Inter-room pollutant transmission routes in naturally ventilated street canyon buildings

Airborne transmission of pollutants is a significant concern in urban areas, especially since the recent pandemic. While previous studies have explored inter-room pollutant crossflow in highrise buildings and isolated scenarios, the ventilation characteristics within street canyon configurations have not received enough attention. Understanding ventilation characteristics in this layout will help limit the spread of pollutants along the street. This study investigates the wind-driven transmission characteristics of pollutants between the indoor rooms situated adjacent to a street canyon, subjected to perpendicular approach flow. Numerical simulations were performed using the 3D steady Reynolds-Averaged Navier-Stokes approach and validated against surface pressure measurements inside a wind tunnel. The impacts of seven building ventilation strategies were compared, each representing a window opening combination on the two buildings. Based on the set environmental conditions, the results indicate that the inter-room transmission is dependent on both the ventilation efficiency and the specific location of the room. Increasing the ventilation rate decreases the cross transmission between the rooms, while rooms at the centre of the canyon have an increased susceptibility of being contaminated. For instance, switching from cross to single-sided ventilation decreased the average air change rate from 42.3h -1 to 1.4h -1 , coinciding with a rise in pollutant concentration by 114-fold and 12-fold at the canyon ’ s centre and side, respectively. Rooms at the edge of the canyon were more contagious and could contaminate other rooms located further inside the street, albeit in small doses. Our findings present new insights within the domain of indoor-outdoor air interaction, shedding light on the interplay between room location and potential inter-unit pollutant transmission. In addition, the pressure measurement results from the wind tunnel experiment provide a valuable resource for other researchers aiming to validate numerical simulations for building engineering applications


Introduction
Air quality plays an important role in affecting the health and mortality rate of a population.The Global Burden of Disease [1] attributed about 6.67 million deaths in 2019 to air pollution alone, accounting for about 11-12% of all annual deaths.The huge economic and wellbeing impacts of polluted air have been a matter of policy discussions and urban management goals, including the Sustainable Development Goals (SDG) [2].It has been argued that air quality should be treated as a fundamental right of individuals [3].Epidemiological studies have repeatedly emphasised its importance and the need to find solutions for a better quality of life and longer life expectancy [4][5][6].A range of air pollutants, including CO, NOx, SOx, etc., can induce primary and secondary health impairments in humans and animals [7,8], as well as exacerbate underlying medical conditions [9][10][11].Therefore, buildings and indoor spaces should be designed to limit pollution and pathogen levels, for a safe and healthy environment.
In urban areas, pollutants can spread between different indoor rooms (or units) and pose significant health risks.Several epidemiological studies have highlighted the airborne transmission of pathogens between apartments which has led to the spread of diseases [12][13][14].Similarly, indoor generated air pollutants from activities such as cooking, heating, mechanical equipment, smoking, construction, etc. [15][16][17], can be transported to other units through airborne routes.In the case of naturally ventilated spaces, the risks can be higher if the incoming air is not filtered.While the nature of infection can vary depending on the microorganism and dosage, it is important to understand the pathways of transmission to limit the inter-unit spread of pathogens as well as other pollutants.
Considering the associated risks from inter-unit cross-transmission, several studies have been conducted to identify possible pathways in buildings and design suitable interventions.Mu et al. [18][19][20] conducted a series of wind tunnel tests to identify the role of wind direction and source location on the inter-unit transmission for an isolated highrise building.Similarly, Liu et al. [21] conducted a decoupled simulation to estimate the risks of cross-transmission in an isolated highrise building.Their results demonstrated that a windward-located pollutant source can intensify the pollution concentrations in adjacent flats.Wang et al. [22] performed simulations to understand the role of window type on the cross-transmission of pollutants in an isolated building and showed that an awning window offers the most protection against the transmission of contaminants.More recently, Huang et al. [23], developed a numerical model to estimate the cross-transmission risk in highrise buildings by measuring the concentrations of a tracer gas in an outdoor environment.Several other studies have explored the effect of building geometry and grouping to assess inter-unit transmission [24][25][26][27][28][29].However, most work in this domain has been limited to the assessment of highrise buildings.In contrast, studies on cross-transmission of pollutants in street canyons, representing an important and ubiquitous urban feature, remain few.
Generally, studies on pollution transmission in street canyons have been limited to outdoor pollution dispersion [30][31][32][33][34][35][36][37][38][39] or transmission of outdoor air pollutants into indoor spaces [40][41][42][43][44].A few studies on inter-unit transmission in street canyons have been conducted in the last couple of years.Dai et al. [45,46] conducted field experiments, using a scaled model of a street canyon, to study the effect of wind and temperature on inter-unit transmission.Rows of concrete blocks were placed to form street canyons, and several model buildings with indoor units placed at the centre of the streets.By measuring the tracer gas concentrations, they observed that the concentration of transmitted pollutants was inversely related to the air change rate (ACH).Transmission primarily occurred in vertically downward direction from sources on the windward façade.Likewise, transmission from leeward façade was vertically upward.The authors associated this to the presence of a stable vortex inside the street canyon.The transmission across the street canyon into rooms located on the opposite buildings was much lower, although at some instances they observed a spike in the pollutant concentration.In fact, they found that the re-entry ratio was higher for a room located further downwind than in the immediate vicinity.While the study makes some important contribution in the field of inter-unit dispersion, a generalised conclusion is difficult to draw and is not easily replicable due to transient nature of the outdoor meteorology.A wind tunnel investigation by Cui et al. [47] attempted to understand the dispersion characteristics of hot smoke released from the mid-height of a building adjacent to a street canyon.They highlighted that wind speed, buoyancy and height of the source unit are critical factors which determine whether the released pollutants will recirculate inside the canyon or get washed downwind.They concluded that the released pollutant was primarily transmitted vertically upwards when released on the windward façade, contrary to the observations by Dai et al. [46].Finally, a recent work by Zheng et al. [48] explores the effect of ventilation strategies in hospital dormitories on the inter-unit transmission of pathogens.The work briefly discusses the effect of aspect ratio of street canyon on dispersion of pathogens.However, further work is warranted to understand the role of ventilation in street canyons and the inter-unit transmission routes.
With the advent of improved computational facilities, several studies on the ventilation of indoor spaces have been conducted [41,49,50].However, one of the major challenges with numerical tools is their reliability and replicability, as they need to be validated against experimental observations.In numerous instances, researchers have adopted a mixed-validation strategy wherein the model being studied has geometric and spatial distinctions from the validation model assessed experimentally [20,[51][52][53][54][55].For example, Yang et al. [50] and Wang et al. [56] simulated multistorey buildings integrated with indoor units, and validated their models using the WT tests conducted by Jiang et al. [57], -which analysed air flow inside an isolated cube.Wu and Niu [58] simulated an isolated multistorey building with indoor units but validated their model using experimental results by Perino and Heiselberg [59], which measured airflow and ventilation inside a chamber.Similarly, Ai and Mak [60] validated their multistorey urban canyon model with indoor units against experimental results by Dascalaki et al. [61] which measured airflow and ventilation rate inside a chamber.While the authors made efforts to ensure accuracy in their results, it should be noted that the simulation domain under investigation differs significantly from the experimental model, potentially altering the physics of the airflow.Therefore, it is possible that the numerical results do not accurately reflect the real flow characteristics.
Noting the scarcity of literature on cross-contamination of indoor spaces adjacent to street canyons, this study aims to provide a comprehensive assessment of the effect of natural ventilation mode on indoor air quality and transmission.A CFD model is developed using Ansys FLUENT ® and validated against two wind tunnel tests: (i) building internal pressure (conducted as part of this study) and (ii) outdoor street canyon pollutant dispersion (obtained from literature).The model was used to simulate various window opening strategies on two parallel buildings forming a street canyon subjected to a perpendicular approach flow.Instead of relying on outdoor measurements to estimate indoor pollution levels, the current study adopts a coupled approach by integrating both indoor and outdoor domains and introducing a source of pollutant inside each individual room.This will help understand the dispersion and airborne routes for inter-unit contamination along street canyons and identify critical scenarios.The novelty of the present work is as follows: • A 1:100 scale model of the street canyon integrated with indoor units is built and tested inside an atmospheric boundary layer (ABL) wind tunnel to understand dominant indoor pressure trends.The results can be used by other researchers to validate their numerical model.• A 3D CFD model is developed and validated to understand transmission routes from one apartment (unit) to other apartments along the length and height of the street.The building is compartmentalised into rooms on three different floors.

Wind tunnel setup for indoor airflow
A series of experiments were conducted inside an atmospheric boundary layer wind tunnel at the Department of Civil Engineering, University of Nottingham, to measure the air pressure inside an indoor space adjacent to a street canyon.The test section is 2.4 m wide by 1.8 m high and has a fetch of 11.5 m to the front of a 2 m diameter circular turntable.The experimental model is based on the earlier works by Gromke et al. [62] and Cui et al. [47].A combination of fence and spires at the inlet, along with cuboidal roughness elements along the fetch, is used to generate the required ABL profile.An image of the canyon model and the test section is shown in Fig. 1, along with the wind profile measured at the centre of the empty turntable.
A reduced-scale model of the street canyon, with a ratio of 1:100, was placed on the turntable, perpendicular to the wind direction.The 1.2 m long model street canyon was formed by two parallel building blocks of height H = 120 mm and depth B = 120 mm.The model blocks were placed apart creating a street of width W = 120 mm.A three-storey indoor domain was made of 2 mm thick transparent sheets and placed at the centre of each of the blocks.Internal dimensions of the rooms measured H × W × D = 37 mm × 62 mm × 116 mm, with a 12 mm square window cut on the external faces of the room.In order to assess internal pressure levels, 15 pressure taps were placed across the internal wall of each room, and evenly arranged in a grid of 20 mm horizontal and 9.5 mm vertical.Vinyl tubes were used to connect the taps to a pressure transducer, a Scannivalve MPS4264, capable of measuring up to 995.4 Pa at ±0.2% accuracy.The reference pressure was measured by placing a pitot-static tube 1.1 m upstream of the model.

Wind tunnel setup for pollutant dispersion
The experimental data for pollutant dispersion in the outdoor domain of street canyons was accessed from the CODASC database [63].The physical dimensions of the scaled street canyon model are the same as described in the previous section.However, no indoor rooms were placed in this experiment and only the outdoor dispersion of a tracer gas was analysed.Four line-sources of pollutants were introduced in the canyon at the street level and a tracer gas was released from them.The mean outdoor concentrations of the tracer gas were measured at the canyon walls and normalised according to Eqn. (1) for line-source emission [64].
where C is the measured mean concentration, u ref is the wind speed at reference height H and Q/l represents the emission strength per unit length of the source.Fig. 2 shows the experimental setup at the Karlsruhe Institute of Technology and more information about same is available online [63].

Numerical setup 2.3.1. Computational domain and boundary conditions
The CFD model of the street canyon consists of two parallel building blocks, each of length L = 120 m and depth of B = 12 m, separated by a road of equal width, i.e., W = 12 m.The buildings are three storeys high, with a total height of H = 12 m, thus achieving an aspect ratio of unity (H/W = 1).The domain size was set based on the recommendations by COST Action C14 [65], summarised in Fig. 3.An inner domain with a refined grid was created around the canyon with dimensions 18H, 3H and 5H in width, height, and depth, respectively, following Salim et al. [66].This is done to ensure a more reliable prediction of flow variables with large gradients within the region of interest.
The computational mesh consisted of hexahedral elements mostly, with the smallest elements located inside the indoor space, measuring δx = δy = δz = 0.01H.The discretisation scheme ensured that at least 10 cells were generated across the length of any edge, as suggested by Franke et al. [65].The mesh consisted of 6.4 million cells after mesh sensitivity analysis.It was found that any further refinement beyond this did not produce any detectable improvement in the results.Instead, it led to higher computational demand and therefore, subsequent research was carried out using this mesh setting.Details of the mesh sensitivity analysis are presented in Appendix A. The inlet wind profile was identical for all simulation test cases.The mean wind speed was based on the fitted power law curve from the WT measurement (section 2.1) and is given by Eqn.(2).The turbulent kinetic energy and the dissipation rate were calculated using Eqns.( 3) and (4), respectively.
Fig. 2. The experimental setup for pollutant dispersion in a street canyon, image courtesy [63].
where the reference velocity u ref = 4.6 m/s, the reference height z ref = H, α is the power law exponent found by curve fitting the wind tunnel data (see Fig. 1(b)) and the empirical constant C μ = 0.09.The streamwise turbulence intensity, I, is the ratio of the standard deviation of the measured velocity to the mean wind velocity.Since the model canyon is a low-rise long building configuration, the contribution of wind-driven air flow is much more significant than other factors [67], and consequently, isothermal conditions were assumed in order to study inter-unit dispersion of pollutants.

Turbulence model and emission source
Turbulent flows with fluctuating velocity fields characterise airflow in urban areas.While the Reynolds-averaged Navier-Stokes (RANS) closure models govern the transport of the averaged flow quantities by modelling the entire range of turbulence scale [68], Large Eddy Simulation (LES) equations explicitly compute the large eddies.Thus, predictions of flow features are more accurate, including dispersion characteristics of pollutants [69].However, LES requires more computational resources and developing sub-grid-scale models is a challenge, including the specification of time-dependent boundary conditions [70].Nevertheless, RANS models, including the Reynolds Stress Model (RSM), have been extensively used in previous studies and validated to predict the dispersion of pollutants in 2D as well as 3D urban canyons [62,[71][72][73].Since the aim of this study was to understand the impact of ventilation choice on the bulk crossflow of pollutants between the various indoor units (or rooms), a knowledge of average pollutant exposure values within a space can provide sufficient information for appraisal.The performance of four different steady-state RANS models was tested, namely the standard k-ε, RNG k-ε, standard k-ω and the RSM turbulence model.The validation of the CFD model revealed the superior performance of the RSM model (section 2.4), hence the assessment of inter-unit dispersion was carried out with the RSM model.
Indoor pollutant concentration was estimated by solving the transport equations for passive tracer gas released at a constant rate of 0.182 x 10 -6 kg/s from point sources at the centre of the rooms.The dispersion of the pollutant was based on the standard steady-state advective-diffusion model, where the turbulent Schmidt number (Sc t ) determines the diffusivity.A range of Sc t values from 0.2 to 0.9 were tested to evaluate the performance, before Sc t = 0.3 was selected [44,[74][75][76].The transmitted pollutant concentration values were then normalised with respect to the source strength, reference height and velocity, given by Eqn.(5) for point-source emission [77].
where C is the average mass concentration of a tracer gas (kg/m 3 ) in the room, Q is the emission rate of the tracer gas and u ref is the reference velocity at the building height H.The SIMPLE scheme was used for pressure-velocity coupling in conjunction with the second-order upwind scheme to discretise the convective and diffusion terms.The enhanced wall function was used for the near-wall treatment [76,[78][79][80][81].

Validation 2.4.1. Validation of internal room flow and pressure field
The results from the wind tunnel measurement and CFD simulations were analysed in terms of pressure coefficient (C p ), given by Eqn. 6 Where p is the measured pressure, p ref is the reference pressure and ρ is the air density.The average C p values along the rooms' internal walls are presented in Table 1.There is under-pressure in all the rooms, indicated by the negative C p .When both windows are open, the flow resembles a jet through the room from window to window.This jet entrains air at its surroundings and consequently, a suction is established in the room (contour plots of C p are presented in Appendix B).However, the magnitude of the negative C p varies between the upstream and downstream rooms, with a higher negative C p in the downstream room, possibly due to the sheltering effect of the upstream building.WT observations indicate C p in the range of [-0.56 to -0.60] for the downstream rooms and C p in the range of [-0.25 to -0.26] for the upstream rooms.The downstream C p predictions by the CFD models are more or less uniform across the three rooms, although the values are generally overpredicted in comparison to the WT observation.The overall difference across the rooms is the least for the RSM model.In the case of the upstream rooms, all four turbulence models overpredict the surface pressure on the top rooms (UT).The standard k-ε model fairly replicates the internal surface pressure in rooms UL and UM.In contrast, the prediction by the RSM model only closely matches the top room UT.
After undertaking a meticulous scrutiny of the flow patterns in the upstream room (across a vertical plane), it was observed that the streamwise flow patterns predicted by the RSM were similar to those of other turbulence models.Specifically, an upward jet flow pattern was identified immediately after the inflowing air entered the window opening, propelling the flow towards the ceiling.Additionally, a recirculation zone was detected in the lower section of the room, which was observed across all turbulence models, including the RSM.This shows that the RSM performs similar to the other turbulence models in simulating internal air flow.One plausible explanation for the observed divergence in the C p values for rooms UL and UM could be attributed to the orientation of the inflowing air in the spanwise direction.Depending on the turbulence model used, the jet flow could be directed more prominently towards one side of the room, thus substantially affecting the C p values, particularly in light of the fact that the measurements were taken on only one side of the room during the experiment.Additionally, airflow near the upstream building is highly 3-dimensional characterised by impinging, separation and vortex shedding [82], making it difficult to predict precisely.
A validation metric was developed to further assess the quality of the simulations.The COST Action 732 [83] document describes some of the performance indicators that can be used to validate a numerical simulation.Namely, the correlation coefficient (R), the normalised mean squared error (NMSE), the fractional bias (FB), the fraction of predictions within a factor of two of the observations (FAC2) and the hit rate (HR) were deemed appropriate for this study.They have been calculated (for all 45 measurement points) and summarised in Table 2 for the different turbulence models.
In terms of R and FAC2, all models meet the acceptance criteria, suggesting that the deviations in C p values are within reasonable limits and that the general trend is as observed in the WT.For the NMSE and FB metrics, which puts more stress on the higher prediction values, the RNG k-ε and standard k-ω models did not meet the acceptance threshold.Meeting NMSE criteria ensures that random scatters are limited.Lastly, the HR was calculated with an absolute difference of 0.25 and a relative difference of 0.08, as indicated in the VDI guidelines [64].It can be clearly seen that only the RSM model meets the acceptance threshold for this metric, which basically assesses the fitness of the model for the investigated purpose [84], i.e., airflow through the indoor domain adjacent to a street canyon.While the RSM turbulence model deviates from the WT observation in the rooms UL and UM (Table 1), the overall metric in Table 2, calculated across both upstream and downstream rooms, outperforms over the other turbulence models.While there remains scope for further improvement in the accuracy of the CFD simulations, the current model performs satisfactorily well and can replicate WT observation at sufficient accuracy.Additional assessments of the numerical model should encompass validating the airflow patterns and air exchange within the room.It is imperative to acknowledge that despite the availability of hot wire anemometers to facilitate the measurement process, the dimensions of the model room proved too small to accommodate the hot wire sensor without causing interference to the airstream.As such, further experimentation is warranted to validate the numerical modelling, perhaps by employing optical measurement techniques.In the case of the downstream rooms, the RSM model provided the most accurate C p prediction across the three rooms.The result is consistent with the literature [62,85,86], which has demonstrated that the RSM is particularly suited for complex flows exhibiting strong anisotropy and recirculation regions.

Validation of outdoor wind flow and pollutant dispersion
For this assessment, the boundary condition imposed in the CFD was based on the CODASC experiment data set and the normalised pollutant concentration (C + ) values were calculated according to Eqn. (1) (for line-source).Fig. 4 shows the C + values extracted along the mid-height of the canyon-facing walls -A (leeward) and B (windward).The x-axis represents the normalised distance along the canyon, normalised with respect to the street length, where the centre of the canyon is at y/H = 0.The distribution indicates higher C + values on wall A as compared to wall B, accompanied by higher pollutant concentration near the centre of the canyon.The predictions by RNG k-ε and standard k-ω are generally higher than WT observations on both walls A and B, especially in the region y/H = [-2 to -1] is evident that the overall dispersion of pollutants is similar between the WT and CFD results, despite a few deviations.Notably, the CFD predictions show a slight drop in C + at the centre of the canyon, forming a local minimum.Similarly, CFD simulation underpredicts the concentration values near the edges of wall A. This is possibly because the computational model is unable to accurately replicate the interaction between the canyon vortex (with a horizontal axis) and the edge vortex (with a vertical axis) near the building corners.Pollutant dispersion near y/H = ±5 is more strongly governed by the edge vortex, developed as a result of separation around the leading corner of the upstream building.Simulation performance indicators, as described in the previous section, were calculated and are summarised in Table 3.Of the four turbulence models tested, all meet the acceptance criteria for correlation coefficient (R), NMSE, and FAC2.This suggests that the simulations can fairly replicate the outdoor flow field and dispersion characteristics, with limited random scatter.All of the turbulence models, except for RNG k-ε, meet the acceptance standards for FB.In terms of HR, the standard k-ε model meets the acceptance criteria while the RSM model nearly meets the acceptance threshold.Dispersion being turbulent in nature, predicting precise values is often challenging, leading to spatial variability [87].Nevertheless, the deviations were considered acceptable since the focus of the research was more towards indoor air quality and the effect of ventilation.We also compared the velocity prediction inside the canyon to assist us in the validation process (see appendix B, Fig. B2).Agreeably, the performance of RSM and standard k-ε was quite close [76], but the choice of RSM was substantiated as the predictions within the region of interest were most close to the experimental data, and the results from both validation studies demonstrated the superior performance of RSM.Therefore, the RSM model was implemented for assessing the seven ventilation cases.

Simulation test cases and ventilation efficiency estimation
To evaluate inter-unit transmission in street canyons, we created a model of internal units (or rooms) within the buildings, at three specific locations along the canyon: the centre and the edges.These locations were based on the results of earlier works, notably [30,62], which demonstrated that pollutant concentration is highest near the centre of the canyon and least near the edges, when the wind direction is perpendicular.Likewise, we created models of three-storey internal rooms located at the centre and edges of the street canyon.Fig. 6 shows the location of the rooms and the names assigned to them.Seven building ventilation scenarios involving a mixture of single-sided and cross-ventilation were investigated in this study.In each case, all rooms of a building were assigned the same window opening strategy.Fig. 7 describes the window opening strategy for each of the ventilation cases.In case 1, the upstream building is cross-ventilated, whereas the downstream building is ventilated from the leeward side.In case 2, the upstream building is ventilated from the windward side, and the downstream building is crossventilated.Cases 3 and 4 represent only single-sided ventilation, one away from the street and the other from the street, respectively.In case 5, both buildings are cross-ventilated.In cases 6 and 7 one building is cross-ventilated while the other is ventilated form the street.The tracer gas, ethane, was released from a point source located at the centre of the room, at a height of 1.5 m, representing average indoor emissions.
While there are multiple combinations of window opening strategies on the façade of a building, simulating all of them is nearly impossible due to high computational demand.We have, therefore, studied seven relevant combinations of ventilation strategies, with a special focus on configurations expected to exhibit extreme pollutant concentration.The study of these selected configurations provides valuable insights into the general transmission characteristics and can inform the development of more advanced computational models for predicting indoor air quality in real-world buildings.Therefore, while the current study is limited in the number of ventilation scenarios, the approach provides a basis for future research in the domain.
There are various indices to measure the ventilation efficiency of a room, including air change rate per hour (ACH) [88], age of air [89] and air exchange efficiency [90].The measure of the net flow of air required to remove all pollutants from a space is known as the purging flow rate (PFR), and it is an essential indicator for evaluating the ventilation efficiency of a local domain [91], such as the indoor rooms studied here.The PFR can also be understood as the equivalent rate at which fresh air is supplied to the domain [92].For this work, the ACH calculated by PFR is used as it offers a better comparison of ventilation efficiency between the simulated rooms [50].It is calculated by Eqn.(7); where Q is the emission strength of 0.182 x 10 -6 kg/s and is constant for all emission sources, C is the mass concentration of the tracer gas and V is the volume of the room.

Results and discussion
The results from the simulation were analysed based on two parameters: the normalised concentration of adventitious pollutants in a room and the ventilation efficiency of that room.The findings are presented in a tabular format, which shows the cross-transmitted pollutant between each pair of rooms.The columns represent the pollutant source while the rows indicate the target room.The pollutant concentration values were normalised using Eqn.(5).Since, the transmission from rooms at one end of the canyon to the rooms at the other end was negligible (K < 10 -10 ), they have not been included in the discussion.Additionally, both sets of side (or corner) rooms played, due to symmetric conditions, an equal role towards the contamination of central rooms, which allowed analysis of one set of corner rooms alone (DSL, UST etc., see Fig. 6).

Ventilation case 1
Table 4 shows the cross-transmitted pollutants between each pair of rooms for the ventilation case 1.In this mode, the upstream building is cross-ventilated, and the downstream building is ventilated from the leeward side.Pollutants are primarily accumulated in the downstream rooms, with contributions from all surrounding rooms.Vertical transmission of pollutants from the lower to upper rooms in the downstream building is the most critical route.The room DCM is severely contaminated with pollutants from DCL (K = 49.99),whereas DCT is majorly contaminated with pollutants from DCM (K = 43).A similar observation is made for the rooms at the side of the canyon (DSL, DSM and DST), but with lowered concentration values (K = 11.57and 9.19, respectively).In contrast, crossventilation in the upstream building produces negligible inter-unit transmission (K < 0.01).
The rooms on the upstream building are also accompanied by high ACH PFR , in the range of 40 h -1 to 80 h -1 , while the downstream building is poorly ventilated with ACH PFR ≈ 1 h -1 .However, it is important to note that the upstream rooms still contribute to the pollution levels in the downstream rooms, even if the inter-unit transmission is limited between them.Transmission from upstream to downstream rooms is more pronounced across-the-street as against along-the-street.This is illustrated in Table 4, where the deeper shade of yellow for across-the-street transmission of pollutants contrasts with the lighter shade for along-the-street transmission, with a difference in K value by two orders of magnitude.A further examination reveals that the strength of transmission from the side rooms on the upstream building (USL, USM and UST) is twice that of the central rooms (UCL, UCM and UCT).
M. Mohammadi et al.

Ventilation case 2
In the 2 nd ventilation case, the upstream building has openings on the windward side and the downstream building is cross ventilated.Table 5 summarises the inter-unit transmission of pollutants observed between each pair of rooms in this case.While this ventilation mode improves the airflow in the downstream rooms, it slightly deteriorates the air quality in the upstream building as compared to the previous configuration.Vertical transmission of pollutants is observed on the upstream building wherein the pollutants are transported from lower to upper rooms, particularly in the central rooms.Pollutants from UCL are accumulated in UCM and UCT.Similarly, pollutants from UCM are transported upwards into UCT.Interestingly, the transmission and accumulation of upstream pollutants into downstream rooms is somewhat reduced, especially near the centre of the canyon.Additionally, the airflow through the upstream rooms is significantly reduced, with an average ACH PFR of 4 h -1 , a reduction of nearly 90% from the previous case.
On the other hand, pollutant concentration is reduced in the downstream rooms, more significantly in the central rooms (DCL, DCM and DCU).In these rooms, ventilation efficiency is simultaneously increased, with an ACH PFR in the range of 10-40 h -1 .The greatest change is observed in the centrally located rooms (DCL, DCM and DCT), where the average ACH PFR increases from 1.1 h -1 (in case 1) to 46.4 h -1 .It is interesting to observe that the emission from DSL represents a major fraction of pollutant concentration in other downstream rooms, suggesting the influence of airflow around the canyon edges.It is possible that the superposition of the edge vortex and the canyon vortex, which results in a canyon-inward directed helical flow, transfers pollutants from DSL towards the central region.The same vortex also assists in the transportation of small amounts of pollutants from the upstream side rooms into the downstream side rooms.
M. Mohammadi et al.

Ventilation case 3
The 3 rd case represents single-sided ventilation on both buildings, in which all rooms are open to air exchange from the face opposite to the street canyon.The ventilation mode on the upstream building is the same as in case 2 while the downstream building follows case 1. Corollary, the major routes of inter-unit pollutant transmission are expected to be the same as those observed in previous instances.Pollutants from downstream rooms are primarily transported upwards into the upper rooms (see Table 6), such as from DCL to DCM and DCT.A small quantity of pollutants is transmitted and accumulated horizontally along-the-street into other downstream rooms (K < 0.1).As expected, there is no transmission of pollutants from downstream to upstream rooms (across-thestreet) since the windows on the downstream building face the downwind direction.The transmission pattern from upstream rooms generally stays the same as it did in case 2, albeit with slightly higher concentrations.In terms of ventilation efficiency, all rooms exhibit poor airflow with an ACH PFR in the range of 0.6-4.3h -1 ; although the upstream rooms show a slightly higher air change rate.

Ventilation case 4
This case represents single-sided ventilation, from the street side, for both buildings.It represents a scenario with low ACH PFR and high transmission, as shown in Table 7. Primarily, vertical transmission is observed for both sets of rooms, with an upward transmission in the upstream building and a downward transmission in the downstream building.For instance, the room UCT is heavily contaminated with pollutants from UCM (K = 101.9),and UCM is majorly contaminated with pollutants from UCL (K = 75.25).In contrast, pollutants from the room DCT are transported into DCM (K = 69.34),and pollutants from DCM are transported into DCL (K = 69.76).
Secondly, a strong horizontal transmission and accumulation of pollutants is observed.Pollutants from the side rooms are transported to central rooms in both buildings.Similarly, pollutants from downstream central rooms (DCL, DCM and DCT) are transported to upstream central rooms (UCL, UCM and UCT).It is worth noting, that the only routes along which little to no transmission occurs is from centre to side rooms (along-the-street).As can be seen by the four light-coloured blocks in Table 7.
In terms of ventilation efficiency, the ACH PFR is quite low for all rooms with an average value of less than 2 h -1 .In fact, the downstream building exhibits slightly better ventilation as compared to the upstream building.The average ACH PFR on the downstream building is 1.6 h -1 , while the upstream building has an average ACH PFR of 1.1 h -1 .This also relates to higher pollutant crosstransmission into upstream rooms as against downstream rooms.
M. Mohammadi et al.

Ventilation case 5
Case 5, representing cross-ventilation on both buildings, shows high ACH PFR and low transmission of pollutants (see Table 8).Opening windows on both sides of the building ensures good airflow through the rooms, with an average ACH PFR of 25 h -1 and 59 h -1 in the downstream and upstream rooms, respectively.While there is negligible transmission of pollutants in the upstream rooms, there is still some degree of cross-transmission of pollutants present in the downstream rooms.In contrast to the previous ventilation cases, no significant vertical transmission is observed in this case.Rather, pollutants are transported horizontally in the streamwise direction (across-the-street) and in the spanwise direction (along-the-street).Pollutants from the upstream side rooms (USL, USM and UST) are transmitted into downstream rooms across the street (DSL, DSM and DST) and also along the canyon into the centre (DCL, DCM and DCT), albeit in smaller measure.Similarly, the upstream central rooms also contribute to the pollution in rooms across the street (DCL, DCM and DCT) as well as along the street (DSL, DSM and DST).Vertical inter-unit transmission occurs only slightly (K = 0.2 ∼ 0.5) between the downstream rooms.For instance, pollutants from DSM are transported into both the lower and upper rooms, DSL and DST respectively.

Ventilation case 6
Case 6 represents a scenario when the upstream building is ventilated from the street side and the downstream building is cross ventilated.Primarily, inter-unit pollutant transmission occurs in the vertical direction near the upstream building (see Table 9).Pollutants released from the lower rooms are transported upwards, contaminating the rooms above.For instance, pollutants from the Table 9 Normalised pollutant concentrations (K) in case 6.
room USL are transmitted into USM (K = 36.08)and UST (K = 14.71), while pollutants from UCL are mainly transmitted and accumulated in UCM (K =K = 70.11)and UCT (K = 10.77).The other transmission route of concern is the streamwise (across-thestreet) flow of pollutants near the centre of the canyon.Assisted by the recirculating vortex, pollutants released from downstream central rooms (DCL, DCM and DCT) are transported into upstream central rooms (UCL, UCM and UCT) in the range K = [5,30].The upstream rooms also demonstrate low levels of ACH PFR (<2 h -1 ), being ventilated from the leeward side.In fact, central upstream rooms have even lower ACH PFR (<1.5 h -1 ), leading to higher transmission and accumulation of adventitious pollutants in them.
Additionally, pollutants from upstream side rooms are transported across-the-street into downstream rooms in lower concentrations (K < 1), however, accumulation of pollutants on the top floor can exceed K = 2.In the case of downstream building, vertical transmission of pollutants is not very prominent, and transmission occurs generally in all directions at a reduced concentration (K < 1), apart from the critical scenario near the centre of the canyon (K > 20).Finally, negligible transmission occurs from central to side rooms for both the upstream and downstream building.In terms of ventilation efficiency, the downstream rooms have ACH PFR ≈ 10 h -1 , and >30 h -1 for the side and central rooms, respectively.The higher air changes at the centre of the canyon also explain why the rooms, in that location, have lower concentrations of transmitted pollutants when compared to the side rooms.

Ventilation case 7
Ventilation case 7 represents a scenario where the upstream rooms are cross-ventilated and downstream rooms are ventilated from the street facing façade.Table 10 shows the normalised concentrations of transmitted pollutants.As is expected, transmission of pollutants into upstream rooms is negligible, due to the high ventilation rate through the rooms (ACH PFR > 38 h -1 ).Instead, pollutants from the upstream rooms are transported into downstream rooms, primarily across the street.For instance, pollutants from UCL are transported across the street into DCL (K = 46.71),DCM and DCT (K < 10).Simultaneously, small concentration of pollutants from upstream side rooms is transmitted and accumulated in downstream central rooms (K < 0.1).

Table 10
Normalised pollutant concentrations (K) in case 7. On the other hand, pollutants released in the downstream rooms are transmitted vertically, with higher pollutant accumulation in the central rooms.Interestingly, vertical transmission of pollutants from downstream side rooms has remained low for all ventilation scenarios described previously.The ventilation rate is also quite poor in the downstream rooms, especially in the central ones, which indicates that adventitious pollutants reside longer in them and, hence, explain the higher concentrations.

Comparison of extreme cases
Fig. 8 shows the relationship between ACH PFR and total pollutant concentration in each room.The size of the square indicates the air changes in the room, while the colour indicates the sum of all adventitious pollutant concentrations in the room.Note that this is different from the pollutant concentration values presented earlier, which was the concentration of pollutants from individual rooms.A large ACH PFR prevents build-up of pollutants and restricts cross-transmission and accumulation between the rooms, demonstrated by the yellow-to-orange shade of the large squares.Downstream rooms in cases 2 and 5, and upstream rooms in cases 1 and 5 record low levels of inter-unit transmission accompanied by the highest air changes.The inverse does not hold true for all cases, i.e., low air changes in a room do not necessarily lead to cross-transmission, although it is an important criterion for inter-unit transmission and accumulation of pollutants.Upstream rooms in case 4 demonstrate low ACH PFR and high transmission, while upstream rooms in case 2 demonstrate low ACH PFR and low pollution.In general, it can be said that apart from the source location, improving the ventilation efficiency in the indoor space can reduce the built-up of cross-transmitted pollutants, indicated by a correlation coefficient of -0.36 between the ACH PFR and the total adventitious pollutant concentration.Similar findings were also reported by Wang et al. [22], who showed that increasing the ventilation rate reduced the transmission of pollutants in a single-side ventilated multistorey building.Additionally, the range of predicted ACH, [0.4 h -1 , 7 h -1 ] and [9 h -1 , 76 h -1 ] for single-sided and cross ventilation scenarios, respectively, fall within the range of real-world observations [93].
Cases 4 and 5 represent the extreme ventilation modes, with the room-averaged total pollutant concentration of 61.6 and 1.6, respectively.We have so far discussed the inter-unit transmission for individual ventilation cases, however, to understand the broader aspect of inter-unit transmission and the impact of these two extreme ventilation scenarios, further discussion is necessary.Table 11 shows the contours of normalised pollutant concentration through the central rooms.The colour scale has been readjusted for clarity.
From Table 11 and it can be inferred that inter-unit transmission predominantly occurs in the single-sided ventilation mode (case 4).The presence of a strong vortex in the street canyon (in case 4) causes the recirculation of pollutants and their subsequent re-entry into other rooms, which can be visualised in the velocity streamline figure.The exhausted pollutants from the upstream rooms are pushed upwards, and the pollutants from downstream rooms are pushed downwards.Vertical transmission into adjacent rooms is in the range of K = [50 to 100] and represents the main inter-unit transmission route.The skimming airflow also transports some of the pollutants into the rooms across the street.Sources of pollutants located downstream cause a moderate degree of inter-unit transmission in the upstream rooms, as seen in the results presented in Table 11 (single-sided ventilation -DCL, DCM and DCT), with normalised pollutant concentration in the range of K = [10 to 50).The height of the source room is also critical.Lower rooms on upstream building and upper rooms on the downstream building are relatively more contagious, transmitting their pollutants into more rooms.For instance, the room DCT affects the air quality of all the rooms in the neighbourhood, as compared to DCL.
In contrast, transmission routes for upstream sources are primarily in the vertical direction, followed by low levels of transmission into rooms across the street (DCT, DCM and DCL) with K < 5.This can be explained by the interaction of the canyon vortex with the freestream airflow above the canyon, which dilutes the pollutant concentration.
The benefits of opening windows on opposite sides of a room are multiple, as demonstrated by the results of the cross-ventilation mode (case 5) in Table 11.This mode of ventilation enables crossflow of air across the rooms, leading to a reduction in crosstransmission and accumulation of pollutants in the neighbouring rooms as well as improved ACH PFR .The velocity streamline figure indicates that the air flows towards the street canyon as it passes through the rooms, resulting in minimal accumulation of crosstransmitted pollutants.As a result, the normalised concentration of any cross-transmitted pollutant remains below one (K < 1).This demonstrates the effectiveness of cross-ventilation in reducing inter-unit transmission of pollutants and improving indoor air quality.
Table 12 shows the contours of normalised pollutant concentration along the side rooms, along with the velocity streamlines.Crosstransmission of pollutants is generally lower here than at the centre of the canyon.The impact of downstream pollutant sources on the upstream air quality is reduced in this zone.On the contrary, upstream pollutants are more dispersed, polluting other rooms in the process.In fact, upstream sources of pollutants impact the air quality of neighbouring rooms in both the upstream and downstream buildings, regardless of the ventilation mode.
In the single-sided ventilation mode, the primary transmission route is in the vertical direction, aided by the recirculating vortex in the street.However, due to the presence of corner eddies, which aids in the dispersion of the pollutants, the inter-unit transmission in the streamwise direction (across the street) is much lower compared to the central rooms.Interestingly, when the building is crossventilated, transmission of pollutants occurs only in the room DST regardless of the source location in the upstream building.This is because airflow in room DST is from the street side rather than the leeward face as is the case in rooms DSM and DSL.The airflow in the street picks up pollutants from the canyon and carries them into the room DST.This can be visualised by the downward flow of air near the left window in DST (see velocity streamline diagram in Table 12).In contrast to the DST, the other downstream rooms (DSM and DSL) have a lower concentration of pollutants that are transmitted and accumulated into the space from the leeward face of the building.This occurs due to the presence of an additional recirculation zone in the wake of the building.The analysis so far has highlighted two routes of inter-unit pollutant transmission which impact indoor air quality.The first route is  Both these routes are impacted by the recirculating vortex in the canyon.However, the presence of corner eddies, which rotate around a vertical axis, introduces a third inter-unit transmission route which is less pronounced, but nevertheless crucial for building ventilation design.This transmission occurs along-the-street (spanwise transmission) and is driven by the corner eddies identified in Fig. 9, which shows the velocity contours and streamlines at the mid-height of the building.A similar flow structure has been identified by other studies, including Gromke and Ruck [94].
The streamlines indicate that the corner vortices drive air towards the centre of the canyon along with the released pollutants.In the case of single-sided ventilation, because the ventilation occurs from the street side, some of the pollutants are transmitted into the central rooms.This route of transmission is demonstrated in Table 7 as well, where rooms USL, USM and UST transmit pollutants into rooms UCL, UCM, UCT, DCL, DCM and DCT (K = [0.08 to 0.8]).Similarly, transmission from the rooms DSL, DSM and DST into central rooms is also observed.On the other hand, the risk of inter-unit transmission from the side to central rooms is much lower in the case of cross-ventilation as the flow field is highly 3-dimensional near the centre of the canyon (with K ≈ 0.03).Additionally, the airflow through the central rooms is primarily from the face opposite the canyon.
There are several implications to the observed transmission routes.Namely, the central rooms are more susceptible to contamination by adventitious pollutants, especially when ventilated from a single side.As expected, the recirculating vortices inside the canyon and at the edges accumulate pollutants near the centre of the canyon, increasing the risk of transmission into the rooms there.Fig. 10(a) explains this with the example of room DCT as a source, under the specific scenario of single-sided ventilation from street canyon.The released pollutants are largely transported to central rooms, below and across the street.On the contrary, the side rooms are more contagious than central room, in the sense that pollutants released from the side rooms are transported along the length of the canyon increasing the risk of inter-unit transmission in all rooms.While the actual pollutant concentration is low, it is nevertheless higher than the transmission from central to side rooms.Fig. 10(b) demonstrates the transmission routes from the room UST, under the influence of cross-ventilation.The choice of ventilation case 5 highlights that even with an improved air flow through the rooms, there is still some transmission and accumulation of pollutants from the side rooms into the central rooms.In fact, other ventilation cases show higher transmission along this route.

Conclusions and recommendations for future research
Pollutants originating from an indoor room can have a significant impact on the wellbeing of residents in neighbouring rooms/ apartments along the street.This study aimed to understand the inter-unit transmission of pollutants in an isolated street canyon under a perpendicular wind flow.A CFD model was created to simulate air flow and pollution dispersion from the internal rooms, which was validated against two experimental results.The study considered seven ventilation strategies, each representing a different window opening combination on the two buildings.Normalised pollutant concentrations inside the rooms and outdoor air flow were analysed.The main insights drawn from the results are discussed below.
• The window opening strategy was found to be instrumental in influencing the pattern of air change rates.Cross ventilation amplified the ACH PFR within the central rooms, whereas single-sided ventilation favoured those located at the street corners.Furthermore, the accumulation of transmitted pollutants was significantly affected by a room's location and air exchange rate, resulting in higher accumulation in the central rooms and those with low ACH PFR .Notably, a negative correlation coefficient (-0.3) exists between pollutant concentration and ACH PFR .• The downstream rooms were typically more susceptible to pollutant accumulation and inter-unit transmission, except for cases 4 and 6, where upstream rooms saw a greater accumulation due to the presence of an undisturbed vortex near the centre of the canyon.Pollutants from central rooms tend to accumulate within the central rooms, while those from side rooms spread along the length of the street, leading to inter-unit transmission in both central and side rooms, albeit at low concentrations.• Three routes of inter-unit pollutant transmission in street canyons were identified: vertical transmission, streamwise transmission (across the street) and spanwise transmission (along the street), listed in decreasing order of relevance.
Our findings present new insights within the domain of indoor-outdoor air interaction, shedding light on the interplay between room location and potential inter-unit pollutant transmission.In addition, the pressure measurement results from the wind tunnel experiment provide a valuable resource for other researchers aiming to validate numerical simulations for building engineering applications.The study was limited to the exploration of wind-driven transmission characteristics.Future investigations should consider a range of environmental and urban variables, such as building configurations and urban elements, including vegetation and vehicles.Follow-up investigations, utilising methods such as tracer gas analysis and transient simulations, promise to provide enhanced insights that can guide urban planners in devising targeted intervention strategies for ensuring healthier indoor environments.z values indicate a stronger vortex inside the canyon and consequently, higher pollutant concentration.While acknowledging that this may not precisely emulate real-world conditions, the heightened predictive capabilities of the model can be leveraged to introduce an enhanced factor of safety.

Fig. 1 .
Fig. 1.(a) The canyon model placed on the turntable with roughness elements; (b) Wind profile in the empty test section; (c) Close-up view of the indoor domain with three floors; (d) Pressure tubes connected to a side wall of the indoor domain; (e) Location of pressure taps inside the rooms.

Fig. 3 .
Fig. 3. Model configuration; (a) Domain size and boundary conditions, (b) Building configuration and location of indoor rooms, (c) Size of indoor room and with surface mesh.
the vertical transmission which transports the bulk of pollutants from the source room (K > 20), if the room is not cross-ventilated.The second route of transmission is across-the-street (streamwise transmission), with a reduced pollutant concentration (K = [0.2 to 20)).

Fig. 9 .
Fig. 9. Velocity contour and streamline along a horizontal plane at mid-height of the building (only half of the canyon is shown for clarity).

Fig. B1 .
Fig. B1.A comparison of contour plots of Cp.along the internal wall of the room Fig. B2 compares the normalised vertical component of velocity (u + z = u z /u ref ), at the centre of the street canyon, between the different turbulent models.The CFD contour plots reveal qualitatively similar flow fields, despite the higher prediction of u + z above the

Fig. B2 .
Fig. B2.A comparison of vertical velocity at the centre of the canyon for the street canyon without internal rooms

Table 1
Comparison of internal wall room average pressure coefficient between WT and CFD data, mean and standard deviation (in brackets).

Table 2
Validation metrics for the internal pressure coefficient.

Table 3
Validation metrics for the normalised pollutant concentration at the walls.

Table 11
Contours of normalised pollutant concentration (K) and velocity streamlines through the central rooms.

Table 12
Contours of normalised pollutant concentration (K) and velocity streamlines through the side rooms.