Natural ventilation versus air pollution: assessing the impact of outdoor pollution on natural ventilation potential in informal settlements in India

Despite the proven benefits of natural ventilation (NV) as an effective low-carbon solution to meet growing cooling demand, its effectiveness can be constrained by poor outdoor air quality. Here, we propose a modeling approach that integrates highly granular air pollution data with a coupled EnergyPlus and differential equation airflow model to evaluate how NV potential for space cooling changes when accounting for air pollution exposure (PM2.5). Given the high vulnerability of low-income populations to air pollution and the dearth of energy and thermal comfort research on informal settlements, we applied our model to a typical informal settlement residence in two large Indian cities: New Delhi and Bangalore. Our results indicate that outdoor PM2.5 levels have a significant impact on NV potential especially in highly polluted cities like New Delhi. However, we found that low-cost filtration (MERV 14) increased the NV potential by 25% and protected occupants from harmful exposure to PM2.5 with a minor energy penalty of 6%. We further find that adoption of low-cost filtration is a viable low-carbon solution pathway as it provides both thermal comfort and exposure protection at 65% less energy intensity—energy intensity reduced to 60 kWh m−2 from 173.5 kWh m−2 in case of adoption of potentially unaffordable full mechanical air conditioning. Our work highlights ample opportunities for reducing both air pollution and energy consumption in informal settlements across major Indian cities. Finally, our work can guide building designers and policymakers to reform building codes for adopting low-cost air filtration coupled with NV and subsequently reduce energy demand and associated environmental emissions.


Introduction
Buildings contribute to over one-third of the total global energy consumption, most of which is consumed by mechanical cooling systems in tropical countries [1]. As cities rapidly urbanize, household incomes increase, and people spend more time indoors, the demand for air conditioning is expected to increase dramatically. Cooling energy demand has already tripled since 1990 and is expected to continue growing leading to possible grid issues in handling peak demands and environmental impacts like climate change [2]. Alternative passive cooling strategies offer a low-energy pathway to provide thermal comfort to occupants while minimizing energy usage and the associated environmental impacts. For example, natural ventilation (NV) is an effective passive cooling strategy that can be utilized to mitigate this growing energy demand by cooling building spaces without the use of mechanical systems and instead relies on the natural flow of wind when the outside weather conditions are favorable [3]. NV introduces higher air exchange rates in buildings, which can increase occupant satisfaction and productivity. NV has also been shown to improve the health of occupants by mitigating sick building syndrome [4]. The term 'sick building syndrome' is used to describe an increase in PM2.5 levels. Daily periodic variations have also been observed with increased PM2.5 levels during weekday traffic rush hours-both during the morning and evening [6]. A weekend effect was also observed when ∼10% higher PM2.5 concentrations were recorded in morning rush hour on weekdays in Indian megacities [17]. Also, in India, PM2.5 levels increase dramatically during the cold-dry months of October-January due to atmospheric conditions and fireworks burnt during the Diwali festival [7]. A study by the global burden of disease estimated 695 000 premature deaths in 2010 due to continued exposure to fine PM in India alone, a number which is expected to grow higher in subsequent years [25].
Atmospheric PM2.5 can penetrate buildings by infiltration of air through cracks or by natural and mechanical ventilation. Since urban residents in industrialized nations spend more than 80% of their time indoors, it is important to study the relationship between outdoor and indoor PM2.5 levels. The indoor/outdoor ratio (I/O ratio) is widely used to evaluate indoor PM2.5 pollution. In naturally ventilated buildings without significant indoor particle sources (such as combustion of coal for cooking/heating), indoor PM2.5 levels are similar to outdoors and the I/O ratio is close to 1 [24]. However, I/O ratios are higher than 1 in naturally or mechanically ventilated buildings with significant indoor sources of pollution. On the other hand, mechanically ventilated buildings with filters exhibit lower I/O ratios because the high-efficiency filters limit the infiltration of particles [26]. Tong et al studied the impact of traffic-related air pollution on indoor air quality in an NV office building using a CFD-based air quality model [27]. They found that indoor pollutant concentrations declined exponentially both with an increase in distance from the roadway and an increase in wind velocities. Furthermore, one study focused on how outdoor air pollutants impacted 28 low-income homes in Denver, USA during the wildfire seasons in 2016 and 2017 [28]. Results showed that indoor PM2.5 concentrations were 4.6 times higher than the outdoors on average and homes with exhaust hoods were observed to have a 55% lower PM2.5 I/O ratio than homes with no stove hoods, thus demonstrating that indoor air quality of low-income homes are affected by a combination of long-range wildfire plumes, proximity to roadways and occupant behavior. While all these previous works estimated the correlation between indoor and outdoor pollutant concentrations in NV mode, few studies have evaluated the impact of outdoor air pollution on NV potential in relation to the tradeoffs between occupant comfort (cooling) and adverse health impacts.
Tong et al investigated the impact of ambient air pollution on energy savings through NV for commercial building for 35 major cities in China [29]. They constrained the use of NV by setting thresholds for acceptable outdoor air temperature and indoor air quality levels according to China's National Ambient Air Quality Standards (NAAQS). The study found that although Beijing demonstrated lower per-square-meter energy savings potential as a result of unfavorable weather and air quality, it still offered the greatest potential for energy savings from NV adoption. Additionally, Martins & Carrilho da Graça studied the impact of PM2.5 specifically on the potential for NV cooling of office buildings for five locations in California, nine European cities and three megacities in Asia [30][31][32]. They employed rule-based controls to examine NV potential in different cities based on outdoor PM2.5 data and used a generic pollutant transport model (available in EnergyPlus) to predict indoor exposures to PM2.5 for occupants along with a detailed building energy simulation model to calculate changes in cooling demand when accounting for outdoor air quality [32]. They found that NV potential can be reduced by 10%-70% and corresponding energy savings by 20%-60% when considering outdoor pollution, as compared to when NV is operated based on outdoor temperature alone [31].
Furthermore, Costanzo et al combined building energy simulations with urban-scale CFD simulations to evaluate the combined effect of a dense urban layout with high pollutant concentrations using similar heuristic-based controls [33]. Based on their case study, NV potential was reduced from 4234 h yearly (neglecting air pollution) to 2707 and 529 h considering thresholds set by national Chinese standards (35 µg m −3 ) and WHO (15 µg m −3 ) [16], respectively. Also, Chen et al applied building energy simulations to investigate the influence of PM2.5, PM10, and ozone on NV usage for commercial buildings across 12 major cities in the US. They found PM2.5 to be the most significant pollutant, where NV potential dropped from 5% to 70% across the US when accounting for air pollutants [34]. In a follow-up study, they also observed the influence of air pollutants to be higher in urban areas as compared to rural areas in the US [35]. Finally, He et al and Eom et al studied the impact of particulate air pollution on domestic energy consumption for residents in Arizona, USA and Seoul, Korea, respectively, using a statistical analysis of consumer-level daily and hourly energy data [18,36]. They concluded that people stay indoors when air is heavily polluted, where their indoor lifestyles result in an increase in daily energy consumption, especially for low-income households in the US. For the Korean households, the rise in PM2.5 concentrations by 75 µg m −3 resulted in electricity consumption increasing by 11.2%-analogous to the impact of a 3.5 • C rise in mean temperatures in summer. While all these studies investigated the impact of ambient air pollution on NV potential and energy consumption, they did not evaluate the benefits of combining indoor particle filtration with NV to potentially reduce occupants' exposure to air pollutants.
Portable air filters can be used to protect occupants from harmful effects of air pollutants when the building is naturally ventilated. Cooper et al explored the impact of home air purifiers on indoor PM2.5 concentrations for 20 households in London [37]. They found that the PM2.5 concentrations in bedrooms reduced by a mean of 45% over 90 minute use of air purifiers. The study also sought to understand how occupants used air purifiers and observed that the occupants used purifiers more when thermal conditions were not favorable rather than the indoor air quality-indicating that it is not air quality that is driving occupants' satisfaction. Furthermore, prior work has also analyzed the impact of air filters on indoor exposure of primary school students to PM in China [38]. Results indicated that up to 70% of the PM2.5 concentration at the beginning of class was reduced over a 40-minute period due to the use of air filters alone. Additionally, a study demonstrated filtration to be more efficacious when the ventilation rates were higher [39]. The authors simulated a variety of filters ranging in efficiency corresponding to the minimum efficiency reporting value or MERV 8-16 and high efficiency particulate air filter (HEPA) and different fixed ventilation rates ranging from 20 to 100 CFM occ −1 for an office building in multiple locations in the US.
Their key takeaway was that improving filtration reduced exposure to excess PM2.5, and simultaneously, coupling higher efficiency filters with higher ventilation rates resulted in more effective filtration. As they increased ventilation rates from 20 CFM occ −1 to 60 CFM occ −1 , they observed the efficacy of filtration to improve by 1.2-1.5 for the commercial offices.
Ji et al identified appropriate ventilation-purification strategies for healthier indoor environments with lower energy consumption in a typical apartment building in Beijing, China [40]. They simulated two sets of strategies-mechanical ventilation with filters and NV with air cleaners-to estimate the effects of these strategies on energy consumption and indoor air quality. NV with air cleaners was found to provide the lowest exposure to indoor PM2.5 and achieved lower cooling and heating energy consumption compared to mechanical ventilation; however, mechanical ventilation with filters was observed to achieve lowest indoor CO 2 concentrations. In another recent study, Liu et al, applied annual building energy simulation to analyze the influence of three different ventilation modes (NV, mechanical ventilation, and hybrid ventilation) together with indoor particle filtration, on the concentrations of PM2.5, CO 2 and other volatile organic compounds as well as energy consumption in a newly constructed residential building complex in China [41]. While mechanical ventilation was observed to guarantee satisfactory conditions for indoor air quality throughout the study period, it was twice as costly to utilize mechanical ventilation compared to NV. The authors recommended using indoor air filtration during severe air pollution (PM2.5 levels above 35 µg m −3 ), coupled with NV to achieve the lowest energy consumption with lesser costs.
Existing works have primarily focused on commercial buildings, and less emphasis has been given to the residential sector. Our approach aims to extend this study to the residential sector with a particular focus on informal settlements where 1 in 8 of the world's population resides [42]. Our analysis focuses on large Indian cities as India is home to 35 of the world's 50 most polluted cities [15] and has significant populations of informal settlement dwellers [22]. One study validated the degradation in IAQ for low-income communities in India when indoor PM2.5 levels spiked more than 300 µg m −3 in the surveyed dwelling-interior design, occupant behavior and outdoor pollution contribute to indoor air pollution in these communities [21].
Additionally, existing research in India has mostly focused on evaluating the potential of NV based on outdoor meteorological data [43] and thermal comfort of occupants [44]. Previous research has been limited in its analysis of the Indian subcontinent and the context-specific appropriate time and duration for which NV should be utilized based on detailed analysis of periodic trends of outdoor air pollution. To address this gap, we combine highly granular air pollution data with building energy simulation and indoor air pollutant models to evaluate how NV potential for space cooling changes when indoor air pollution exposure is taken into account for informal settlements in the Indian subcontinent. Finally, we extended our previous work that presented a modeling approach to estimate the impact of air pollution on NV potential for informal settlements in India [45] to evaluate how NV potential changes when introducing indoor air filtration. Overall, this study aims to inform guidelines for building designers and architects, especially those engaged with in-situ retrofits and rehabilitation of informal settlements to better estimate the NV potential for different regions and to explore low-cost air filtration methods to mitigate the adverse impacts of air pollution.

Methods
In this section, we describe the methodology used to study the effects of outdoor PM2.5 on NV potential for major cities in India (figure 1). We utilize and pre-process highly granular 10-minute interval publicly The total number of favorable hours for NV was evaluated for a typical residential building representative of informal settlements. For the baseline model, feasibility of NV was based on zone apparent temperature. However, in the proposed model, indoor PM2.5 concentrations were modeled based on outside PM2.5 data using a first-order differential equation model. available PM2.5 concentration data from August 2019 to July 2020 for the Indian cities of New Delhi and Bangalore. They were chosen as the case study cities for the following reasons: • New Delhi experiences extremely high levels of outdoor PM2.5 pollution averaging over 200 µg m −3 , whereas Bangalore experiences comparatively lower levels of PM2.5 pollution averaging around 70 µg m −3 (for context WHO guidelines recommend less than 15 µg m −3 [16]). • According to the climatic zones defined by the Energy Conservation Building Code, India [46], New Delhi experiences a warm and humid climate, whereas Bangalore experiences a temperate climate. We aim to understand the difference in NV potential across two different cities in two different climates while keeping air quality in mind.
We modeled a typical validated residential building representative of informal settlements in India used in previous studies [47,48] in EnergyPlus [49,50]. We note that we only model the current horizontal form of informal settlements, specified as M1 in the previous studies [47,48]. M2 and M3 are the complete redevelopment scenarios for informal settlements proposed by the authors in [47] and [48]. Even though M1 is characterized by deep, narrow alleys representative of poor ventilation, we intended to study the impact of air pollution on the current scenario of informal settlements, here M1, to understand and quantify the harmful effects of air pollution on the existing vulnerable dwellers of informal settlements. Given the challenge of heat stress and air pollution in informal settlements in India and across the world [50], we concentrated our analysis on potential retrofit and rehabilitation strategies that could be implemented in-situ now and not wait for full-scale redevelopment to occur as this is likely to take several years or decades. Combining this building energy simulation with a differential equation airflow model allows us to evaluate indoor exposure of occupants to outside pollutants, together with indoor particle filtration. Finally, we run Table 1. Simulation parameters for the model (informed by [47,48]

Outdoor PM2.5 data and pre-processing
We obtained outdoor PM2.5 concentration data from the Open Air Quality (OpenAQ: openaq.org) and Purple Air (purpleair.com) databases for 15 min intervals from 1 August 2019 to 31 July 2020. We utilized the low-cost secondary sources of OpenAQ and Purple Air because they are ubiquitously available across many Indian cities, have sensors located close to informal settlement areas, and provide adequate accuracy for conducting our first-order analysis on the impact of NV potential [52]. To perform the data pre-processing to ensure data quality, we considered OpenAQ to be the primary database because OpenAQ is third party verified and PurpleAir to be secondary because PurpleAir utilizes crowd-sourced data. We utilized PurpleAir Data to fill in the gaps where OpenAQ data is not available for more than 24 h continuously. Missing data for less than 24 h in OpenAQ database was linearly interpolated, as recommended in [26]. Missing data for more than 24 h were replaced with high-resolution data from the Purple Air database. We also treated concentrations greater than 700 µg m −3 as outliers and replaced them with missing data, which were also linearly interpolated.

Detailed building energy simulation model
In this study, we used a validated residential model for informal settlements as developed by authors in [47,48]. The model had a horizontal morphology as depicted by the Building Model in figure 1. The building is composed of 28 21 m 2 individual units each dispersed over a single floor, the 28 units were considered to be individual thermal zones for the purpose of building simulation. It is characterized by deep, narrow alleys representative of poor housing conditions with low exposure to sunlight and low ventilation.
The window-to-wall ratio is set at 15%, and we assume that the building will be operated for 24 h. The occupant density was set at 0.64 people m −2 , and NV was observed to be the only source of fresh air exchange in the simulated zone. This model was validated using onsite surveys and in-situ sensors which measured indoor temperature and relative humidity. Data-driven manual iterative calibration technique was employed and achieved an acceptable hourly mean bias error of 1.07% and coefficient of variation for root mean square error of 2.26% [48]. This residential model formed the typical archetype of informal settlement dwelling in our study for the two study cities. The building energy model is simulated in EnergyPlus v9.1-and the other simulation parameters are listed in table 1.

Air filters in EnergyPlus
To account for indoor particle filtration, we adopted the method of modeling portable fans in EnergyPlus, which have the properties of air filters embedded in them from [39]. Air filters have a minimum efficiency reporting value (MERV), which is a subjective measure of efficiency. The pressure drop across the filter is a function of filter type and MERV, as well as the manufacturer and age of the filter [53]. David and Waring [39] estimated the pressure drops across the non-HEPA filters to be ranging from 111 to 157 Pa and the pressure drop across the HEPA filter to be 374 Pa. Since air filters cannot be modeled directly in EnergyPlus, we modeled an air intake fan using a ZoneVentilation object. As suggested by David and Waring [39], we added the pressure values corresponding to the filters to the pressure rise across the fan, indicating that more work is required by the fans to move the same amount of air for a filter with a larger pressure drop. We designed this method to mimic the effect of air filters realistic in our study area, wherein, these filters would be attached to an 'intake fan' such as an exhaust fan.

Indoor air pollutant modeling
We carried out indoor air pollutant modeling using a first-order differential equation based on the conservation of mass to quantify the indoor exposure to occupants. This approach has already been validated for its effectiveness in simulating the transport of air pollutants from the outdoor environment to the indoor [34]. Using the following equation, we calculate the rate of change of indoor concentration C i in building zone k at each timestep t as a function of different sources (S i ) and losses (L i ) of the pollutant in that environment: is the concentration of pollutant i at timestep t in thermal zone k. Pollutants can travel into the indoor environment via airflow through windows, neighboring zones and the ventilation system. Difference losses include the transport of indoor pollutants outdoors through windows, and transport of pollutants from zone k into neighboring zones. We simulated different airflow rates for transport of pollutants through the AirflowNetwork Model in EnergyPlus. As per outputs from EnergyPlus that analyzed the air change rates at each node, typical air exchange rates were observed to vary between 3 and 5 ach. Deposition rates were found to range from 0.2 h −1 up to 1.2 h −1 in a range of field studies [34]. We ran a sensitivity analysis with different deposition rates of 0.2, 0.8 and 1.2 h −1 . Increase in deposition rate resulted in loss of air pollutants. As the deposition rate increased from 0.2 to 0.8 h −1 , indoor air pollutant concentration was found to decrease at an average of 4%-5%. In this paper, we chose the deposition rate to be 0.2 h −1 as a conservative estimate for air pollution and thus provide a lower-bound for energy saving potential. To model indoor particle filtration, we considered different filter efficiencies and recirculation rates to account for recirculation loss of the pollutant in that particular environment.
We assumed that within each timestep, generation rates of source and loss rates of pollutant i remained constant. We note here that our study is focused on the impact of outdoor pollutants and therefore, indoor sources of the pollutant and reactions between different pollutants were not modeled in this study.
We verified the indoor air pollutant model by comparing the fluctuations in indoor air pollutant concentration with the outdoor levels while operating the building in NV mode. Figure 2 represents the pollutant modeling results for New Delhi and Bangalore respectively from August 2019 to July 2020. The red line represents indoor concentrations whereas the blue line represents the outdoor concentration. As expected, it can be seen that indoor PM2.5 concentrations follow closely with outdoor PM2.5 levels.

Operational scenarios for evaluating NV potential
In this section, we describe three different scenarios for utilizing NV, considering influence of air pollutants against two different thresholds set by WHO and NAAQS versus without their influence, to assess the NV potential for the two cities. We estimate NV potential by carrying out building energy simulations in EnergyPlus using the AirflowNetwork module and calculating the number of hours when the conditions are favorable for NV. The scenarios are summarized in table 2 and modeled as follows: • Baseline Scenario (I): A scenario where NV is controlled on the basis of heat index or apparent temperature. Heat index is a function of indoor operative temperature and humidity (Heat Index-National Weather Service). The organization has defined low, moderate, and high caution levels for heat index where each level is defined as having an apparent temperature less than or equal to 32 • C, 39 • C and 46 • C respectively. We calculate the number of favorable hours for NV corresponding to all three caution levels without considering the influence of air pollutants. • NAAQS Scenario (II): NV is favorable according to rules defined in Baseline scenario and when indoor PM2.5 levels are below the NAAQS India threshold of 60 µg m −3 [51]. • WHO Scenario (III): NV is favorable according to rules defined in Baseline scenario and when indoor PM2.5 levels are below the WHO threshold of 15 µg m −3 [16] We also note that previous work has indicated that inhabitants in Indian subcontinent are customarily more comfortable in warmer temperatures [54], we base our further results on the moderate caution levels (i.e. 39 • C indoor apparent temperature) for the summer months and low caution levels (i.e. 32 • C indoor apparent temperature) for the winter months for the Indian subcontinent. In this paper, this subsequently will be referred to as 'Recommended Caution Level' .
Going forward, we will use the nomenclature from table 2 for referring to scenarios: Baseline Scenario with Low Caution will be referred as IA, Baseline Scenario with Moderate Caution will be referred as IB and so on.

Results
In this section, we present our results for assessing feasibility of NV potential across the three scenarios for two Indian cities: New Delhi and Bangalore. We evaluate NV potential by calculating the number of hours when NV can be utilized for a particular scenario. We also explored indoor particle filtration as a cost-effective mitigation strategy to combat the negative effects of air pollution. Our results indicate three main takeaways from our study: (1) NV potential decreases significantly with the rise in outdoor PM2.5 concentrations but filtration aided NV could reduce exposure to outdoor PM2.5.
(2) Ventilation strategies should be localized based on climate and pollution patterns that vary across locations and seasons. (3) Low-cost filtration is a viable low-carbon solution because it protects occupants from harmful exposure at a significantly lower energy intensity than mechanical cooling systems

Filtration aided NV could reduce exposure to outdoor PM2.5 and enhance viable hours
We first analyzed the impact of air pollution on NV potential, without considering indoor particle filtration. New Delhi experiences a warm and humid climate (average temperature is 32 • C ± 7 • C in summer and 15 • C ± 6 • C in winter), whereas Bangalore experiences a temperate climate with average temperature 24 • C ± 5 • C for the entire year. Bangalore's more temperate climate gives it a higher starting potential for utilizing NV year-round, as can be seen from figure 3(B). When compared to the hotter climate of New Delhi, Bangalore's starting potential is 28% higher (Baseline-I A). As expected, as the heat index caution level becomes less strict (i.e. as we increase the threshold from Low to Moderate to High caution), the acceptable comfortable hours for NV increase for both New Delhi and Bangalore across the Baseline Scenarios-I A, I B and I C but the NV potential drops much more in New Delhi than Bangalore in the  )). This is consistent with previous works that evaluated the NV potential in Indian cities [43,55].
Overall, we found that the NV potential decreased dramatically owing to adverse outdoor air pollution conditions. New Delhi's NV potential drops significantly at each new heat index threshold (see figure 3(A)). We observe an average reduction of 57% in NV potential for the NAAQS Scenarios-II A, II B and II C and a decrease of 90% for the WHO Scenarios-III A, III B and III C compared to the corresponding baseline scenarios. The significant reduction in New Delhi is due to very high concentrations of PM2.5 where more than 99% of the total population in New Delhi is exposed to average PM2.5 levels higher than 200 µg m −3 [7]. However, Bangalore has a smaller drop under NAAQS standards (3%), but when under the WHO standard, NV potential drops much lower by 65% relatively when compared to the baseline ( figure 3(B)). While Bangalore's NV potential is influenced primarily by PM2.5 levels, Delhi's NV potential is restricted by both constraints of apparent temperature and high outdoor air pollution.

Yearly feasibility for NV with filters
As previous studies have demonstrated, air filters can mitigate the harmful effects of air pollution by reducing the concentration of indoor air pollutants (here, PM2.5). The amount of pollutant filtered depends on the type of air filter, its efficiency, ventilation rates, and recirculation rates. Informal settlements are typically equipped with some form of low-energy mechanical ventilation systems, such as ceiling fans or exhaust fans, in order to supplement NV for their comfort needs. Therefore, for the purpose of this study, we assume that air filters would be attached to an 'intake fan' such as an exhaust fan to model air filters in EnergyPlus. We chose this approach as this would be a feasible method of deployment in an informal settlement context. Here, we analyze filters with different filter efficiencies and recirculation rates to determine how NV potential changes for both New Delhi and Bangalore under the three different operational scenarios. We conducted a sensitivity analysis, where we considered three filter efficiencies corresponding to 30% (MERV 10), 70% (MERV 14) and 99.7% (HEPA) [39] and five recirculation rates corresponding to 2 ach, 4 ach, 6 ach, 8 ach, and 10 ach. We analyze the increase in yearly percentage favorability for NV compared to the Baseline (without any air filter). Through sensitivity analysis, we quantify the uncertainty in the potential for NV based on different filter efficiencies and recirculation rates. Figure 4 shows increase in percentage favorability for NV in New Delhi for the WHO-III D ( figure 4(A)) and NAAQS Scenarios-II D ( figure 4(B)) respectively, compared to the Baseline in which air filters were not used. We observe that as filter efficiency increases, the potential for utilizing NV increases. We also observe that as the recirculation rate increases, NV potential rises. HEPA is the most efficient filter, which filters polluted air with 99.7% efficiency, followed by MERV 14 which has efficiency of 70%. Correspondingly, we observe that HEPA corresponds to the highest percentage favorability for NV, followed by MERV 14 irrespective of recirculation rates. When the recirculation rate of air filters is taken into account, we observe that NV potential increases as the recirculation airflow rate increases. This is expected because the recirculation rate corresponds to the rate at which air is changing inside the zone, higher recirculation means better ventilation and filtration.
Deploying HEPA increases yearly percentage favorability of NV in New Delhi under the WHO Scenario-III D by up to 16% ( figure 4(A)) and the NAAQS Scenario-II D by up to 26% ( figure 4(B)) at the highest recirculation rate considered (10 ach). Without the use of air filters, yearly NV potential for the WHO Scenario-III D was 7% while yearly NV potential for the NAAQS Scenario-II D was 30%. On the other hand, deploying MERV 14 increases the yearly percentage of NV potential by up to 13% under WHO Scenario-III D and up to 19% under NAAQS Scenario-II D. High levels of ambient PM2.5 concentrations in New Delhi still limit the usage of NV, but indoor particle filtration could help in reducing the harmful exposure of PM2.5, especially for residents of informal settlements who rely heavily on NV to fulfill their cooling requirements.
In contrast, Bangalore, with its temperate climate and lower levels of PM2.5 concentrations, exhibited higher NV potential under both WHO-III D (31%) and NAAQS-II D (94%) Scenarios without air filters alone ( figure 3(B)). With the use of air filters, we observe from figure 5 that yearly NV potential for Bangalore increased up to 83% in the WHO Scenario-III D and 99% in the NAAQS Scenarios-II D for both MERV 14 and HEPA Filters. Hence, we observed that air filters help reduce the concentrations of indoor PM2.5 leading to better indoor environmental quality. Air filters with better efficiency and higher recirculation rates are more effective in combating the negative effects of air pollution. Figure 5. Increase in yearly percentage favorability of NV with the use of air filters corresponding to different efficiencies and recirculation rates for WHO Scenario-III D in Bangalore, as compared to without the use of air filters. NV potential is observed to increase with increase in filter efficiencies and recirculation rates. Note that the increase in percentage favorability is less than 5% for NAAQS Scenario-II D in Bangalore which would be hardly interpretable on the heat map, hence we do not have a corresponding figure for NAAQS Scenario.
Recirculation rates of up to 8 ACH are achievable, especially because residents in India rely heavily on ceiling fans [56]. Therefore, we choose air filters with a recirculation rate 8 ACH for further analysis. HEPA filters are more than twice as expensive as a MERV 14 filter-Average cost of a HEPA filter is $300 whereas average cost of a MERV 14 filter is $100 [57]. Since our target group in this study is low-income communities residing in informal settlements, we recommend using MERV 14 filters which offer 70% efficiency. We find that HEPA filters consume 40% more fan electric energy than MERV 14 filters but provide only an additional 6% increase in favorable NV hours. As a result, we base our deeper analysis of the results on assumptions of MERV 14 filter adoption and a recirculation rate of 8 ach. However, we also quantified uncertainty in NV potential for further results based on a lower bound corresponding to recirculation rate of 6 ACH, filter efficiency of 30% and a higher bound corresponding to recirculation rate of 10 ACH and filter efficiency of 99% (see figure 6 and supplementary material-tables 1-6).

Ventilation strategies should be localized based on climate and pollution patterns
Based on figure 2(A), we observe that outdoor PM2.5 levels are extremely high during the dry winter months of October-February in New Delhi. This is likely due to burning of firecrackers during the Indian festival of Diwali [7] and due to transport of pollutants during the crop residue burning season in late September [58] and/or the subsequent winter atmospheric conditions that trap air pollution at the ground level [17]. PM2.5, owing to its extremely small size, can remain suspended in air for a long time, especially due to the cold surface of earth and low wind circulation during the winter months [6]. Consequently, we find the NV feasibility in New Delhi varies dramatically by season ( figure 6). Specifically, the feasibility for NV drops drastically in the winter months in New Delhi, even though NV potential for Baseline Scenario-I D is almost 100%, which demonstrates the impact of outdoor air pollution on NV Potential.
The climate is highly favorable for NV during winters as can be seen from the line corresponding to the heat index which varies from 19 to 25 • C during these winter months. For reference, low caution level for the heat index was defined as apparent temperature less than or equal to 32 • C. Despite good climatic conditions, potential for NV drops to an average of 40% for the NAAQS Scenario II-D during October-February, while it is less than 5% for the WHO Scenario-III D in New Delhi when indoor air filtration (MERV 14) is taken into account. Without the use of air filtration, feasibility for NV was found to be lower than 10% for the NAAQS Scenario-II D and almost negligible for the WHO Scenario-III D, on average. Extremely high outdoor PM2.5 levels, averaging over 250 µg m −3 is the primary reason for lower feasibility for NV in New Delhi during winters. But at the same time, low-cost air filtration techniques can reduce a building occupants' exposure to PM2.5, thus mitigating the harmful effects without sacrificing NV potential.  Month-wise percentage favorability of NV for Bangalore where heat index threshold is moderate for summer (March-October) and low for winter (November-February). Bangalore experiences smaller drops in NV potential for both NAAQS and WHO Scenarios as compared to the Baseline Scenario due to its favorable temperate climate and better outdoor air quality.
Improved outdoor air quality is observed in New Delhi for the summer months, even though average concentrations of PM2.5 are still above 60 µg m −3 . However, during summer, hotter climate contributes to lower favorability for NV as we can observe from the heat index line which ranges in between 30 and 48 deg C. NV potential drops from an average of 35% for the Baseline scenario-I D to 32% for the NAAQS-II D and 12% for the WHO scenario-III D during summer in New Delhi. Therefore, we found that extremely high outdoor PM2.5 levels directly influence NV potential during the thermally favorable winter months in New Delhi.
On the other hand, Bangalore, due to its favorable temperate climate and better outdoor air quality, experienced smaller drops in NV potential for both the NAAQS-II D and WHO-III D Scenarios when coupled with indoor air filtration, as shown in figure 7. The highest reduction in NV potential for WHO Scenario-III D, 45% (figure 7), was observed for the month of October (no drop was seen for NAAQS Scenario-II D). Similarly, as seen with the rise in New Delhi, this was possibly due to the firecrackers burnt during the festival of Diwali [7]. For the subsequent winter months, NV potential for NAAQS Scenario-II D remained the same as the Baseline Scenario-I D and an average reduction of less than 25% was observed for WHO Scenarios-III D after deploying indoor air filters. In the absence of air filters, Bangalore experienced a reduction of 30%-40% in NV potential during the winter months. As discussed earlier, low temperatures could restrict the circulation of air, trapping the air pollutants and contributing to air pollution. During the summer months, Bangalore's potential for NV is not affected, owing to both favorable climate and outdoor air quality especially when compared to New Delhi. Bangalore does not experience a drop in NV potential for the NAAQS Scenario-II D and less than 10% drop in the WHO Scenario-III D with the use of air filters (it did experience a drop of 20% during summer when air filters were not deployed). We conclude that outdoor PM2.5 levels and corresponding NV potential varies significantly across the summer and winter months. We also observed that average concentrations of outdoor PM2.5 reached their peak in the morning, decreased in the afternoon, and peaked again in the evening. This is possibly because anthropogenic activity increases as a result of traffic congestion during commuting hours [23]. To assess the influence of daily patterns of outdoor PM2.5 on NV, we analyzed daytime versus night-time favorability for NV for the entire year (figure 8). We found that daytime potential reduced more than night-time potential mostly across both NAAQS ( figure 8(B)) and WHO Scenarios (figure 8(C)) as compared to Baseline ( figure 8(A)), except for the winter months when outdoor PM2.5 levels are consistently greater than 150 µg m −3 throughout the day.
Finally, we analyzed intraday variations in PM2.5 and the heat index to provide insights into potential ventilation strategies. Specifically, we analyzed intraday variations of PM2.5 and heat index for a typical summer and winter day in New Delhi and assessed the feasibility of NV on those days. We observe from figure 9(A) that high PM2.5 levels are clearly the reason for infeasible NV potential on a typical winter day to protect the occupants from harmful exposure to air pollution. Heat index levels are well within the acceptable thresholds. On the other hand, during the summer, high temperatures drive the unfavorable conditions for NV especially during daytime as heat index levels greater than 39 • C are extremely unsafe for the occupants as shown in figure 9(B). PM2.5 levels during summer are observed to be much lower compared to the winter months. Therefore, our results reiterate that strategies for ventilation need to be tailored depending on regional, seasonal, and daily variations of PM2.5.

Energy impact of lost NV potential due to air pollution
Given the need to balance occupant comfort and health, we modeled a hybrid approach that utilizes NV when air quality conditions are acceptable and mechanical systems with filtration otherwise. Adding a mechanical ventilation system incurs an additional energy load that we classify as the 'energy penalty' incurred due to air pollution of the outdoor air. Typically, informal settlement dwellers have access to electricity to cover basic energy needs, such as ceiling fans, as they rely on such fans in addition to NV. Apart from few instances of load shedding, the electricity is available consistently [59]. However, it is difficult for occupants of these settlements to afford additional electricity but there is ongoing research on slum rehabilitation which is looking into the feasibility of low-cost mechanical ventilation systems, such as air conditioners in addition to providing adequate housing [60]. While this approach is 'academic' , we believe that this study is a critical first step that will help policymakers understand how much the cooling/heating requirements are met and how much additional energy will be required to provide thermal comfort to occupants.
In order to assess the energy penalty, we utilized the Ideal Loads Air System module in EnergyPlus to model the mechanical system. We then calculated comfortable indoor temperatures according to the Indian Model of Adaptive Comfort [54] and used them as setpoints for the thermostat. Finally, we calculated the energy penalty incurred by occupants as a trade-off for NV when mechanical systems restrict unacceptable exposure to unsafe PM2.5 levels outdoors. We quantify the energy penalty as the extra energy used by the mechanical systems in NAAQS and WHO Scenarios, as compared to the Baseline Scenario which does not consider air pollution as the limiting factor for NV and is only based on zone apparent temperatures (as discussed in Methods section and table 2). Figure 10 depicts yearly energy usage per building area for the three operational scenarios for NV in New Delhi and Bangalore. As shown in figure 10, Bangalore has lower energy consumption than New Delhi owing to the temperate climate in Bangalore and warm/humid climate in New Delhi. We note that filter fan energy consumption is already included in the values indicated in figure 10.
In New Delhi, we observed an increase in energy usage of 6% in NAAQS scenario from Baseline (60 kWh m −2 in NAAQS from 56.7 kWh m −2 in baseline) and 27% in WHO Scenario from Baseline (72 kWh m −2 in WHO from 56.7 kWh m −2 in Baseline) with the deployment of indoor air particle filtration. Similarly, in Bangalore, we observed an energy penalty of 0.5% for the NAAQS Scenario and 25% for the WHO Scenario, when compared to the Baseline Scenario individually. This is because the threshold of acceptable PM2.5 levels reduces and becomes stricter as we transition from Baseline to NAAQS to WHO Scenarios. Without the use of air filters, the energy penalty was much higher for New Delhi (up to 9%) and Bangalore (up to 13%). Overall, we observed that outdoor PM2.5 levels can have a big impact on the NV potential.
Furthermore, we modeled a simple scenario in which informal settlement dwellers are coerced to adopt potentially unaffordable full mechanical air conditioning. Under the NAAQS standard and an assumed setpoint of 26 degrees Celsius, we found that in New Delhi energy use intensity increased from 60 to 173.5 kWh m −2 . This further underscores the potential of low-cost filtration as a viable interim low-carbon solution pathway as it provides both thermal comfort and exposure protection at 65% less energy intensity than adoption of full mechanical air conditioning.

Discussion
NV has proved to be one of the most affordable and sustainable solutions for conserving energy in buildings and improving overall comfort of the occupants. However, utilizing NV in dense urban (residential) environments with extremely high outdoor PM2.5 concentrations is a challenge. This paper represents a novel analysis of the effect of outdoor PM2.5 on feasibility of NV in a subcontinent where PM2.5 is abundant. Studying the impact of air pollution on NV is particularly important for highly vulnerable dwellers of informal settlements in which 800 million+ people reside worldwide.
Our results demonstrate that filtration-aided NV increases the favorable hours for NV and protects occupants from harmful exposure to PM2.5 at a lower energy penalty and lower capital/operational costs than a full mechanical air conditioning system, which is particularly important for the low-income population being studied in this paper. The outdoor PM2.5 levels can also vary significantly across locations, months, and hours of the day. Therefore, there is a need for finding ventilation strategies tailored to regional climate and locations to optimize for occupant health and thermal comfort while conserving energy through the use of NV. In the context of informal settlements, air filters could be attached with low-cost mechanical ventilation systems, such as exhaust fans or portable table fans, typically available in such settlements to facilitate the implementation of these tailored strategies. We also quantified the subsequent energy penalty associated with mechanical ventilation when filtration is added to reduce pollution exposure and found that a modest 6% increase in energy usage can provide 1,343 additional safe NV potential hours using the Indian NAAQS standard in highly polluted New Delhi. Although this approach is 'academic' for current informal settlements, there is ongoing research on exploring the feasibility of low-cost air-conditioners for these settlements as part of slum rehabilitation programs [60].
One of the most related studies is by Ji et al [40] which also used building energy simulation to investigate the reduction in feasibility of NV for a typical apartment building in Beijing, China. Beijing is one of the most polluted cities in China, with average outdoor PM2.5 concentrations recorded to be 80 µg m −3 . Feasibility for NV was observed to be reduced by almost 55% annually to adhere to acceptable air quality standards of China. When coupled with indoor air filtration, the exposure to PM2.5 could be reduced sufficiently to allow for NV almost throughout the year. However, in India, where average outdoor PM2.5 concentrations is as high as ∼200 µg m −3 for New Delhi, NV coupled with air filtration could provide acceptable indoor air quality for up to 60% of the year. In contrast, for Bangalore, which records average outdoor PM2.5 levels around 60 µg m −3 , air filtration is sufficient to maintain adequate indoor air quality throughout the year. Therefore, our results are consistent with existing studies which found that NV potential is affected drastically for places with high outdoor pollution, and indoor air filtration coupled with NV could help reduce the harmful exposure to air pollutants significantly.
In addition to yearly analysis of impact of outdoor PM2.5 on NV, it is important to consider the seasonal and hourly patterns of PM concentrations. The worst PM2.5 pollution is observed in winter, followed by spring and summer. This is mostly likely due to weaker circulation of wind during winters which forces small pollutants like PM2.5 to stay in the air for a longer time. In fact, high concentrations of PM2.5 could coincide with favorable seasons for NV, especially for countries with a tropical climate like India. Apart from the seasonal pattern in PM2.5 levels, we also noticed hourly patterns-PM2.5 typically peaks during the morning, declines during afternoon, and goes up in the evenings again. The morning and evening peaks could be representative of traffic patterns during typical commuting hours. These observed patterns of PM2.5 concentrations are consistent with previous studies [6,34] which depicted similar trends across seasons and similar peaks on a daily basis. Hence, intraday dynamics of pollution, regional climate and study location all play an integral role in aiding occupants and building operators to decide whether to utilize NV by opening or closing windows.

Limitations and future work
Our model quantifies the impact outdoor PM2.5 has on NV potential when accounting for occupant comfort and overall well-being. We focused our analysis on outdoor sources of PM2.5 as indoor PM2.5 data is not readily available and varies significantly across households. Future work aims to collect indoor PM2.5 data for typical residential buildings and consider their effect on NV potential and understand balancing outdoor fresh air intake and indoor pollution. Additionally, while we focused on PM2.5 as the only significant source of outdoor air pollution and we were constrained by data availability for other sources of outdoor air pollution, future studies are required to take the other sources into account such as PM10, nitrogen oxides, and ozone, to better estimate feasibility for NV. Since low-cost outdoor air quality data are used from secondary sources in our study, future work aims to work with air pollution modelers to improve the accuracy of such data inputs in future iterations of our analysis. We also aim to study the role of noise and security in estimating NV potential. Finally, while this study does not consider how/when occupants open and close windows, future work aims to utilize new and existing data sources to study the effects of different environmental and air quality parameters on occupant window opening behavior and incorporate them into our analysis.

Conclusions
This study quantified the impact of outdoor air pollution (specifically, PM2.5) on the potential for NV in informal settlements across two major cities in India (New Delhi, Bangalore). The average population in India is exposed to extremely high levels of average outdoor PM2.5 concentrations (greater than 100 µg m −3 ) and is highly reliant on NV for cooling purposes-hence, we studied the influence of ambient air pollution on NV, specifically for the vulnerable populations in informal settlements that make up the majority of the country's population. We combined publicly available air quality data with a building energy model and a differential equation airflow model and took into account indoor air filtration. We established three operational scenarios for NV (considering the influence of heat index and varying thresholds of air pollution concentrations) to assess the NV potential for each city.
We observed that NV potential drops when the outdoor PM2.5 levels are high, and that the drop is very sharp for the highly polluted city of New Delhi. Also, high concentration of air pollutants coincided with favorable winter months of NV for New Delhi. However, filtration-aided NV has the potential to significantly increase the number of hours favorable for NV. Acceptable indoor air quality and thermal comfort could be maintained up to 60% of the time (annually) for New Delhi and almost throughout the year for the less polluted city of Bangalore. Furthermore, we also find that NV potential varies significantly across summer and winter months for both New Delhi and Bangalore. Low-cost and higher efficiency filters, attached with portable table fans or exhaust fans, can provide a pathway for mitigating the harmful effects of PM2.5 by reducing the occupants' exposure, especially under higher ventilation rates. Finally, we also quantified the energy penalty incurred for adoption of low-cost filtration and found that it results in modest 6% increase in energy use intensity and yields 25% more favorable NV hours in highly polluted New Delhi. While in Bangalore, low-cost filtration contributes to only 0.5% increase in energy use intensity, resulting in favorable NV hours for 99% of the year. We also found that low-cost filtration remains a viable interim low-carbon solution pathway as it provides both thermal comfort and exposure protection at 65% less energy intensity than adoption of potentially unaffordable full mechanical air conditioning. Our work demonstrated that assessing NV potential based on spatiotemporal analysis of outdoor PM2.5 data can help building operators to implement better season-specific ventilation strategies that can both enhance thermal comfort and protect residents from hazardous air pollution.
Overall, this work contributes to the growing body of literature on understanding the interdependencies between occupant thermal comfort, air pollution exposure, and building energy usage in urban areas across the world with a particular focus on residents of informal settlements. Our results could help inform low-carbon and safe cooling in-situ retrofit and rehabilitation strategies being developed by building designers, architects, and policymakers in Dharavi (e.g. Maharashtra Housing and Area Development Authority (MHADA) and Slum Rehabilitation Authority (SRA)) and across the world. This work can also guide current and future policies on NV and provide a direction for reforming building codes based on outdoor air quality on a broad scale without the need for performing detailed energy simulations. In the end, addressing the energy needs, comfort and pollution exposure for informal settlement and other low-income dwellers is critical for an environmentally just transition towards a low-carbon energy future.

Data availability statement
The data that support the findings of this study are openly available at the following URL/DOI: https://doi. org/10.5281/zenodo.7563318.