The model evaluation and sensitivity study is conducted around an Helsinki Region Environmental Services Authority (HSY) air quality monitoring site, hereafter referred to as the “supersite”, in Helsinki, Finland (60∘11′47′′ N, 24∘57′07′′ E). The site is located 3 km north–northeast from the Helsinki city centre, and it is characterised as an urban street-canyon curbside station with a traffic rate of around 28 000 on a workday, of which 12 % are heavy-duty vehicles . The street canyon is 42 m wide and the mean building height is around 19 m on the southwestern and 16 m on the northeastern side of the street see Fig. 1 in, resulting in a height-to-width ratio of 0.42. The supersite consists of a container (length 8.0 m, width 1.7 m and height 2.7 m) equipped with standard air quality measurement devices measuring from 4 m above ground level.

To get information about the spatial variability of air pollutants around the supersite, a 2-week measurement campaign was conducted in summer (6–16 June) and winter (28 November–11 December) 2017. During both campaigns, the horizontal distribution of air pollutants in the neighbourhood was monitored on non-rainy days using a mobile laboratory, and additionally during two intensive observation periods the vertical profiles of aerosol particles were measured using a drone.

The Sniffer mobile laboratory measured the horizontal distribution of trace gases and aerosol particle concentrations and size distribution. The measurements were done in 1–2 h slots with a 1 s temporal resolution during the morning and afternoon rush hours, around noon and in the late evening. During each observation period, the Sniffer was driving along a main street (Mäkelänkatu) and a side street as well as standing at the supersite, opposite the supersite and on a field 185 m from the main street (hereafter the “background”). The instrumentation of the Sniffer is given in Table  and the measurement locations in Fig. S1 in the Supplement. The main inlet was situated above the windshield at 2.4 m and a global positioning system (model GPS V, Garmin) recorded the van speed and position. For a detailed description of the Sniffer, see .

During the intensive observation periods, a multi-rotor drone (X8, VideoDrone Finland Ltd.) carried an electrical particle sensor (Partector, Naneos GmbH) to measure the vertical distribution of the alveolar lung-deposited surface area (LDSA) of aerosol particles, which describes the total aerosol surface area penetrating the deepest parts of the lungs see, e.g.and references within. The measurement were done on both sides of the street canyon when the Sniffer was simultaneously driving. The drone was flown 10 times up and down between z=2 and 50 m during one 30 min measurement interval, after which measurements were repeated on the other side. Each intensive observation period started by measuring LDSA at the supersite and ended on the other side. Measurements were started at 3 m from the building wall and the horizontal location was kept constant with a GPS sensor of the drone. Additionally, LDSA was measured at the supersite by a Pegasor AQ Urban sensor (Pegasor Ltd.) and on the other side by a DiSCmini (Testo Ltd.) or with another Partector at 1 m and in winter also at 14 m. For the details of the instrumentation, see .

The sensitivity of the results to the PALM model boundary conditions is examined during the following three periods: the morning (07:16–09:15 LT) and evening (20:26–21:14 LT) of 9 June, and the morning (07:20–09:14 LT) of 12 December.

In addition to the Sniffer and drone measurements, we use stationary aerosol observations from the supersite and two urban background monitoring sites: Kallio site operated by HSY and SMEAR III Station for Measuring Ecosystem Atmospheric relations; around 1.0 km southwest and 0.8 km northeast from the supersite, respectively (Fig. ). See Table S1 for the instrumentation. In addition to aerosol observations, meteorological data (wind speed, wind direction, air temperature) from the SMEAR III measurement tower (z=31 m) and Kivenlahti meteorological measurement mast 17.4 km west from the supersite are used in the study.

This study applies the PALM model system, version 6.0 (revision 4416) , which features an LES core for atmospheric and oceanic boundary layer flows. PALM solves the non-hydrostatic, filtered, incompressible Navier–Stokes equations of wind (u, v and w) and scalar variables (subgrid-scale turbulent kinetic energy e, potential temperature θ and specific humidity q) in Boussinesq-approximated form. PALM is especially suitable for complex urban areas, due to its features such as a Cartesian topography scheme and a plant canopy module, which are applied here to include the aerodynamic impact of both solid buildings and permeable vegetation on the flow. Furthermore, the so-called PALM-4U (short for PALM for urban applications) components have recently been implemented in PALM , including the SALSA aerosol module, the online chemistry module and the self-nesting and offline nesting features, which are all applied in this study.

SALSA describes an aerosol size distribution by a number of size bins (10 by default), and each bin can be composed of different chemical components. Chemical components included are sulfuric acid (H2SO4), organic carbon (OC), black carbon (BC), nitric acid (HNO3), ammonia (NH3), sea salt, dust and water (H2O). SALSA contains the following aerosol dynamic processes: coagulation, nucleation, dry deposition on solid surfaces and resolved-scale vegetation, and condensation and dissolutional growth by gaseous H2SO4, HNO3, NH3 and semi- and non-volatile organics (SVOC and NVOC). The gaseous compounds can be transferred to SALSA from the online chemistry module, which is based on the Kinetic Pre-Processor KPP; version 2.2.3 and an adapted version of the KP4 pre-processing tool . The implementation is flexible, allowing the user to choose the chemical mechanism and components being considered. In this study, a simplified mechanism describing photochemical smog is applied (see Sect. S3.2 in the Supplement). Photolysis is parameterised based on . However, the transfer of different organic vapours from the chemistry module to SALSA is still under development.

To capture the dominant turbulent eddies of the atmospheric boundary layer (ABL) in LES, the horizontal extent of the modelling domain should span over several ABL heights; see, e.g. . At the same time, to resolve most of the kinetic energy within street canyons, a high enough grid resolution (on the order of ∼ 1 m) is needed . Furthermore, uncertainty arising from the lateral boundary conditions usually decreases with increasing horizontal dimensions. To fulfill these contradicting requirements, a self-nesting feature has been included in PALM . In self-nesting, one or several child domains are nested within a parent domain and the child obtains its boundary conditions from its parent. Furthermore, PALM incorporates an automated mesoscale offline nesting with a mesoscale operational weather prediction model, which allows realistic, non-cyclic and non-stationary boundary conditions for the flow. As the mesoscale data do not contain resolved-scale turbulence, turbulence must first be developed within the PALM domain. To reduce the time and distance for the mesoscale flow field to adjust and turbulence to develop within the LES modelling domain, a synthetic turbulence generator within PALM can be applied.

The model simulations are conducted over a root domain of 6.9 km × 6.9 km, within which two smaller domains, parent and child, are nested progressively (Fig. ). The dimensions (Lx, Ly, Lz), number of grid points (Nx, Ny, Nz) and grid resolutions (Δx, Δy, Δz) of each domain are given in Table . In this study, the focus is on the child domain which matches with the area of the spatial aerosol measurements around the supersite.

Information on the building and vegetation height and land surface elevation are taken from high-resolution raster maps for Helsinki . The manipulation of the domain input files is done using the Python library P4UL . Only vegetation higher than zv,min=4.0 m is included in the simulations. Due to the lack of observational data on the leaf area density (LAD) of vegetation, a constant LAD value is applied for all tree crowns above zv,min. In summer, LAD=1.2 m2m-3 for broadleaf trees , while in winter LAD is decreased to 20 % of the summertime value.

We apply both modelled and observed data as meteorological boundary conditions, which are set dynamic; i.e. they change with time.

As modelled data, numerical weather prediction data from MetCoOp Ensemble Prediction System MEPS; are applied. MEPS data were downloaded from the data archive using the File Interpolation, Manipulation and EXtraction (Fimex) library . MEPS has a horizontal resolution of 2.5 km (see Fig. ), 65 vertical levels and 10 ensemble members. It ran four times daily with a 3-hourly cycling for data assimilation (3D-Var). The lateral boundary data are from the European Centre for Medium-Range Forecasts (ECMWF) high-resolution (HRES) atmospheric model. In this study, the MEPS control run, i.e. ensemble member 0 with unperturbed initial and lateral boundary conditions, is used.

Data from the Kivenlahti mast are downloaded from Finnish Meteorological Institute (FMI) Open Data service as 10 min averaged data. On the mast, meteorological observations are conducted at three to eight measurement levels between z=2-327 m. Despite being located 17.4 km from the supersite, the closest observations of the vertical profile of basic meteorological variables are conducted at Kivenlahti. Meteorological observations from the SMEAR III station at z=31 m are downloaded using the SmartSMEAR tool .

The initial conditions and dynamic meteorological boundary data are provided to PALM in a so-called dynamic driver. Of the MEPS data, the dynamic driver was created by the following procedure. First, the sigma coordinates were translated to pressure coordinates and further to height coordinates applying the hypsometric equation. Then, u, v, w, θ and water vapour mixing ratio qv were interpolated from the MEPS grid to the PALM grid: first in horizontal over a two-dimensional grid using the cubic spline method and then in vertical using the linear interpolation. The Kivenlahti mast observations, instead, were linearly interpolated in the vertical to the highest observation level, after which a constant value was used. When applying the SMEAR III data, constant values are used for the entire vertical profile. The dynamic driver created from the observational data does not include any horizontal variation.

A mesoscale interface, INIFOR, has been developed to transform mesoscale modelling data into PALM-readable boundary data. However, it is currently only available for COSMO-DE/D2 datasets, which do not cover Finland.

Similar to the meteorological boundary conditions (Sect. ), both modelled and observed air quality data are used as background concentrations in the simulations. As in the previous model evaluation study , the modelled background aerosol particle number and trace gas concentrations are produced with the trajectory model for Aerosol Dynamics, gas and particle phase CHEMistry and radiative transfer ADCHEM;. ADCHEM is operated as a one-dimensional column trajectory model along the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) air mass trajectories, starting 7 d backwards in time (see Figs. S2 and S3 in the Supplement). The gas and aerosol particle compositions and size distributions are simulated along the back trajectories arriving to the coordinates of the supersite. For the emission inventories and parameterisations applied, see Sect. S3.4 in the Supplement. Detailed descriptions of the aerosol and cloud microphysics, new particle formation and gas-phase chemistry mechanisms in ADCHEM are provided by and references therein.

To investigate the impact of the background aerosol size distribution (PSD) and concentration on the model simulations, PSD measurements from SMEAR III (see Sect. ) are applied as an alternative for the modelled values. For simplicity, ADCHEM data are always used for the chemical composition of aerosol particles and gaseous concentrations.

For each PALM simulation, the concentrations are averaged over the simulation time and these temporally constant vertical profiles are then introduced to the simulation domain by a decycling method, in which background concentrations are fixed at the lateral boundaries.

In this study, air pollutant emissions only from traffic combustion are included, as traffic is the main pollutant source within the modelling domain . Traffic-lane maps separating different road categories, i.e. main streets, collector roads and residential streets, have been generated by combining lane and street type information from the Map Service . The lane width is 3.5 m. Emissions are introduced as dynamic surface fluxes.

Aerosol particle emission inventories are typically provided as total mass emission factors EFPM. In SALSA, these would need to be translated to number emission factors EFN, assuming some size distribution for the emitted aerosol particles. However, converting aerosol mass to number is highly sensitive to the assumed size distribution. Therefore, in this study, we choose to apply a number emission factor EFN=4.22×1015 kgfuel-1 based on fuel consumption and a number size distribution estimated by at the supersite in May 2017 (see Fig. S4 in the Supplement).

For gaseous compounds, mass composition of aerosol particles and fuel, unit emission factors EF[compound] (Table ) are calculated using emission inventory by the European Environmental Agency for 2017 and specifically the tier 3 method, which applies information on the mileage per vehicle category and technology, and driving speed. However, since no information on the cumulative mileage for different Euro classes was available, EFNH3(g) and EFN2O(g) are based on the tier 1 method seeEq. 28. Furthermore, the following estimates were applied: EFSVOC(g)=0.01EFNMOG Fig. 2, EFRH(g)=0.4EFNMOG Fig. 4, where NMOG stands for non-methane organic gas and RH(g) for alkanes, and EFH2SO4(g)=0.1EFSO2 . Emitted aerosol particles smaller than 15 nm in diameter are assumed to be composed of 75 % OC and 25 % H2SO4, whereas larger particles contain 72 % BC, 21 % OC and 7 % H2SO4 as estimated from EFPM, EFBC and EFOC.

The hourly vehicle fleet compositions for the neighbourhood are obtained from the Helsinki Region Environmental Services Authority (HSY and Urban Environment Division of the City of Helsinki, personal communication, 1 October 2018), the mileage for each vehicle technology from the ALIISA model and the fuel sulfur content from the LIPASTO database . The traffic rates in the neighbourhood are estimated by normalising the mean traffic volumes per each street with traffic counts from an online traffic-monitoring station located in the northwestern corner of the child domain (City of Helsinki, personal communication, 3 March 2018). Traffic volumes for both southward and northward traffic are measured separately.

The length of the morning simulations on 9 June and 12 December is 2 h, and only 1 h for the evening simulation on 9 June. Simulation times correspond to the observation periods.

For all simulation times, two simulations using either modelled (M) or observed (O) boundary conditions for the flow and background aerosol particle number size distribution (PSD) are conducted. The first setup, hereafter MMETMPSD, applies the modelled meteorological (MET) boundary conditions from the MEPS data and the modelled PSD from the ADCHEM model. The second setup, hereafter OMETOPSD, applies the observed meteorological data from the Kivenlahti mast and the observed PSD from SMEAR III. Furthermore, two types of sensitivity tests are conducted for the summer morning. Firstly, model sensitivity on the background PSD is studied by running a simulation with the modelled meteorological boundary conditions and observed background PSD (MMETOPSD). Secondly, the influence of WD on pollutant dispersion is investigated by replacing WD in the MEPS data by WD measured on the Kivenlahti mast (OWD,mastOPSD) or at the SMEAR III station (OWD,SMEAROPSD). As WDSMEARIII is only measured at z=31 m, the wind direction at the model boundaries in OWD,SMEAROPSD is set constant with height. In total, nine different simulations have been conducted.

The aerosol and chemistry modules are run only within the child domain to limit computational costs. In all simulations, the aerosol processes of condensation and dissolutional growth, coagulation, dry deposition and sedimentation are included and calculated every 1.0 s. The aerosol particle size distribution is described by 10 size bins, of which three are within the first subrange (2.5–15 nm) and seven within the second subrange (15 nm–1 µm). Aerosol particles are assumed to be internally mixed and hygroscopic, and can contain H2SO4, OC, BC, HNO3 and/or NH3. The chemical reactions are calculated at every time step of the PALM model.

The advection of both momentum variables and scalars is based on the fifth-order advection scheme by together with a third-order Runge–Kutta time-stepping scheme . The pressure term in the prognostic equations for momentum is calculated using the iterative multigrid scheme . The roughness height is z0=0.05 m and the drag coefficient applied for the trees is CD=0.3.

Simulations were first run only for the root domain for 1 h, called the precursor run here, after which the final simulations were started. The final simulations including also the nested parent and child domains were initialised using the final state of the precursor run. Offline nesting is used as forcing for the root domain and the parent and child are nested within using one-way self nesting. As SALSA and chemistry are run only within the child domain, for them the nesting is not applied and the boundary conditions of air pollutants are set at the child boundaries. The data output was collected starting after the first 15 min of the final simulation. Simulations were performed on the Centre for Scientific Computing (CSC) Puhti supercluster. Using in total 394 Intel Xeon processor cores, each simulation required 39–80 h of computing time.

The following performance measures are applied in the evaluation: fractional bias (FB), normalised mean squared error (NMSE), factor of 2 (FAC2) , normalised mean bias factor (NMBF) and normalised mean absolute error factor (NMAEF) . See Table  for the acceptance criteria and Appendix  for the equations. In short, FB and NMBF measure systematic error (i.e. bias), NMSE and NMAEF both systematic and random errors, and FAC2 the correct concentration scales. Additionally, the statistical significance of the model error (i.e. the absolute difference between the observations and modelled values) is estimated with a Student's t test for the horizontal distributions of Ntot.

In order to compare the data, both the mobile Sniffer measurements containing its geographical coordinates and the PALM data output have been horizontally aggregated to a 5m×5m grid, with a threshold of at least three measurement points per grid to calculate the median value. A comparison between the measured and modelled median total aerosol particle number concentration (Ntot) values is illustrated in Fig. . In general, the model captures the large concentration gradient between the main street (in the middle from northwest to southeast) and the side street on the northeast side of the main street. However, the model overestimates Ntot at the northwestern end of the Sniffer route at all simulation times, which is likely due to overestimation of the traffic emissions at an adjacent cross section. Ntot is well modelled also along the side street, except during the winter morning in MMETMPSD (Fig. d–e and i–j), when the model slightly underestimates Ntot. The performance measures in simulating the horizontal distribution of Ntot and whether the acceptance criteria are fulfilled are shown in Fig.  for all simulation times. Overall, FB, NMSE and FAC2 show mostly acceptable model performance. However, NMBF often exceeds the acceptance criteria, showing that the model tends to over- or underestimate the observations by 25 % or more. NMAEF never fulfills the criteria, indicating that the absolute gross error between the observed and modelled values is always over 35 % larger than the mean observation.

During the first hour in the summer morning, MMETMPSD performs better than OMETOPSD with respect to all performance measures. OMETOPSD clearly overestimates Ntot along the main street (Fig. k) despite a stronger temperature inversion in the MEPS model data (MMETMPSD) compared to the Kivenlahti (OMETOPSD) observations (Fig. c). This likely stems from the underestimation of the wind speeds above 217 m in OMETOPSD (Fig. a), which would lead to lower mechanical turbulence production and to lower mean turbulent kinetic energy (TKE) in OMETOPSD (Fig. a). During the second hour, no large differences in Ntot are observed, as the wind speed and direction of the input data become more equal.

Instead in the summer evening, OMETOPSD performs slightly better than MMETMPSD (e.g. FAC2=0.75 and FAC2=0.67, respectively), but both acquire good performance values and even NMBF is within the acceptance criteria. This is surprising considering the clearly stronger winds in MEPS at z<200 m than what is observed on the Kivenlahti mast. Yet, MEPS predicts a more stable stratification, which leads to nearly equal TKE values (Fig. b). This can justify why the difference in the spatial variability of aerosol particle concentrations between MMETMPSD and OMETOPSD is not that large.

During the winter morning, MMETMPSD fulfills the acceptance criteria during the first hour, except for NMAEF, and performs better than OMETOPSD. However, during the second hour, the difference is small. Interestingly, FAC2 is higher for OMETOPSD than MMETMPSD over the whole simulation. Contrary to the summer evening, MEPS predicts clearly lower wind speeds in the winter morning, which would lead to weaker mixing, but at the same time the observed temperature inversion on the Kivenlahti mast is stronger than the modelled by MEPS especially during the first hours. Hence, the stronger stability and suppression of turbulence (Fig. c) can explain the higher concentrations in OMETOPSD. However, it should be noted that for the first hour the differences in the model absolute error between MMETMPSD and OMETOPSD are not significant, and for the second hour a Student's t test cannot be performed (see Table S2 in the Supplement).

The modelled vertical profile of alveolar LDSA is evaluated against the observed one over a 5m×5m area next to the supersite on the northern side of the container (Fig. ) and opposite the supersite on the other side of the main street (Fig. ). In the summer morning, MMETMPSD performs well opposite the supersite (Figs. a–b and ) especially during the first hour, but it clearly overestimates LDSA at the supersite (Figs. a–b and ). On the contrary, OMETOPSD successfully reproduces the LDSA profile at the supersite (Fig. f–g) but not opposite it (Fig. f–g). This can be explained by the wind direction: according to the meteorological boundary condition analysis in Sect. , the wind direction predicted by MEPS is more westerly than the one observed at Kivenlahti. Therefore, a canyon vortex forming in the main street canyon pushes pollutants upwind to the western side of the street in MMETMPSD. On the other hand, the opposite is observed for OMETOPSD for which the wind direction is more from the north. In the summer evening, MMETMPSD fulfills all acceptance criteria, while OMETOPSD overestimates LDSA on both sides of the street canyon. This is contradictory to the horizontal distribution of Ntot, based on which OMETOPSD showed better performance. In the winter morning, OMETOPSD shows slightly better performance at the supersite, whereas MMETMPSD underestimates LDSA, but it also has a lower background concentration. Instead, opposite the supersite, OMETOPSD clearly overestimates LDSA below the building height. For the summer evening and winter morning, the differences in the shape of the vertical LDSA profiles are not as notable as in the summer morning, which can be explained by the good correspondence of the Kivenlahti wind direction observations to the MEPS data.

Figure  illustrates the observed and modelled PSD for MMETMPSD separately along the main and side streets, at the supersite and opposite it, and in the background during the first hour of the summer morning on 9 June. In addition to the Sniffer measurements with EEPS and ELPI, the modelled values are compared against differential mobility particle sizer (DMPS) measurements at the supersite and SMEAR III. The model successfully reproduces PSD along the main street and specifically at the supersite, for which FB, NMSE and FAC2 are within the acceptance criteria (see Table S3 in the Supplement). Also in the background, the Sniffer measurements agree with the model based on FB, NMSE and FAC2 even though the concentration of the smallest (the mean bin diameter Dmid<25 nm) aerosol particles is underestimated. Instead, along the side street, the modelled values are clearly lower than the observed and, for instance, based on NMBF the model underestimates the EEPS and ELPI observations by a factor of 3.45 and 5.48, respectively.

Comparing the two simulations with different boundary conditions, MMETMPSD performs better along the main street and hence also at the supersite and opposite it during the first hour of the summer morning (Tables S3 and S4). However, during the second hour, the difference between MMETMPSD and OMETOPSD is minor, which was also observed for the horizontal distribution of Ntot. OMETOPSD, which uses the observed background PSD as boundary conditions, performs better along the side street and in the background. This is also observed in the summer evening (Tables S8 and S9) and winter morning (Tables S10 and S11). In the summer evening, both MMETMPSD and OMETOPSD perform equally well along the main street and equally poorly at the supersite overestimating the EEPS measurements by a factor of 3.68–4.36 based on NMBF (Tables S8 and S9). In the winter morning, MMETMPSD produces better results than OMETOPSD along the main street. Instead, at the supersite, both MMETMPSD and OMETOPSD perform equally well and fulfill the acceptance criteria for FB, NMSE and FAC2 (Tables S10 and S11).

The chemical composition of aerosols was measured at the supersite by an ACSM (aerosol chemical speciation monitor) and the black carbon (BC) concentration by a MAAP (multi-angle absorption photometer; see Table S1 in the Supplement). Additionally, the Sniffer measured BC within PM1. In general, the modelled and observed horizontal distributions of BC compare tolerably well based on the performance measures FB, NMSE and FAC2, while NMBF and NMAEF are not within the acceptance criteria (see Fig. S13 in the Supplement). Overall, the performance is best for MMETMPSD in the summer morning during the first hour. In the summer and winter mornings, OMETOPSD suffers from high positive bias and absolute error and MMETMPSD in the winter morning from high absolute error. Instead, in the summer evening, both simulations show NMSE and FAC2 within the acceptance criteria but still overestimate BC. Comparison to the measured chemical composition of aerosols at the supersite (Tables  and S12–13) shows that the modelled concentration of organic carbon (OC) is in general on the right order of magnitude. Furthermore, in the summer morning (Table ), especially the concentrations of sulfates (SO4-) and nitrates (NO3-), and also BC and PM2.5 are correctly reproduced by OMETOPSD, while ammonium (NH4+) is highly overestimated especially by MMETMPSD (NMBF=18.05). In the summer evening (Table S12), MMETMPSD corresponds better to observations than OMETOPSD and correctly reproduces SO4-, OC and PM2.5, while overestimating the rest. Also in the winter morning (Table S13), MMETMPSD performs slightly better than OMETOPSD in modelling OC and PM2.5 on the right order of magnitude, but MMETMPSD still overestimates the mass concentrations of the other chemical components by a factor of around 2.7–6.5. Hence, the difference in the chemical composition is not systematic. Comparing modelled values with point observations in a street canyon is very sensitive to the correct wind direction because a perpendicular wind component leads to accumulation of pollutants on the leeward side of the street canyon. As the vertical dispersion of LDSA was also shown sensitive to the wind direction, the results of the performance of modelling the correct chemical composition correspond to those of the vertical dispersion of LDSA (see Sect. ).

Sensitivity of the modelled aerosol concentrations to the background PSD is investigated by applying the modelled PSD (MMETMPSD) from ADCHEM and the observed PSD at SMEAR III (MMETOPSD) as the background PSD. Regarding all variables used in the evaluation (Ntot, LDSA, PSD and aerosol chemical composition), only minor differences are observed between MMETMPSD and MMETOPSD. For instance, for the horizontal distribution of Ntot, FB=-0.02 and FB=0.01, and NMSE=1.17 and NMSE=1.15, respectively (see Fig. ). The difference in the horizontal distribution of Ntot (Fig. b) is mainly within -20–20 % with slightly higher (lower) concentrations in the southern (northern) part of the domain. This difference stems from roughly 80 % lower observed than modelled background Ntot, and thus the air being advected from northwest is cleaner in MMETOPSD compared to MMETMPSD.

Similar to the horizontal distribution, the vertical profile of LDSA for MMETOPSD does not differ from MMETMPSD within the street canyon (Fig. ). Only a small decrease in model performance is observed opposite the supersite when applying the observed PSD as the boundary condition (e.g. FB is increased from 0.20 to 0.34 and NMSE from 0.07 to 0.14 during the first hour; Fig. ). However, above z>30 m, the difference gradually approaches 65 %–160 %, i.e. the relative difference in the background Ntot between the modelled ADCHEM values and SMEAR III observations.

With respect to PSD, MMETMPSD and MMETOPSD perform mostly equally well or poorly, except for slightly better performance of MMETOPSD on the side street. The wind speed and/or direction influence the modelled PSD more than the background PSD (see Fig. S14 in the Supplement).

In Sect. , the simulation using the observed data as boundary conditions (OMETOPSD) was shown to perform worse than when using the modelled MEPS data (MMETMPSD) in the summer morning. As the observed wind speed at Kivenlahti and the one modelled by MEPS differ (see Sect. ), we separately investigate the model sensitivity to the incoming wind direction.

In general, OMETOPSD and OWD,mastOPSD, for which the meteorological boundary conditions are taken from MEPS but the incoming wind direction is replaced with the one measured on the Kivenlahti mast, result in a similar pattern for the difference in the horizontal distribution of Ntot compared to MMETMPSD (Fig. a, c). However, the differences are larger for OMETOPSD, for which the incoming upper-level wind speed is up to 2 ms-1 slower than in OWD,mastOPSD during the first hour (Fig. a). This results in the aerosol particles being transported more to the southwest side of the main street. As the wind is more from the north in OMETOPSD than in OWD,mastOPSD, also the impact of wind direction on the street-canyon vortex along the main street is observed by clearly lower (higher) Ntot on the southern (northern) side of the street canyon. The similar patterns of ΔNtot for OMETOPSD and OWD,mastOPSD but the higher absolute values of ΔNtot for OMETOPSD indicate that the horizontal distribution is strongly controlled by the wind direction, while the absolute values depend on the wind speed. Instead, OWD,SMEAROPSD, for which the wind direction is from SMEAR III and does not vary with height, shows clearly higher concentrations (up to +150 %) along the main street but smaller differences in its surroundings (-40–80 %).

Replacing the modelled wind direction with the one observed on the Kivenlahti mast (OWD,mastOPSD) or SMEAR III (OWD,SMEAROPSD) improves the model performance for the horizontal distribution of Ntot during the first summer morning hour (Fig. ). However, for the second hour, OWD,SMEAROPSD is shown to perform even worse than OMETOPSD based on a lower FAC2 and higher NMBF and NMAEF. During the first hour, OWD,SMEAROPSD performs better than OWD,mastOPSD based on FB, NMBF and NMAEF, indicating that there is more bias in OWD,mastOPSD, while during the second hour OWD,SMEAROPSD performs better only based on NMSE. Presumably, the wind has too much of a westerly component in OWD,SMEAROPSD compared to the northerly winds in the MEPS and Kivenlahti data, which results in the traffic emissions downstream being flushed along the main street.

The observed vertical profile of LDSA at the supersite on the summer morning corresponds better to the one modelled by OMETOPSD and OWD,mastOPSD than MMETMPSD (Fig. ). Hence, modifying the MEPS wind direction to correspond to the observed one at Kivenlahti increases the model performance. During the second hour, OWD,mastOPSD performs even slightly better than OMETOPSD (e.g. FB=-0.07 and FB=-0.09, and NMSE=0.01 and NMSE=0.03). However, applying the wind direction from SMEAR III improves the model performance only for the first hour, while during the second hour OWD,SMEAROPSD performs the worst. Whereas OWD,mastOPSD results in lower (higher) LDSA than MMETMPSD below (above) z=30-40 m, OWD,SMEAROPSD shows higher values up to the building height (z=19 m) during the second hour, above which it gradually starts to follow the ΔLDSA for OWD,mastOPSD. Opposite the supersite, up to 5-fold values compared to MMETMPSD are observed in OMETOPSD and OWD,mastOPSD within the first hour (Fig. b), and the model performance is clearly decreased when using the observed wind direction from Kivenlahti (e.g. FB is increased from 0.20 to 0.47 and FAC2 decreased from 1.00 to 0.67; Fig. ). Instead, OWD,SMEAROPSD performs better than MMETMPSD. Within the second hour, ΔLDSA compared to MMETMPSD is mainly within ±50 % below the building height, but above OWD,SMEAROPSD deviates from the other profiles showing 2-fold values. All ΔLDSA profiles gradually approach ∼100 %, which is a result of using different PSDs for MMETMPSD and for the rest of the simulations.

Similarly for the aerosol size distribution, changing the modelled wind direction by MEPS to the one measured at Kivenlahti improves model performance in the background and slightly decreases elsewhere during the first hour, whereas during the last hour PSD is modelled better along the side street and at the supersite (see Tables S3 and S6 in the Supplement). Applying the wind direction from SMEAR III generally does not improve model performance (Table S7). Along the main street, the difference in PSD is mainly governed by the emission, which is shown by the peak at Dmid≈30 nm in Fig. S14 (in the Supplement), while in the background ΔN follows the difference in the boundary conditions for aerosol particles.