan was installed in the bathroom of each residential module. The bathroom fan was set to exhaust at 109 m3/h. The Exhaust Fan was powered by a DC motor with a consumption of 5.1 W and a flow rate of 109 m3/h during the study.

Data completeness

Data completeness was presented daily for all devices and was ranked on a scale from 0 to 100%. For all data, we adopted criteria that designated ‘complete data’ for any given day for a participant must contain ≥ 60% of the expected data based on a particular device’s sampling frequency. We considered days that did not fulfill those criteria to be ‘incomplete data’ and removed these data from the analysis. If data was collected with a very high frequency, we first resampled the data to a lower sampling frequency and then applied the above completeness criteria.

Indoor PM2.5 data

Data completeness for PM2.5 sensors was calculated as a percentage by taking the number of non-missing 2-min measurements and dividing by the total number of 2-min intervals during the study period for each participant on each day (Fig. 2a). The x-axis of Fig. 4 represents the day number (Day 1 through 20), and the y-axis represents the unique participant IDs (IAQ05 to IAQ20). For clarification, the four participants in Pilot Study were assigned IDs from IAQ01 to IAQ04, which were excluded from the analysis. Therefore, in the Main Study, participant ID started from IAQ05 to IAQ20. IAQ14 was recruited but dropped out of the Main Study just before the experiment started.Figure 4 Hourly average concentrations for the Standard and Advanced Control Conditions in the kitchen. Lines represent the hourly average PM2.5 concentration for all participants, and bands represent the 95% confidence interval of the hourly average.

The completeness was higher than 80% for all participants and days except for IAQ05 on days 1 and 2, IAQ19 on days 3 and 4, and IAQ20 on day 2. Missing measurements were random in nature except for the period between 19:40 on day 2 and 4:20 on day 3 for IAQ19 and IAQ20 due to a network outage. Over 95% of missing values were random single-value gaps in the data. In this case missing measurements were filled with the nearest available forward-in-time data point for exposure calculation purposes. Beyond one missing value we treated data as missing and did not include that in the exposure calculation.

Stove hood data

The Stove hood data completeness is presented in Fig. 2b. Due to the data pipeline malfunctioning, data was lost during the last three days for participants IAQ15 and IAQ16 and the first 10 days for participants IAQ17 and IAQ18. Missing stove hood data were treated assuming participants did not cook between 11 pm and 5 am. We populated missing data in the following way: (1) if the data point before the missing period showed the stove hood off (this is true for most of the missing periods), and the data missing period is between 11 pm and 5 am, then it was assumed that the stove hood was off for the entire missing period; (2) if the data point before the missing period showed the stove hood on (there are three such missing periods), then the cooking stop time was used as the time when the stove hood was turned off. These periods occur when cooking was expected (5 am until 11 pm); 3) IAQ17 and IAQ18 were excluded from calculations of the stove hood use and period because data were completely lost for one study condition.

Data analysis methods

PM2.5 concentration was calculated for the kitchen PM2.5 sensor located at the side of the stove hood. We compared the performance of both conditions: (i) the Advanced Air Pollution Control Condition and (ii) Standard Control Condition.

Hourly PM2.5 concentrations in the overall apartment and in the kitchen

Collected PM2.5 data was aggregated on an hourly basis. We calculated each hour’s mean and confidence intervals and presented the average PM2.5 mass concentration hourly with mean and 95% confidence intervals.

We conducted our data analysis on two different levels. First, we assessed the entire residential apartment by combining data from all the sensors. This analysis aimed to understand the general air pollution trends in the apartment, considering the impact of cooking emissions and all deployed interventions. Second, we examined individual spaces or zones within the apartment. For the analysis in a specific zone, we used data collected by the sensors installed in those specific areas. This analysis aimed to understand how effective specific intervention was considering source proximity. For example, we provided the kitchen's hourly average concentrations in Fig. 4.

Cooking PM2.5 emission event detection

We used multiple data sources and cross-controlled them to have high confidence in detecting a cooking PM2.5 emission event. We consider cooking emission events occur when the stove is ON, detected by a dedicated circuit monitor, and when the PM2.5 level measured on the side of the stove hood increases rapidly above the background level. In this study, the background PM2.5 level was always less than 10 μg/m3.

Since the study focused on cooking emission mitigation, detecting the start time of cooking emissions was critical. Based on measurements, we detected a time when PM2.5 sharply increased compared to the background level. The cooking emission start time was determined as a timestep before PM2.5 concentration at least doubled compared to the background level.

PM2.5 concentrations in the kitchen during cooking

After determining the start time of the cooking emission event, we used that time to calculate the median PM2.5 mass concentration 15 min, 30 min, and 60 min after the cooking emission event started. We aggregated PM2.5 mass concentration measurements for a time interval of interest, calculated the median, first, and third quartile, and compared Advanced Control Condition and Standard Control Condition.

Standard vs Advanced Control PM2.5 concentrations for each participant

The PM2.5 data collected were aggregated for each participant and separately analyzed. Participants in the study spent time in the residential module but also left it to attend daily activities like going to work. We tracked occupants' movements in and outside of the residential module. In the analysis in this paper, we focus on exposures in the residential module, and therefore, we included only periods when participants were in the module.

Statistical analysis

The distribution of PM2.5 is not normally distributed. The lower values are clustered, and the larger values spread widely. Since we are interested in comparing the median and the spread of the distribution Wilcoxon Signed Rank Test can be used to conduct the nonparametric version of paired t-test to compare the median. We used Wilcoxon Signed Rank Test to compare results.

Stove hood use

The stove hood use ratio represents the ratio between the number of times the stove hood was ON while cooking and the number of cooking PM2.5 emission events.stovehoodratio=numberoftimesstovehoodwasonwhilecookingnumberofcookingevents

We collected information about the stove hood fan speed (i.e., 0, 1, 2, 3, and 4) every 10 min or when the fan status changed (i.e., when the fan speed changed or was powered on/off).

Collected raw Stove hood use data were first filtered to include only periods when cooking took place, and then removed periods with missing data. Within the clan data subset we created, we then calculated the number of times when the stove hood fan was on. This data set represents the basis of the stove hood ratio calculation. Since the data set contained information about the stove hood fan being ON or OFF and the fan speed, we calculated the total duration of the fan operation at each of the four speeds and compared these data between the Advanced Air Pollution Control and Standard Operation Conditions.

Results and discussion

Hourly PM2.5 concentrations in the apartment

The results in Fig. 3 for the overall apartment show that integrated concentration in the apartment during Advanced Control is ~ 40% lower than during the Standard Control Condition. The Advanced Control Condition reduced peak concentration by > 50% and exposure time in the apartment.

Pantelic et al.41 compared cooking-emitted PM2.5 mitigation effectiveness of automated stove hood combined with the HVAC system and showed ~ 90% reductions in integrated cooking-emitted PM2.5 concentrations for the system with an automated stove hood combined with the HVAC system compared to HVAC system alone. Results from the current study agree with the trend previously observed, but the difference in concentration reduction between Standard and Advanced Controls was smaller than expected. To understand this phenomenon further, we investigated PM2.5 concentrations in the kitchen in the next section.

Hourly PM2.5 concentrations in the kitchen

Results in Fig. 4 show that Advanced Control reduces hourly averaged peak concentration from 26 ± 4.1 to 18 ± 3.1 μg/m3 in the morning (during breakfast cooking) and from 24 ± 3.9 to 16 ± 2.8 μg/m3 in the evening. Alongside reduction in peak concentration Advanced Control reduces overall exposure time.

When integrated PM2.5 concentration for the standard control with manual stove hood is compared to the Advanced Control for 14 study participants, results in Fig. 4 show a difference of ~ 20%. This difference is lower than ~ 40% in the overall apartment. To better understand such a small difference, we investigated particle concentration in the kitchen during the cooking PM2.5 emission event, the manual use of the stove hood, how overall automation worked, and how control sequences worked.

When we compare the results for overall PM2.5 concentration in Fig. 3 with the PM2.5 concentration profile in the kitchen depicted in Fig. 4 and the living room, bedroom, and bathroom PM2.5 concentration profiles depicted in Fig. A4, we can observe that profiles had similar shapes, but different peak values. This suggest that residential module environment was not well-mixed during cooking periods.

PM2.5 concentrations in the kitchen during cooking emission events

Results in Fig. 5 show PM2.5 concentrations after cooking was started with three different timeframes of 15 min, 30 min, and 60 min.Figure 5 Comparison of PM2.5 concentrations in the kitchen under Standard and Advanced Control within 15 min, 30 min, and 60 min after cooking started for all participants. The boxes represent data within 1st and 3rd quartiles; the whiskers represent data within 1.5 times the interquartile range of the 1st and 3rd quartiles; the line in the box represents the median.

Results in Fig. 5 show that 15 min after cooking started, Standard Control with a manual stove hood produces statistically insignificantly (t-test, p > 0.05) different PM2.5 concentrations in the lower PM2.5 concentration in the kitchen compared to Advanced Control for cooking breakfast and dinner. This difference was statistically significant (p < 0.05), and it’s important to put this result in the context of manual and automatic stove hood operation. Several examples of stove hood operation along with PM2.5 concentration are depicted in Fig. 6. Results in Fig. 6a, b show that the participants manually turned on the stove hood (Standard Control condition) before the PM2.5 concentration reached 6 μg/m3. This is earlier than turning on by the control algorithm (Advanced Control Condition) at the PM2.5 concentration threshold above 15 μg/m3. Examples in Fig. 6 show that in the Advanced Control Condition, the PM2.5 concentration level can reach even beyond 100 μg/m3 (Fig. 6d) when the stove hood is activated (PM2.5 concentration in black dots, Fig. 6). Additional examples of Advanced Control operation in Fig. 6c, e show activation when PM2.5 mass concentration was above 30 μg/m3. This delay in starting the stove hood is attributable to the sensing frequency of 2 min. This contributes to the higher mean PM2.5 concentration for automated operation during the first 15 min after cooking started (Fig. 5). On average, the control algorithm turned on the stove hood 7.5 ± 2.2 min after cooking started41. Further investigation on stove hood operation is discussed in the next section, where we compare how occupants used stove hoods at home and in residential modules during the study.Figure 6 Examples of Standard and Advanced Controls for stove hood. x-axis is time, y-axis on the left side shows PM2.5 mass concentration, on the right side in fan speed. Line represents status of the stove hood and the dot represent PM2.5 mass concentration.

30 min and 60 min after the cooking starts, Advanced Control reduces PM2.5 concentrations to a lower level than the standard control with a manual stove hood (Fig. 5). Results show that 30 min after cooking started for both breakfast and dinner, Advanced Control reduced concentration to ~ 50% lower than standard control, from 16.8 ± 7.2 μg/m3 for standard to 7.7 ± 4.1 μg/m3 during Advanced Control Condition. (Fig. 5). T-test showed that difference was statistically significant p < 0.05. This is an expected trend because Advanced Control is designed to activate several IAQ interventions, including the stove hood and the portable air cleaners in the living room and bedroom that reduce PM2.5 concentration with a high clean air delivery rate24. This result also suggests that faster activation of the stove hood manually becomes less influential on the PM2.5 concentration as the time from the beginning of cooking increases.

Survey of participants' domestic use of air quality interventions during cooking

The survey responses show that 10 out of 14 participants have a central air system for heating at homes. Most of the participant's homes have operable windows and doors. 10 out of 14 participants have an electric stove at home, and the remaining 4 have a gas stove. Among the participants in our study, 12 cook meals at home 1–3 times per week. Most of the cooking methods include baking, frying, and microwaving. A large majority of participants cook red meat, chicken, fish, or bacon often. This food that contains fat, which results in high emission rates, especially during frying.

7 out of 14 participants, or 50%, have a stove hood at home. 4 out of 7 use it often at home, 1 uses it sometimes, and 2 use it rarely. The main reason participants avoided using a stove hood at home was the noise level the stove hood produced. Results from the survey show that 35.7% (combining answers often and sometimes but excluding rare use) of participants in the current study use a stove hood at home which is in line with pre previously reported range.

Regarding the methods to mitigate cooking emissions, survey results suggest that 10 out of 14 participants frequently opened windows while cooking at home; 2 out of 14 participants closed doors to other spaces during cooking. 5 out of 14 study participants observed indoor air quality changes while cooking. PACs were not frequently used among the study participants, with 2 out of 14 participants having portable air filters at home, and both had PACs because of pets.

Advanced and Standard Control Condition operation in WLL residential modules

Results in Table 1 show each participant’s stove hood use ratio and how long the stove hood was on during their stay in the WLL residential modules. We use the stove hood use ratio as a metric to see how often the stove hood was used during the cooking PM2.5 emission event. The stove hood ratio shows that the stove hood was active at 66.5% and 96.5% during the cooking PM2.5 emission events, respectively, in Standard Control and Advanced Control Conditions. While three study participants had a stove hood use ratio of less than 5% for the manual mode, 7 participants had a stove hood use ratio above 80% (Table 1) in the Standard Control Condition with a manually operated stove hood. Table 1 Stove hood use in Standard (manual) and Advanced (automated) controls.

Participant ID	Condition	Phase	Number of times fan was ON	Number of cooking events	Stove hood use ratio (%)	Total (h)	1	2	3	4	
IAQ05	Advanced control	2	17	19	89.5	3.1	0	0	3.1	0	
Standard control	1	0	19	0	0	0	0	0	0	
IAQ06	Advanced control	2	15	15	100	3.1	0	0	3.1	0	
Standard control	1	0	11	0	0	0	0	0	0	
IAQ07	Advanced control	1	17	20	85	4.4	0	0	4.4	0	
Standard control	2	15	19	78.9	23.8	18.7	5.1	0	0	
IAQ08	Advanced control	1	18	20	90	7	0.2	0.1	6.8	0	
Standard control	2	19	19	100	16.3	3.6	1.4	11.3	0.2	
IAQ09	Advanced control	2	18	18	100	5.8	0	0	5.8	0	
Standard control	1	6	12	50	25.1	25.1	0	0.1	0	
IAQ10	Advanced control	2	18	18	100	6.6	0	0	6.6	0	
Standard control	1	1	19	5.3	0.3	0.2	0	0.1	0	
IAQ11	Advanced control	1	20	20	100	6.6	0	0	6.6	0	
Standard control	2	16	19	84.2	20.6	0	17.1	3.5	0	
IAQ12	Advanced control	1	20	20	100	6.9	0	0	6.9	0	
Standard control	2	19	19	100	10.4	0	0.4	10	0	
IAQ13	Advanced control	2	19	19	100	7.6	0	0	7.6	0	
Standard control	1	17	20	85	10.2	2.1	5.6	2.6	0	
IAQ15	Advanced control	1	19	20	95	7.8	0	0	7.8	0	
Standard control	2	10	12	83.3	7.9	2.3	4.6	1	0	
IAQ16	Advanced control	1	19	20	95	6.2	0	0	6.2	0	
Standard control	2	7	12	58.3	2	0.8	0	1.2	0	
IAQ19	Advanced control	1	20	20	100	10.8	0	0	10.8	0	
Standard control	2	17	18	94.4	183.3	52	0	131.1	0.2	
IAQ20	Advanced control	1	20	20	100	12.3	0	0	12.3	0	
Standard control	2	18	19	94.7	23.8	1.2	17.9	4.8	0	
Total	Advanced control		240	249	96.4	88	0.2	0.1	87.8	0	
Standard control		145	218	66.5	323.6	105.9	51.9	165.5	0.3	

The total hours of stove hood use and stove hood use ratio in Standard Control condition was much higher for the participants who have stove hoods in their own home (total 45 h; stove hood use ratio 67) in comparison to the participants who do not have stove hood (total 11 h; stove hood use ratio 57). Moreover, among the participants who have stove hoods in their homes, participants who reported more frequent use of stove hood at home resulted in higher total hours of stove hood use and stove hood use ratio in Standard Control Condition. This suggests that our study results properly reflected participants' behaviors in their homes.

The stove hood operated for 323.6 h (or 140.3 h if we exclude IAQ19 as an outlier) in the Standard Control Condition (Table 2). Results in Table 2 for Advanced Control Condition show that the stove hood operated for 88 h (or 77.2 h if we exclude IAQ19 as an outlier), PAC in the bedroom for 33.7 h, and PAC in the living room for 61.9 h. These three interventions in total, operated for 183.6 h (Table 2) in the Advanced Control Condition. The bathroom exhaust was manually operated during Standard Control Condition and worked for 1563.2 h. Analysis showed that occupants frequently left the fan running when they left the module, so the fan had more than ten hours of uninterrupted operation. The bathroom exhaust fan was very quiet, and most probably, study participants could not hear it and had a tendency to leave it ON for extended periods. During the Advanced Control Condition bathroom exhaust operated for 237.2 h (Table 2). Based on the results in Tables 1 and 2 we can conclude that Advanced Control Condition had lower operation time than the Standard control condition with manual operation of stove hood and bathroom exhaust. The combination of more frequent use and ~ 3.5 (~ 2 times longer if IAQ19 is excluded) longer operation of the stove hood in Standard Control attributed to small differences in cooking emitted PM2.5 concentrations measured in the kitchen of WLL residential modules (Figs. 3 and 4). Table 2 Operation of Standard and Advanced control condition.

Participant ID	Condition	Stove hood (h)	PAC bedroom (h)	PAC living room (h)	Bathroom exhaust (h)	
IAQ05	Advanced control	3.1	1.1	0.9	15.1	
IAQ05	Standard control	–	–	–	176.0	
IAQ06	Advanced control	3.1	1.1	2.4	3.2	
IAQ06	Standard control	–	–	–	0.2	
IAQ07	Advanced control	4.4	2.0	3.3	1.5	
IAQ07	Standard control	23.8	–	–	115.7	
IAQ08	Advanced control	7.0	2.3	7.8	3.4	
IAQ08	Standard control	16.3	–	–	96.0	
IAQ09	Advanced control	5.8	2.8	5.2	4.2	
IAQ09	Standard control	25.1	–	–	120.0	
IAQ10	Advanced control	6.6	4.1	4.4	5.0	
IAQ10	Standard control	0.3	–	–	72.0	
IAQ11	Advanced control	6.6	4.3	4.7	2.8	
IAQ11	Standard control	20.6	–	–	–	
IAQ12	Advanced control	6.9	2.1	5.8	3.2	
IAQ12	Standard control	10.4	–	–	208.8	
IAQ13	Advanced control	7.6	1.5	2.7	3.3	
IAQ13	Standard control	10.2	–	–	121.2	
IAQ15	Advanced control	7.8	1.9	3.3	6.5	
IAQ15	Standard control	7.9	–	–	144.0	
IAQ16	Advanced control	6.2	1.8	4.7	103.8	
IAQ16	Standard control	2.0	–	–	144.0	
IAQ17	Advanced control		1.4	1.4		
IAQ17	Standard control		–	–		
IAQ18	Advanced control		2.4	4.1		
IAQ18	Standard control		–	–		
IAQ19	Advanced control	10.8	2.0	4.6	15.3	
IAQ19	Standard control	183.2	–	–	150.8	
IAQ20	Advanced control	12.3	3.0	6.8	69.9	
IAQ20	Standard control	23.8	–	–	214.5	
Total	Advanced control	88.0	33.7	61.9	237.3	
Standard control	323.6	–	–	1563.2	

PM2.5 concentrations 15 min from the start of the cooking show mixed effects (Fig. 7a): for some participants, the automated stove hoods in Advanced Control produce lower indoor concentration than in Standard Control, but for some, it is the reverse. This might be due to a personal cooking style that was not specified in the study protocol43, therefore, participants could have used higher or lower temperatures and cooked for a different time depending on their cooking style. The results for 30 min after the cooking started show that for most participants, the Advanced Control had a lower median PM2.5 concentration of cooking-emitted particles in the kitchen (Fig. 7b). The PM2.5 concentrations 60 min after the cooking is started are either lower or similar under the Advanced Control compared to the Standard Control (Fig. 7c). The participants in this study used the stove hood manually very frequently. During the Standard Control, as we compare the 66.5% stove hood ratio with the 10% to 36% range from the previous studies or an estimated 35.7% from the survey results. There are some differences between the studies, which makes direct comparison challenging. In the current study, we directly measure the stove hood use. In contrast, other studies relied on the self-reported frequency of the stove hood use, or the environmental parameters measurements to derive stove hood use data indirectly. Besides differences in measurement methodology in the current study, participants resided in the simulated apartment modules for one month. In the existing studies, participants were either surveyed, or environmental measurements were conducted in their actual homes. Being involved in the research study might have made the participants more aware of the necessity of using a stove hood.Figure 7 PM2.5 concentrations in the kitchen during (a) 15 min after cooking started, (b) 30 min after cooking started, (c) 1 h after cooking started. Participant IAQ16, although depicted in this figure, dropped out of the study. Since we had an incomplete data set for IAQ14, all data was removed from the previous analysis. In the current figure, we showed the results as an illustration.

On a conceptual level, we implemented decentralized air quality intervention in each zone (room) of the space and controlled them based on the PM2.5 levels in that zone. In other words, the logic embedded in the Advanced Control algorithm that operated decentralized air quality interventions was to activate intervention where it is needed when it is needed instead of relying on manual user operation. The cooking emission affected one-bedroom apartments where this study occurred, dispersed throughout all the spaces, increasing PM2.5 concentration above the threshold for activating air quality interventions. The key strength of decentralized air quality control implemented in this study is that it will activate air quality interventions in the spaces that might be affected by some other sources of pollution, like ambient air, that are more often the dominant source of indoor pollution44.

Ching-Hsuan et al.27 found that automated PAC reduced PM2.5 levels by 31% while PAC operated manually reduced PM2.5 levels on average by 4%. Cooper et al.45 pointed out that the use of PAC was motivated primarily by thermal comfort needs and not air quality perception, suggesting decision-making based on misperceptions. Pei et al.26, found that 81.4% of households with PAC do not use it at all, and 18.6% use PAC intermittently. These findings show that automation of the PAC increases it’s usability and effectively reduces PM2.5 concentrations compared to manual operation. Such studies do not exist for other air quality interventions. As pointed out in the introduction section Lawrence Berkeley National Laboratory (LBNL) research concluded that stove hoods suffer from the same issue as manually operated PAC. This study adds to the existing body of knowledge and advocate for technology solution through automation. The principle of decentralized automated air quality interventions will be effective in any home and for any type of pollution source, not just cooking emissions.

Limitations and future work

While participants knew they were participating in a study regarding indoor air quality, at no time were they informed of the study Condition in which they were participating. Instead, they were only given instructions at baseline (i.e., the beginning of the first study Condition) and the start of the third week (i.e., the beginning of the second study Condition) regarding what they could or could not do as it pertains to controlling or not controlling the stove hood and bathroom exhaust. They were otherwise left to live as they normally would within the residential modules. Notably, the study employed a crossover design, with participants randomized to a specific sequence of study Condition participation. Our data suggested that those participating in the sequence that saw them participate in the Advanced Control Condition first and the Standard Control Condition second used the stove hood longer during the Standard Control Condition than participants who experienced the study Conditions in the opposite sequence. One of the advantages of the crossover design is our ability to control for a sequence effect like that observed in our study during an a posteriori linear mixed effect model analysis, thus allowing us to start to tease out the effect to which you are alluding. Importantly, another advantage of the crossover design is that each participant serves as their own control (i.e., we compare intra-participant data between conditions), which provides sufficient statistical power for the linear mixed effects model used to analyze this type of design–even with a lower number of participants. Regardless, we did not employ this type of analysis for the current manuscript and plan to do that in our next publication. In this manuscript, we recognize this as a limitation.

We could not determine when exactly the cooking started or stopped. The exact set of activities participants do before cooking and after cooking is complete differ so we could not assume it’s the time when the stove was turned ON (burner can be turned on early then placing food on it) or OFF (a person can remove whatever is cooked to the adjacent burner or on the table and turn the burner later), which we monitor. We measured several proxies, like stove status ON/OFF, and concentration of particles in the space, but these are all proxies. There are no databases of time-resolved cooking emissions we can use to benchmark against and determine the exact start of cooking.

The hardware setup in this study with 2 min measurement frequency represents one of the limitations. As depicted in Fig. 6, there were instances when one sensor measurement was below 15 μg/m3 but the consecutive one was over 90 μg/m3. This impacted the effectiveness of cooking emitted particle removal. If the sensor had a higher measurement frequency stove hood would be activated faster. This issue can be resolved with changes in the hardware we intend to implement in future studies. Sensor accuracy is another hardware limitation. PurpleAir reported measurement errors for their sensor ± 10 μg/m3 for measurements under 100 μg/m3. We paid special attention to the lower detection limit during the calibration process. We calibrated sensors to detect as low as 6 μg/m3, which we used as a lower threshold to stop air quality interventions. In this study, we did not measure the chemical composition of cooking emissions. Chemical composition impacts the accuracy of the optical measurements method we deployed in this study. This might have made an impact on measurement accuracy. Although this is a technical limitation, practically, we are choosing a low value for control purposes, and uncertainties will make a negligible impact on the exposure.

Conclusion

We implemented a crossover study design exposing 15 participants to two control conditions in the Well Living Lab in Rochester, MN. Each participant spent four weeks living in WLL residential module. We implemented the Standard Control Condition, which resembled the ventilation and air supply of a typical home in the USA with a manually operated stove hood for cooking emission control. Besides standard ventilation and air supply, the second or Advanced Control Condition had an automatically operated: stove hood, two portable air cleaners, and bathroom exhaust. Automation was based on the residential modules' PM2.5 measurements in 4 locations, the kitchen, living room, bedroom, and bathroom.

Considering the whole apartment Advanced Control produced ~ 40% lower integrated concentration than Standard Control. In the kitchen, Advanced Control produced ~ 20% lower integrated concentration than Standard Control with a stove hood manually operated. Study participants used stove hoods in 66.5% of cooking PM2.5 emission events. The frequency of stove hood used in the study exceeded the use frequency of up to 36% previously reported in the literature. We surveyed all study participants about the frequency of stove hood use and, based on the survey results, estimated that participants turn on stove hoods in 35.7% of cooking emission events at their homes. Measurement of different indoor air quality interventions use suggests that participants operated the stove hood for 323.6 h in manual mode compared to 88 h in automated mode alongside PAC in the living room for 61.9 h and PAC in the bedroom for 33.7 h. This result suggests that automatic air quality control of cooking emissions is more effective because it operates in ~ 96% of cooking PM2.5 emission events, reduces indoor PM2.5 concentration, and overall operates for a shorter period then manually operated stove hood.

In particular, the Advanced Control reduces hourly averaged peak concentration from ~ 25 μg/m3 to 17 μg/m3 and exposure time. 15 min after cooking begins, Standard Control with a manually operated stove hood produces lower PM2.5 concentration in the kitchen than Advanced Control for cooking breakfast and dinner. During the first 15 min after cooking started, this start of operation lag was important enough to cause a concentration difference. 30 min after cooking started, Advanced Control reduced concentration to 50% below Standard Control.

Results for each study participant show mixed effects of control conditions on PM2.5 concentrations within 15 min after cooking started. For some participants, 15 min after cooking started, automated stove hoods under Advanced Control produced lower indoor concentration, but for some, it was reversed. This can be attributed to differences in participants habits related to the use of stove hoods. Besides personal stove hood use habits, there were also differences in cooking style (cooking temperature and duration) that were not specified and standardized in the protocol. Personal results for 30 min after cooking started to show that for most participants, advance control had a lower median PM2.5 centration of cooking-emitted particles in the kitchen. 60 min after cooking starts, PM2.5 concentration under Advanced Control is either lower or not different than Standard Control.

Supplementary Information

Supplementary Information.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-024-69731-7.

Acknowledgements

This study was funded by Panasonic Corporation of North America and Delos Living, LLC. The funding sources did not influence any aspect of this study. We acknowledge Delos Living LLC and Panasonic Corporation North America for providing funds to complete this project.

Author contributions

JP—wrote the main manuscript text MT—prepared all figures YK—performed sensor calibration YS—executed parts of the study, participated in data analysis KB—designed part of the study, participated in data analysis All authors reviewed the manuscript.

Data availability

The datasets generated and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request. Data will be available upon request The request should be made to Jovan Pantelic, PhD via email jovan.pantelic@delos.com.

Competing interests

The authors declare no competing interests.

Publisher's note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
==== Refs
References

1. Wan M-P Wu C-L Sze To G-N Chan T-C Chao CYH Ultrafine particles, and PM2.5 generated from cooking in homes Atmos. Environ. 2011 45 6141 6148 10.1016/j.atmosenv.2011.08.036
Wan, M.-P., Wu, C.-L., Sze To, G.-N., Chan, T.-C. & Chao, C. Y. H. Ultrafine particles, and PM2.5 generated from cooking in homes. Atmos. Environ. 45, 6141–6148 (2011).10.1016/j.atmosenv.2011.08.036
2. Zeng L Occupant exposure and ventilation conditions in Chinese residential kitchens: Site survey and measurement for an old residential community in Shanghai J. Build. Eng. 2020 31 101406 10.1016/j.jobe.2020.101406
Zeng, L. et al. Occupant exposure and ventilation conditions in Chinese residential kitchens: Site survey and measurement for an old residential community in Shanghai. J. Build. Eng. 31, 101406 (2020).10.1016/j.jobe.2020.101406
3. Patel S Indoor particulate matter during HOMEChem: Concentrations, size distributions, and exposures Environ. Sci. Technol. 2020 54 7107 7116 10.1021/acs.est.0c00740 32391692
Patel, S. et al. Indoor particulate matter during HOMEChem: Concentrations, size distributions, and exposures. Environ. Sci. Technol. 54, 7107–7116 (2020).32391692 10.1021/acs.est.0c00740
4. Park YM Personal exposure monitoring using GPS-enabled portable air pollution sensors: A strategy to promote citizen awareness and behavioral changes regarding indoor and outdoor air pollution J. Expo Sci. Environ. Epidemiol. 2023 33 347 357 10.1038/s41370-022-00515-9 36513791
Park, Y. M. et al. Personal exposure monitoring using GPS-enabled portable air pollution sensors: A strategy to promote citizen awareness and behavioral changes regarding indoor and outdoor air pollution. J. Expo Sci. Environ. Epidemiol. 33, 347–357 (2023).36513791 10.1038/s41370-022-00515-9
5. See SW Balasubramanian R Risk assessment of exposure to indoor aerosols associated with Chinese cooking Environ. Res. 2006 102 197 204 10.1016/j.envres.2005.12.013 16457802
See, S. W. & Balasubramanian, R. Risk assessment of exposure to indoor aerosols associated with Chinese cooking. Environ. Res. 102, 197–204 (2006).16457802 10.1016/j.envres.2005.12.013
6. Wei See S Karthikeyan S Balasubramanian R Health risk assessment of occupational exposure to particulate-phase polycyclic aromatic hydrocarbons associated with Chinese, Malay and Indian cooking J. Environ. Monit. 2006 8 369 376 10.1039/b516173h 16528421
Wei See, S., Karthikeyan, S. & Balasubramanian, R. Health risk assessment of occupational exposure to particulate-phase polycyclic aromatic hydrocarbons associated with Chinese, Malay and Indian cooking. J. Environ. Monit. 8, 369–376 (2006).16528421 10.1039/b516173h
7. See SW Balasubramanian R Chemical characteristics of fine particles emitted from different gas cooking methods Atmos. Environ. 2008 42 8852 8862 10.1016/j.atmosenv.2008.09.011
See, S. W. & Balasubramanian, R. Chemical characteristics of fine particles emitted from different gas cooking methods. Atmos. Environ. 42, 8852–8862 (2008).10.1016/j.atmosenv.2008.09.011
8. Li Y-C Characteristics of PM2.5 emitted from different cooking activities in China Atmos. Res. 2015 166 83 91 10.1016/j.atmosres.2015.06.010
Li, Y.-C. et al. Characteristics of PM2.5 emitted from different cooking activities in China. Atmos. Res. 166, 83–91 (2015).10.1016/j.atmosres.2015.06.010
9. Dockery DW Effects of inhalable particles on respiratory health of children Am. Rev. Respir. Dis. 1989 139 587 594 10.1164/ajrccm/139.3.587 2923355
Dockery, D. W. et al. Effects of inhalable particles on respiratory health of children. Am. Rev. Respir. Dis. 139, 587–594 (1989).2923355 10.1164/ajrccm/139.3.587
10. Moshammer H Hutter H-P Neuberger M Gas cooking and reduced lung function in school children Atmos. Environ. 2006 40 3349 3354 10.1016/j.atmosenv.2006.01.040
Moshammer, H., Hutter, H.-P. & Neuberger, M. Gas cooking and reduced lung function in school children. Atmos. Environ. 40, 3349–3354 (2006).10.1016/j.atmosenv.2006.01.040
11. Tan YQ Association between microenvironment air quality and cardiovascular health outcomes Sci. Total Environ. 2020 716 137027 10.1016/j.scitotenv.2020.137027 32044485
Tan, Y. Q. et al. Association between microenvironment air quality and cardiovascular health outcomes. Sci. Total Environ. 716, 137027 (2020).32044485 10.1016/j.scitotenv.2020.137027
12. Naseri M The impact of frying aerosol on human brain activity NeuroToxicology 2019 74 149 161 10.1016/j.neuro.2019.06.008 31254560
Naseri, M. et al. The impact of frying aerosol on human brain activity. NeuroToxicology 74, 149–161 (2019).31254560 10.1016/j.neuro.2019.06.008
13. Zhang X Chen X Zhang X The impact of exposure to air pollution on cognitive performance Proc. Nat. Acad. Sci. 2018 115 9193 9197 10.1073/pnas.1809474115 30150383
Zhang, X., Chen, X. & Zhang, X. The impact of exposure to air pollution on cognitive performance. Proc. Nat. Acad. Sci. 115, 9193–9197 (2018).30150383 10.1073/pnas.1809474115
14. Alves CA Cooking activities in a domestic kitchen: Chemical and toxicological profiling of emissions Sci. Total Environ. 2021 772 145412 10.1016/j.scitotenv.2021.145412 33581534
Alves, C. A. et al. Cooking activities in a domestic kitchen: Chemical and toxicological profiling of emissions. Sci. Total Environ. 772, 145412 (2021).33581534 10.1016/j.scitotenv.2021.145412
15. Atamaleki A Emission of aldehydes from different cooking processes: A review study Air Qual. Atmos. Health 2022 15 1183 1204 10.1007/s11869-021-01120-9
Atamaleki, A. et al. Emission of aldehydes from different cooking processes: A review study. Air Qual. Atmos. Health 15, 1183–1204 (2022).10.1007/s11869-021-01120-9
16. Buonanno G Morawska L Stabile L Particle emission factors during cooking activities Atmos. Environ. 2009 43 3235 3242 10.1016/j.atmosenv.2009.03.044
Buonanno, G., Morawska, L. & Stabile, L. Particle emission factors during cooking activities. Atmos. Environ. 43, 3235–3242 (2009).10.1016/j.atmosenv.2009.03.044
17. Amouei Torkmahalleh M Gorjinezhad S Unluevcek HS Hopke PK Review of factors impacting emission/concentration of cooking generated particulate matter Sci. Total Environ. 2017 586 1046 1056 10.1016/j.scitotenv.2017.02.088 28233614
Amouei Torkmahalleh, M., Gorjinezhad, S., Unluevcek, H. S. & Hopke, P. K. Review of factors impacting emission/concentration of cooking generated particulate matter. Sci. Total Environ. 586, 1046–1056 (2017).28233614 10.1016/j.scitotenv.2017.02.088
18. Sharma R Balasubramanian R Evaluation of the effectiveness of a portable air cleaner in mitigating indoor human exposure to cooking-derived airborne particles Environ. Res. 2020 183 109192 10.1016/j.envres.2020.109192 32062480
Sharma, R. & Balasubramanian, R. Evaluation of the effectiveness of a portable air cleaner in mitigating indoor human exposure to cooking-derived airborne particles. Environ. Res. 183, 109192 (2020).32062480 10.1016/j.envres.2020.109192
19. He C Morawska L Hitchins J Gilbert D Contribution from indoor sources to particle number and mass concentrations in residential houses Atmos. Environ. 2004 38 3405 3415 10.1016/j.atmosenv.2004.03.027
He, C., Morawska, L., Hitchins, J. & Gilbert, D. Contribution from indoor sources to particle number and mass concentrations in residential houses. Atmos. Environ. 38, 3405–3415 (2004).10.1016/j.atmosenv.2004.03.027
20. Ditto JC Abbatt JPD Chan AWH Gas- and particle-phase amide emissions from cooking: Mechanisms and air quality impacts Environ. Sci. Technol. 2022 56 7741 7750 10.1021/acs.est.2c01409 35671398
Ditto, J. C., Abbatt, J. P. D. & Chan, A. W. H. Gas- and particle-phase amide emissions from cooking: Mechanisms and air quality impacts. Environ. Sci. Technol. 56, 7741–7750 (2022).35671398 10.1021/acs.est.2c01409
21. Torkmahalleh MA PM2.5 and ultrafine particles emitted during heating of commercial cooking oils Indoor Air 2012 22 483 491 10.1111/j.1600-0668.2012.00783.x 22486983
Torkmahalleh, M. A. et al. PM2.5 and ultrafine particles emitted during heating of commercial cooking oils. Indoor Air 22, 483–491 (2012).22486983 10.1111/j.1600-0668.2012.00783.x
22. Xiang J Hao J Austin E Shirai J Seto E Characterization of cooking-related ultrafine particles in a US residence and impacts of various intervention strategies Sci. Total Environ. 2021 798 149236 10.1016/j.scitotenv.2021.149236 34340070
Xiang, J., Hao, J., Austin, E., Shirai, J. & Seto, E. Characterization of cooking-related ultrafine particles in a US residence and impacts of various intervention strategies. Sci. Total Environ. 798, 149236 (2021).34340070 10.1016/j.scitotenv.2021.149236
23. Sankhyan S Witteman JK Coyan S Patel S Vance ME Assessment of PM2.5 concentrations, transport, and mitigation in indoor environments using low-cost air quality monitors and a portable air cleaner Environ. Sci. Atmos. 2022 2 647 658 10.1039/D2EA00025C
Sankhyan, S., Witteman, J. K., Coyan, S., Patel, S. & Vance, M. E. Assessment of PM2.5 concentrations, transport, and mitigation in indoor environments using low-cost air quality monitors and a portable air cleaner. Environ. Sci. Atmos. 2, 647–658 (2022).10.1039/D2EA00025C
24. Liu Q Healthy home interventions: Distribution of PM2.5 emitted during cooking in residential settings Build. Environ. 2022 207 108448 10.1016/j.buildenv.2021.108448
Liu, Q. et al. Healthy home interventions: Distribution of PM2.5 emitted during cooking in residential settings. Build. Environ. 207, 108448 (2022).10.1016/j.buildenv.2021.108448
25. Novoselac A Siegel JA Impact of placement of portable air cleaning devices in multizone residential environments Build. Environ. 2009 44 2348 2356 10.1016/j.buildenv.2009.03.023
Novoselac, A. & Siegel, J. A. Impact of placement of portable air cleaning devices in multizone residential environments. Build. Environ. 44, 2348–2356 (2009).10.1016/j.buildenv.2009.03.023
26. Pei J Dong C Liu J Operating behavior and corresponding performance of portable air cleaners in residential buildings, China Build. Environ. 2019 147 473 481 10.1016/j.buildenv.2018.08.009
Pei, J., Dong, C. & Liu, J. Operating behavior and corresponding performance of portable air cleaners in residential buildings, China. Build. Environ. 147, 473–481 (2019).10.1016/j.buildenv.2018.08.009
27. Huang C-H Impacts of using auto-mode portable air cleaner on indoor PM2.5 levels: An intervention study Build. Environ. 2021 188 107444 10.1016/j.buildenv.2020.107444
Huang, C.-H. et al. Impacts of using auto-mode portable air cleaner on indoor PM2.5 levels: An intervention study. Build. Environ. 188, 107444 (2021).10.1016/j.buildenv.2020.107444
28. Singer BC Delp WW Price PN Apte MG Performance of installed cooking exhaust devices Indoor Air 2012 22 224 234 10.1111/j.1600-0668.2011.00756.x 22044446
Singer, B. C., Delp, W. W., Price, P. N. & Apte, M. G. Performance of installed cooking exhaust devices. Indoor Air 22, 224–234 (2012).22044446 10.1111/j.1600-0668.2011.00756.x
29. Lunden MM Delp WW Singer BC Capture efficiency of cooking-related fine and ultrafine particles by residential exhaust hoods Indoor Air 2015 25 45 58 10.1111/ina.12118 24750219
Lunden, M. M., Delp, W. W. & Singer, B. C. Capture efficiency of cooking-related fine and ultrafine particles by residential exhaust hoods. Indoor Air 25, 45–58 (2015).24750219 10.1111/ina.12118
30. Delp WW Singer BC Performance assessment of U.S. residential cooking exhaust hoods Environ. Sci. Technol. 2012 46 6167 6173 10.1021/es3001079 22568807
Delp, W. W. & Singer, B. C. Performance assessment of U.S. residential cooking exhaust hoods. Environ. Sci. Technol. 46, 6167–6173 (2012).22568807 10.1021/es3001079
31. Xie W Exhaust rate for range hood at cooking temperature near the smoke point of edible oil in residential kitchen J. Build. Eng. 2022 45 103545 10.1016/j.jobe.2021.103545
Xie, W. et al. Exhaust rate for range hood at cooking temperature near the smoke point of edible oil in residential kitchen. J. Build. Eng. 45, 103545 (2022).10.1016/j.jobe.2021.103545
32. He L Experimental studies of natural make-up air distribution in residential kitchen J. Build. Eng. 2021 44 102911 10.1016/j.jobe.2021.102911
He, L. et al. Experimental studies of natural make-up air distribution in residential kitchen. J. Build. Eng. 44, 102911 (2021).10.1016/j.jobe.2021.102911
33. Yi KW Kim YI Bae G-N Effect of air flow rates on concurrent supply and exhaust kitchen ventilation system Indoor Built Environ. 2016 25 180 190 10.1177/1420326X14541558
Yi, K. W., Kim, Y. I. & Bae, G.-N. Effect of air flow rates on concurrent supply and exhaust kitchen ventilation system. Indoor Built Environ. 25, 180–190 (2016).10.1177/1420326X14541558
34. Piazza, T. Study of Ventilation Practices and Household Characteristics in New California Homes (2007).
35. Chan, W. R., Kim, Y.-S., Less, B. D., Singer, B. C. & Walker, I. S. Ventilation and Indoor Air Quality in New California Homes with Gas Appliances and Mechanical Ventilation. https://www.osti.gov/biblio/1509678. 10.2172/1509678 (2019).
36. Klug, V. Characteristics of Range Hoods in California Homes Data Collected from a Real Estate Web Site (2011).
37. Singer BC Chan WR Kim Y-S Offermann FJ Walker IS Indoor air quality in California homes with code-required mechanical ventilation Indoor Air 2020 30 885 899 10.1111/ina.12676 32304607
Singer, B. C., Chan, W. R., Kim, Y.-S., Offermann, F. J. & Walker, I. S. Indoor air quality in California homes with code-required mechanical ventilation. Indoor Air 30, 885–899 (2020).32304607 10.1111/ina.12676
38. Zhao H Factors impacting range hood use in California houses and low-income apartments Int. J. Environ. Res. Public Health 2020 17 8870 10.3390/ijerph17238870 33260667
Zhao, H. et al. Factors impacting range hood use in California houses and low-income apartments. Int. J. Environ. Res. Public Health 17, 8870 (2020).33260667 10.3390/ijerph17238870
39. Stratton, J. C. & Singer, B. C. Addressing Kitchen Contaminants for Healthy, Low-Energy Homes. https://www.osti.gov/biblio/1220415. 10.2172/1220415 (2014).
40. Transformational IoT sensing for air pollution and thermal exposures. Front. Built Environ.10.3389/fbuil.2022.971523 (2022).
41. Pantelic J Son YJ Staven B Liu Q Cooking emission control with IoT sensors and connected air quality interventions for smart and healthy homes: Evaluation of effectiveness and energy consumption Energy Build. 2023 286 112932 10.1016/j.enbuild.2023.112932
Pantelic, J., Son, Y. J., Staven, B. & Liu, Q. Cooking emission control with IoT sensors and connected air quality interventions for smart and healthy homes: Evaluation of effectiveness and energy consumption. Energy Build. 286, 112932 (2023).10.1016/j.enbuild.2023.112932
42. Ligomenides, P. A. Modeling uncertainty in human perception. in Uncertainty in Knowledge-Based Systems (eds. Bouchon, B. & Yager, R. R.). 337–346 10.1007/3-540-18579-8_33 (Springer, 1987).
43. Pantelic J The impact of automated control of indoor air pollutants on cardiopulmonary health, environmental comfort, sleep quality in a simulated apartment: A crossover experiment protocol Front. Built Environ. 2023 9 5 10.3389/fbuil.2023.1117992
Pantelic, J. et al. The impact of automated control of indoor air pollutants on cardiopulmonary health, environmental comfort, sleep quality in a simulated apartment: A crossover experiment protocol. Front. Built Environ. 9, 5 (2023).10.3389/fbuil.2023.1117992
44. Morawska L Airborne particles in indoor environment of homes, schools, offices and aged care facilities: The main routes of exposure Environ. Int. 2017 108 75 83 10.1016/j.envint.2017.07.025 28802170
Morawska, L. et al. Airborne particles in indoor environment of homes, schools, offices and aged care facilities: The main routes of exposure. Environ. Int. 108, 75–83 (2017).28802170 10.1016/j.envint.2017.07.025
45. Cooper E Wang Y Stamp S Burman E Mumovic D Use of portable air purifiers in homes: Operating behaviour, effect on indoor PM2.5 and perceived indoor air quality Build. Environ. 2021 191 107621 10.1016/j.buildenv.2021.107621
Cooper, E., Wang, Y., Stamp, S., Burman, E. & Mumovic, D. Use of portable air purifiers in homes: Operating behaviour, effect on indoor PM2.5 and perceived indoor air quality. Build. Environ. 191, 107621 (2021).10.1016/j.buildenv.2021.107621
