Introduction

Despite the fact that levels of atmospheric pollutants have fallen over the last few years, urban air pollution continues to be the largest environmental health risk in Europe (EEA, 2022). In order to meet the EU standard regulated (Air Quality Directive, 2008/50/EC), the National Emissions Ceiling Directive (NEC, 2016/2884/EU) was established to commit to European countries reducing their annual pollutant emissions. In this context, the 1st Spanish National Air Pollution Control Programme (NAPCP) (Ministerio para la Transición Ecológica 2019) was developed including a set of existing (WEM scenario) and additional emission reduction measures (WAM scenario). The NAPCP is composed of 50 measures for eight emission sectors: (1) energy mix (8 measures); (2) transport: road transport, rail, aviation and ship** (6 measures); (3) improved energy efficiency in the industrial and manufacturing sectors (3 measures); (4) improved energy efficiency (7 measures); (5) waste management (8 measurements); (6) use of fertiliser plans (9 measures); (7) reduction in emissions from burning of pruning (2 measures); and (8) manure and housing management for cattle, pigs and poultry (7 measures). A detailed description can be found in Vivanco et al. (2021). The impacts of these emission scenarios projected for 2030 on air quality and human health were investigated at the national level using mesoscale modelling by Vivanco et al. (2021) and Gamarra et al. (2021a), respectively. In addition, the impact of selected individual measures from NAPCP on air quality and human health was studied at the national level by Gamarra et al. (2021b).

However, air quality assessment and estimating population exposure in urban areas is challenging due to the strong heterogeneities of pollutant distribution within the city (Santiago et al. 2021, 2022a; Woodward et al. 2023). The interactions between the atmosphere and urban surfaces produce complex airflow patterns and reduced ventilation in the streets (Buccolieri and Hang 2019). This fact is linked to traffic emissions, one of the most important pollutant sources in cities, induce high levels of pollution (mainly NO2) and very large spatial variability of air pollutant concentrations in the streets (Vardoulakis et al. 2011; Amorim et al. 2013; Vos et al. 2013; Gromke and Blocken 2015; Borge et al. 2016; Beauchamp et al. 2018; Santiago et al. 2020). Hence, the spatial representativeness of urban air quality monitoring stations (AQMS) is very limited (Santiago et al. 2013; Martín et al. 2015; Kracht et al. 2018), and high spatial resolution is necessary to capture those strong gradients of concentrations. Computational fluid dynamic (CFD) modelling is able to explicitly resolve the airflow and pollutant dispersion around buildings and trees at high spatial resolution, being an adequate tool for assessing air quality in real urban environments (Kwak et al. 2015; Jeanjean et al. 2017; Sanchez et al. 2017; Borge et al. 2018; Rivas et al. 2019). The main disadvantage of these models is the large computational cost required. This fact limits the size of the simulated computational domain and the simulated time, and in most cases, it is unfeasible to carry out unsteady simulations of long time periods (e.g. one year to obtain annual mean concentrations). However, several numerical methodologies based on CFD simulations have been developed to obtain annual mean concentration maps (needed to evaluate the exceedances of EU standards) at high spatial resolutions (Parra et al. 2010; Solazzo et al. 2011; Vranckx et al. 2015; Santiago et al. 2017a, b; Rivas et al. 2019; Sanchez et al. 2017; Reiminger et al. 2020; Rafael et al. 2021).

In this context, the impacts of WEM and WAM emission scenarios from 1st Spanish NAPCPs on NO2 concentrations at street level were investigated by Santiago et al. (2022b) in three neighbourhoods of Madrid (Spain). That study showed that the annual mean limit value could be exceeded in some areas within the neighbourhood (dimension of a mesoscale modelling cell) in spite of the spatially averaged NO2 concentration being below the limit value. Therefore, to evaluate the effects of mitigation measures on air quality and population exposure, it is important to consider the spatial variability of NO2 concentration within the districts in an urban environment. In this context, the present study is focused on individual measures related to the urban environment, more specifically, the implementation of electric cars and efficient boilers. Taking into account these previous results, several research questions arise.

Is the impact of those individual measures projected to 2030 from the 1st Spanish NAPCP on street-level NO2 concentrations enough to obtain concentrations below the EU annual limit values? And how much additional traffic emission reductions are required to not exceed the limit value for NO2 concentrations?

Hence, this paper aims to estimate the effects of specific emission reduction measures related to urban environments (the implementation of electric cars and efficient boilers) projected to 2030 on street-level NO2 concentrations in a real neighbourhood and to estimate the exceedances of the EU annual limit value. Mesoscale and CFD modelling are used to evaluate their effectiveness at high spatial resolution. Moreover, additional local traffic emission reductions are simulated to estimate the reductions needed to ensure the annual mean limit value not being exceeded, not only in terms of the spatially averaged concentration in the study zone but also in some areas at the microscale within a neighbourhood.

The study area is described in the ‘Study area’ section. The modelling setup and the methodology used to estimate annual NO2 concentration maps are presented in the ‘Methodology’ section. The ‘Description of scenarios’ section shows the emission scenarios proposed. Results are presented in the ‘Results’ section. Model evaluation is carried out in the ‘Model evaluation’ section. The effects of emission scenarios from individual measures related to the introduction of more efficient boilers and electric cars are presented in the ‘Impact of individual measures (ECOBOIL and EC) and the combination of ECOBOIL + EC’ section, and the impact of additional local restrictions related to electric cars are shown in the ‘EC scenarios’ section. Finally, the summary and the main conclusions from the study are presented in the ‘Summary and conclusions’ section.

Study area

The study area is located around the Escuelas Aguirre (EA) air quality monitoring station (AQMS) in the central area of Madrid city, close to El Retiro park (Fig. 1). EA-AQMS is placed in a garden near the intersection of two avenues with intense traffic (orange point in Fig. 1b,c). The annual mean NO2 concentration measured by this monitoring station (58 µg m−3) exceeded the EU annual mean limit value for NO2 (40 µg m−3) for the 2017 year (the reference year for this study) (Madrid City Council, 2017), and every year, usually, it is one of the stations with the highest annual mean values measured in Madrid (Madrid City Council, 2016, 2017, 2018, 2019, 2020, 2021, 2022).

Fig. 1
figure 1

a General location of the neighbourhood within Madrid, Spain. b Study area around Escuelas Aguirre (EA) AQMS. c Numerical domain including vegetation (in green) and mean daily traffic. d Details of computational mesh. Orange point shows the location of AQMS and LAT, and LON indicates the latitude and longitude of the AQMS

Full size image

The study area covers a large green area (the north part of El Retiro park) and several streets and avenues with intense traffic (Fig. 1c) including many buildings of different heights. The tallest building is approximately 90-m high, but most of the buildings range from 18- to 24-m high. NO2 concentrations in the streets are strongly influenced by local traffic emissions, while the urban background concentration is the main contribution inside the park. All these characteristics make it interesting to focus this study on this neighbourhood. Air quality in this neighbourhood was previously investigated and modelled in several studies. Long-term-averaged concentrations at high spatial resolution for two different periods of around 3 weeks were computed through a CFD-based approach by Santiago et al. (2017a). These results fitted satisfactorily the passive sampler data recorded during two different experimental campaigns. In addition, annual NO2 concentration maps were computed by Santiago et al. (2022b) for this neighbourhood for the 2016 year and for two scenarios of emission reductions projected to 2030 from the 1st Spanish NAPCP (WEM and WAM scenarios). The present study is focused on the effectiveness of specific measures for pollutant emission reductions related to urban areas (the implementation of electric cars and efficient boilers). In this case, the 2017 year is taken as a reference for the emission reductions of the individual measures studied.

Methodology

The impact of the emission reduction measures is evaluated in terms of annual mean NO2 concentrations. Hence, maps of annual mean NO2 concentrations at a pedestrian level for the different emission scenarios need to be computed at high spatial resolution. These kinds of maps can be estimated through CFD modelling; however, the high computational cost required makes it, in most cases, unfeasible to carry out unsteady CFD simulations over a year. To address this issue, a numerical methodology (weighted average CFD-RANS, WA CFD-RANS) based on a set of steady CFD simulations of non-reactive pollutant dispersion for different meteorological conditions (16 wind directions) is used (Parra et al. 2010; Santiago et al. 2013, 2022b; Santiago et al. 2017a, b; Sanchez et al. 2017; Rivas et al. 2019). Following this methodology, the annual mean NOx concentration maps are estimated by means of a weighted average of the steady-state CFD simulations (described in detail below). Similar methodologies were successfully applied in different urban environments. Finally, NO2 concentrations are estimated from NOx concentrations using the NO2/NOx ratio at each hour computed by mesoscale simulations at ground level in the mesoscale cell corresponding to the microscale domain (Santiago et al. 2022b). Hourly meteorological conditions and background concentrations used in the numerical methodology (WA CFD-RANS) are computed by mesoscale modelling. In the next sections, the CFD modelling and the numerical methodology used to estimate the annual mean concentrations are described in detail.

Description of CFD modelling

CFD simulations (performed with STAR-CCM + code) are based on Reynolds-averaged Navier–Stokes (RANS) equations with Realisable k-ε, where k is the turbulent kinetic energy and ε is its dissipation rate. Since the park induces non-negligible dynamical effects on pollutant dispersion, the aerodynamic effects of vegetation are modelled assuming trees as a porous medium. It is modelled by adding a term of sink of momentum to the momentum equations and sink and source terms to the k and ε equations. More details about the vegetation model used in the present study can be found in Santiago et al. (2013), Santiago et al. (2017b) and Santiago et al. (2022c). The formulation is based on Green (1992), Liu et al. (1996) and Sanz (2003), and similar approaches were also used by Mochida et al. (2008), Amorim et al. (2013) or Gromke and Blocken (2015). Note that the vegetation model was tested in these previous studies. Inside the park, a vegetation canopy of 18-m high (average height of trees) is modelled, but for street trees, only the crown is considered. NOx dispersion is simulated by means of a transport equation for a non-reactive pollutant.

The distance between buildings and boundaries of the computational domain should have no influence on wind flow patterns obtained by CFD simulations, and the size of the numerical domain is determined considering the best practice guidelines of COST action 732 (Franke et al. 2007). The building area is 700 m × 800 m approximately, the distance between the buildings and the lateral boundaries is greater than 8H, and the height of the domain is greater than 6H, where H is the height of the tallest building. In order to minimise the possible overestimation of wind at the entrance of the building area, the results are analysed on a smaller area of 600 m × 500 m. The domain is discretized using an irregular mesh of 3·106 computational cells including prism layers close to the ground and buildings with a resolution ranging from 1 to 3 m. Each building and street is resolved with at least 10 grid cells. Figure 1d shows details of the mesh. A test of the grid independence of the results was carried out, showing that this resolution is appropriate (Santiago et al. 2013).

Symmetry boundary (all fluxes are 0) conditions are imposed at the top of the domain, and buildings and ground are modelled using standard wall functions. Ground is considered a roughness wall with a roughness length (z0 = 0.05 m); 16 wind directions (N, NNE, NE and so on) are simulated, imposing wind profiles with different angles at the inlet boundaries of the numerical domain. Neutral inlet vertical profiles of wind speed (u), turbulent kinetic energy (k) and ε are used (Eqs. (1)–(3)):

$$u\left(z\right)=\frac{{u}_{*}}{\kappa }ln\left(\frac{z+{z}_{0}}{{z}_{0}}\right)$$
(1)
$$TKE=\frac{{u}_{*}^{2}}{\sqrt{{C}_{\mu }}}$$
(2)
$$\varepsilon =\frac{{u}_{*}^{3}}{\kappa \left(z+{z}_{0}\right)}$$
(3)

where κ is von Karman’s constant (0.4), \({u}_{*}\) is the friction velocity and Cµ is a model constant (0.09). These profiles are widely used by CFD simulations over urban environments (Richards and Hoxey 1993; Buccolieri et al. 2011; Santiago et al. 2017a). For all these simulations, the friction velocity is set as \({u}_{*}\) = 0.24 m s−1 indicating a wind speed logarithmic profile at the inlet with 3.2 m s−1 at 10 m. NOx emissions are modelled through a source term in the transport equation of NOx. They are located on the roads from the ground to 1-m high to consider the initial dispersion and distribution, taking into account the mean daily traffic count of each street (Santiago et al. 2017a). It is assumed that the traffic emissions are distributed following the traffic count of each street. In the numerical methodology, the temporal evolution of traffic emissions is considered using typical profiles of traffic for Madrid (Quaassdorff et al. 2016; Sanchez et al. 2017), and the traffic emission factors are modified in order to provide the same annual average NOx concentration as that recorded at the EA-AQMS in 2017 (Santiago et al. 2017a, 2022b). Only traffic NOx emissions (which are clearly dominant in this heavily trafficated area) are considered in CFD simulations, and contributions of other sources outside of the domain are included through the background concentration provided by mesoscale simulations. Hence, the emissions from boilers are only considered in the mesoscale simulations.

Methodology for estimating annual averageNO2concentrations

For the computation of NO2 concentration maps, firstly NOx concentrations are computed by using WA CFD-RANS methodology. This methodology can only be applied to non-reactive pollutants; hence, it is applied to NOx (NO + NO2), whose behaviour is similar to that of a non-reactive pollutant due to the fast conversion between NO and NO2 (Sanchez et al. 2016). In this study, NO2 concentrations are estimated from NOx concentrations using the NO2/NOx ratio at each hour computed by mesoscale air quality simulations at ground level in the mesoscale cell corresponding to the microscale domain (Santiago et al. 2022b). The annual mean NOx concentration map is computed by means of a weighted average of the steady CFD simulations. For this computation, hourly meteorological conditions and background concentrations are needed. In this case, mesoscale simulations are used. Air quality at the mesoscale is simulated by the CHIMERE chemistry transport model (Menut et al. 2013). The model was applied to a domain covering the Iberian Peninsula at a spatial resolution of 0.1° × 0.1° nested within a European domain at 0.15° × 0.15°. The National Official Emission Inventory for 2017 provided by the Ministry for the Ecological Transition and Demographic Challenge (METDC) was used, elaborated at a 0.1° × 0.1° grid resolution (the same as the EMEP grid) for subsnap categories on an annual basis. The temporal evolution of these emissions was estimated by applying seasonal, weekly and hourly profiles. CHIMERE has been extensively evaluated in Spain (Vivanco et al. 2009), and the performance is comparable to that of other air quality models applied in Europe (Bessagnet et al., 2016). The meteorological fields were adapted from simulations at the European Centre for Medium-Range Weather Forecasts, ECMWF (www.ecmwf.int (accessed date: 10 January 2022)), known as the Integrated Forecasting System (IFS), for 2017, and obtained from the MARS archive at ECMWF through the access provided by Spanish State Meteorological Agency (AEMET) for research projects. Specifically, 3 hourly data from the HRES (Atmospheric Model high resolution) model with a horizontal resolution of approximately 9 km × 9 km and 137 vertical levels were used, interpolated to hourly values as used in CHIMERE. More details can be found in Vivanco et al. (2023) and Gamarra et al. (2021b).

The main steps of the WA CFD-RANS methodology are the following.

  1. 1)

    For each hour, the selection of CFD simulation (NOx_CFD) from the set of simulations according to the wind direction (WD) provided by the IFS meteorological mesoscale simulation.

  2. 2)

    For each hour, modification of NOx concentrations according to the temporal evolution of traffic emissions depending on the type of day, hour and month (TE(t)). Moreover, an additional factor is used for scenarios with reductions in traffic emissions.

  3. 3)

    For each hour, modification of NOx concentrations considering the ratio between the reference velocity used in the CFD simulations (uref(CFD)) and the reference velocity obtained from mesoscale simulation for that hour (uref(meso,t)). The methodology is based on the assumption that, for neutral inlet profiles, the concentrations due to traffic emissions inside the CFD domain are inversely proportional to wind speed. At each hour, this ratio of friction velocities is used to transform the simulated concentrations by CFD into the modelled concentrations consistent with the atmospheric conditions at that hour. The friction velocity from the mesoscale model provides information about the momentum flux considering wind speed and temperature. Hence, the ratio of friction velocities indirectly includes the effects of other physical processes such as thermal effects included in the calculation of the friction velocity in the mesoscale model. Sanchez et al. (2017) found better agreement with experimental measurements using the friction velocity (\({u}_{*}\)) as reference velocity in the methodology than using the wind speed at a certain height as reference velocity. Hence, the use of \({u}_{*}\) as reference velocity is adopted in this study.

  4. 4)

    For each hour, the background concentration is added to concentrations computed by CFD simulations. Background concentrations are computed by interpolating the vertical concentration profiles of the mesoscale air quality simulations at 1.5H, being H the height of the tallest building (90 m in this case). The first vertical layer of CHIMERE is from 0 m to approximately 24 m above the ground. It is assumed that the contribution of the surface emissions from the studied domain is low enough at this height to consider the pollutant concentration at such a level represents the background concentration coming from other sources beyond the domain boundaries. This assumption minimises the double counting of traffic emissions in both models (mesoscale and microscale) (Santiago et al. 2022b).

  5. 5)

    Finally, hourly concentration maps are averaged to obtain the annual mean concentration maps.

One of the advantages of using this methodology is that NOx and NO2 maps for the different traffic emission scenarios are obtained without any additional CFD run. Hence, the number of scenarios investigated can be large.

Description of scenarios

The base case is the scenario with the pollutant emissions for the 2017 year. Two individual measures of emission reductions included in the NAPCP and projected for 2030 are investigated. The first one is the ECOBOIL scenario, which consists of the substitution of the current boilers for more efficient boilers for all fuel mixes. For the ECOBOIL scenario, emission reductions are compared with domestic heating sector emissions for 2017. Since CFD simulations only consider traffic emissions, the emission reductions for this scenario are applied to the domestic heating sector in the mesoscale simulations. Therefore, the impact of this individual measure in the CFD simulations is due to the decrease in background concentration. The second scenario is the EC scenario, which corresponds to the projected substitution of passenger cars based on combustion engines by electric cars for 2030 at the national level. The reduction of NOx emissions for this scenario is compared with traffic emissions for 2017. These reductions are applied to road transport emissions in both modelling approaches (microscale and mesoscale). Therefore, for this scenario, both background concentration and traffic emissions inside the neighbourhood are reduced.

Moreover, 9 additional local traffic emission reduction scenarios are investigated in order to determine the NOx emission reduction required for the EU limit value for the annual mean NO2 concentration not being exceeded. These local measures are modelled considering emission reductions in the microscale simulations. The road transport emissions in the mesoscale simulations are the emissions for the EC scenario. These scenarios consist of several degrees of substitution of light and heavy-duty vehicles with combustion engines by electric vehicles. A brief description of all scenarios is shown in Table 1. Note that NO2 concentrations for different emission scenarios are compared for the same meteorology (2017 year).

Table 1 Emission scenarios
Full size table

Results

Model evaluation

Firstly, the time series of NO2 concentrations recorded at EA-AQMS for 2017 are compared with modelled NO2 concentrations. To minimise the uncertainties of the modelling approach, the traffic emission factors are modified in order to provide the same annual average NOx concentration as that recorded at the EA-AQMS in the 2017 scenario. It is noteworthy that different versions of the WA CFD-RANS methodology were successfully evaluated in different urban environments (Parra et al. 2010; Sanchez et al. 2017; Rivas et al. 2019). In particular, the performance of the methodology was investigated in the neighbourhood studied in the present paper. The spatial distribution of modelled time-averaged NO2 concentrations was assessed by Santiago et al. (2017a) for two different periods (January–February 2011 and November 2014). Two experimental campaigns deploying passive samplers (26 in 2011 and 95 in 2014) inside the neighbourhood were carried out. Modelled time-averaged NO2 concentrations were in agreement with the observations. In addition, the NO2 time series recorded at EA-AQMS were also compared with modelled concentrations for 2016 by Santiago et al. (2022b).

Figure 2 shows the time series, scatter plot and mean diurnal variation of observed and modelled NO2 concentrations at EA-AQMS. In addition, statistical metrics such as normalised mean square error (NMSE), the correlation coefficient (R) and the fraction of predictions that are within a factor of two of the observations (FAC2) are computed (Table 2). Modelled concentrations at EA-AQMS are in agreement with observations. The concentration peak in the morning and the midday hours are slightly underestimated. On the contrary, the evening concentration peak is overestimated. In general, the mean daily variation of observed NO2 concentrations is rather well simulated by the modelling approach. Similar behaviour was found in 2016 by Santiago et al. (2022b). The statistical metrics (Table 2) show a fair model performance, similar to Santiago et al. (2022b) results for the 2016 year and within the criteria proposed by Chang and Hanna (2004) (good performance for NMSE < 3; FAC2 > 0.5) and Goricsan et al. (2011) (fair performance for R > 0.5).

Fig. 2
figure 2

a Time series, b scatter plot and c mean diurnal variation of observed and modelled NO2 at EA AQMS for the year 2017

Full size image
Table 2 Statistical metrics for model evaluation
Full size table

Impact of individual measures (ECOBOIL and EC)and the combination of ECOBOIL + EC

Firstly, the impact of ECOBOIL and EC scenarios on NO2 concentrations at the street level is investigated in the EA-AQMS neighbourhood. The spatial distributions of annual mean NO2 concentrations at the pedestrian level for both scenarios are compared with the base case for 2017 (Fig. 3). Areas exceeding the EU limit value for annual mean NO2 concentrations (40 µg m−3) are noted in red colours. It is observed that concentrations in most parts of streets and avenues are above limit value. However, there are no exceedances in the area corresponding to El Retiro park. These facts indicate the importance of traffic emissions in the final concentrations in these streets. In November 2014, Santiago et al. (2017a) observed a similar strong concentration gradient not only for modelled NO2 but also for NO2 measured by means of passive samplers deployed in this neighbourhood. NO2 concentration maps for the 2016 year obtained by Santiago et al. (2022b) were similar distributions but with slightly higher values in the El Retiro park due to a little higher contribution of the background concentrations.

Fig. 3
figure 3

Maps of annual average NO2 concentrations for the (a) 2017 scenario, (b) ECOBOIL scenario, (c) EC scenario and (d) ECOBOIL + EC scenario

Full size image

NO2 concentrations for the ECOBOIL scenario (Fig. 3b) show negligible differences in comparison with the base case for 2017. It is due to the fact that traffic emissions are the most important contributions to NO2 concentrations at the pedestrian level in these streets, and in addition, the decrease of background concentration (computed by the CHIMERE model) due to this measure is small in this area. To quantify these differences, the spatially averaged concentrations over the whole neighbourhood are computed for all scenarios (Fig. 4a). Spatially averaged NO2 concentration hardly decreases from 55 µg m−3 for the base case to 54.6 µg m−3 for the ECOBOIL scenario. In addition, in order to determine the size of the exceedance area, the ratio between the area with concentrations higher than the annual mean limit value for NO2 (Aover40) and the total study area (Atotal) is investigated (Fig. 4b). For the base case, this ratio is 52%, and it barely decreases until 51.6% for the ECOBOIL scenario. These values are lower than the ratio obtained for 2016 (Aover40/Atotal = 100%) by Santiago et al. (2022a), despite the spatially averaged NO2 concentrations being similar for both years (58.3 µg m−3 for 2016 and 55 µg m−3 for 2017). It is due to the fact that, for 2016, the contribution of background concentrations induces NO2 concentrations above 40 µg m−3 in the park for 2016. However, for 2017, the concentrations in the park, although close to 40 µg m−3, are below this value. In addition, considering the revision of the Ambient Air Quality Directive by the European Commission, the results are also discussed in terms of the value proposed for the annual limit value for NO2 (20 µg m−3). The area with concentrations higher than 20 µg m−3 (Aover20) is also computed, and it is observed that the concentrations in the whole exceed this value (Aover20/Atotal = 100%) for 2017 and for ECOBOIL scenario since the background concentration is already above 20 µg m−3.

Fig. 4
figure 4

a Spatially averaged annual NO2 concentrations for the study neighbourhood for 2017, ECOBOIL, EC and ECOBOIL + EC scenarios. b Ratio (in %) between the area above the annual mean limit value for NO2 (40 µg m−3) (Aover40) and the total study area (Atotal) for the study neighbourhood for the 2017, ECOBOIL, EC and ECOBOIL + EC scenarios. c The same as (b) but for the proposed limit value of 20 µg m−3 (Aover20)

Full size image

As expected, the most important effects are found for the EC scenario. In Fig. 3c, it is observed that NO2 concentrations decrease compared with that obtained for the base case, reducing its values in several streets below the limit value. The spatially averaged concentration over the whole study area decreases from 55 µg m−3 for the base case to 40.3 µg m−3 for this scenario (Fig. 4a). However, this concentration is still above the limit value. Regarding the area with concentrations above the limit value, Aover40/Atotal is 39.2%. For this scenario, the background concentration is slightly lower than 20 µg m−3, and considering this new limit value, the size of the exceedance area increases but does not cover the whole domain (Aover40/Atotal = 86.1%).

Considering both measures at the same time (ECOBOIL + EC), the spatially averaged NO2 concentration is slightly below the limit value (39.9 µg m−3). However, the sizes of the exceedance area for both limit values are very similar to the EC scenario (Aover40/Atotal = 38.7%; Aover20/Atotal = 83.1%). These results indicate that additional measures are needed for reducing annual mean NO2 concentrations below the limit value at the street level. With this purpose, additional scenarios with different degrees of local traffic emission reductions (EC_2, EC_3, etc.; Table 1) are investigated in the next section since this type of measure has more impact on NO2 street-level concentrations.

EC scenarios

Scenarios considering additional local measures with more intense reductions of traffic emissions are simulated. It consists of different degrees of substitution of combustion engine vehicles (light and heavy duty) for electric vehicles. The objective is to determine the required reduction of local traffic emissions for the annual mean NO2 concentration to not exceed the EU limit value. These additional scenarios are described in Table 1, and the annual mean NO2 concentration maps for them are shown in Fig. 5. It is observed that the size of exceedance areas for the current limit value (40 µg m−3, in red colours in Fig. 5) decreases as emission reductions increase. For the EC scenario, the exceedance areas cover not only avenues with intense traffic but also narrow streets with low traffic intensity. This is due to the low ventilation of these deep street canyons with a large aspect ratio (ratio between the height of buildings and the width of the street). For scenarios with stronger emission reductions, the size of the exceedance area in those streets is reduced as well. From the EC_5 scenario onwards, the exceedances are mainly located in the main avenues. For these scenarios, the spatially averaged concentrations over the whole neighbourhood and the ratio between the area with concentrations above the annual mean limit value for NO2 and the total study area (Aover40/Atotal) are computed (Fig. 6). The spatially averaged NO2 concentration is already less than the limit value for EC_2. However, the exceedance area is still large (Aover40/Atotal = 36.8%). This ratio is already less than 10% for the EC_8, but the exceedance area in the neighbourhood disappears only for the EC_10. Considering the proposed annual limit value of 20 µg m−3, exceeded area increases (Fig. 4c), and for all scenarios, there are exceedances in several areas of the domain. This is due to the fact that the background concentration for these scenarios is already close to 20 µg m−3. In the Appendix, in order to better see the differences between scenarios, Fig. 7 shows the differences in NO2 concentrations between the EC_X scenarios and the EC scenario. It shows that concentrations in the park (south of the domain) are similar since the main contribution is the background concentration. However, the decrease of concentrations in the streets is much larger for larger local traffic emission reductions. The decrease of NO2 concentration is approximately 100 µg m−3 in certain areas for EC_10.

Fig. 5
figure 5

Maps of annual average NO2 concentrations for the (a) EC_2 scenario, (b) EC_3 scenario, (c) EC_4 scenario, (d) EC_5 scenario, (e) EC_6 scenario, (f) EC_7 scenario, (g) EC_8 scenario, (h) EC_9 scenario and (i) EC_10 scenario

Full size image
Fig. 6
figure 6

a Spatially averaged annual NO2 concentrations for the study neighbourhood for 2017 for all EC scenarios. b Ratio (in %) between the area above the annual mean limit value for NO2 (40 µg m−3) (Aover40) and the total study area (Atotal) for the study neighbourhood for the 2017 and all EC scenarios. c The same as (b) but for the proposed limit value of 20 µg m−3 (Aover20)

Full size image

Summary and conclusions

The main objective of this paper is to assess the impact of individual measures for NOx emission reduction on the NO2 concentrations at very high spatial resolution in an urban district of Madrid City. Microscale CFD modelling has been used for the studied domain, and the contribution of other sources beyond the modelling domain boundaries has been taken into account using the CHIMERE model. The employed WA CFD-RANS methodology based on a set of CFD-RANS simulations for 16 meteorological scenarios (16 wind directions) has to allow to make affordable the computation of long-term (annual) NO2 concentrations at very high spatial resolution (a few metres) in an urban area. This methodology was evaluated in other studies; in particular, an ad-hoc evaluation was done in the studied urban domain using data from two experimental campaigns of passive samplers. These campaigns were carried out during three weeks in January–February 2011 and November 2014. The results showed good estimates of long-term NO2 concentrations. This paper focuses on annual average concentrations, and these evaluations were carried out in winter. However, in winter, NO2 concentrations are higher, and their contribution to annual average concentrations is more important. In addition, in the present study, modelled NO2 time series at the EA-AQMS were satisfactorily evaluated against the recorded measurements at the EA-AQMS for 2017 (the base year for this study).

Two scenarios included in the NAPCP with NOx emission reductions from the installation of more efficient boilers for domestic heating (ECOBOIL) and from the partly substitution of passenger cars with combustion engines by electric cars (EC) are investigated. This analysis is extended to 9 additional scenarios (EC_2, EC_3 and so on) of more ambitious implementation of electric vehicles (including heavy and light duty vehicles) in order to determine what the NOx emission reduction required for the EU limit value for the annual mean NO2 concentration not being exceeded is.

Results show that the ECOBOIL scenario has a very weak impact on the NO2 concentrations and in the area of exceedance of the annual limit value. However, the EC scenario implies a very significant reduction of the NO2 concentrations, whose spatial average almost complies with the limit value (40.3 µg m−3), and the ratio of exceedance area decreases from 52.0 to 39.2%. Nevertheless, both scenarios imply improvements; they are not enough to fully remove NO2 limit value exceedances in the studied area.

The analysis of the additional scenarios with different degrees of implementation of electric cars shows that NO2 concentration and exceedance areas decrease consequently. A scenario with a small additional (compared with the EC scenario) implementation of electric cars (EC_2, − 55% of the current fleet of Pcars (except hybrids) substituted by electric cars) fulfils that the spatially average NO2 concentration be lower than the EU limit value, but the area with exceedances is still very large. However, stronger emission reductions corresponding to the most ambitious scenarios are needed in order to achieve that at least 95% of the domain is without EU limit value exceedances. This happens for the EC_9 scenario (− 100% of the current fleet of Pcars (except hybrids) substituted by electric cars and − 50% of the current fleet of HV substituted by electric vehicles) where the NOx emissions are reduced by 80%. Only for the scenario of total substitution of combustion engine vehicles by electric ones (EC_10, NOx emission reductions of 92%) is there no limit value exceedance area within the neighbourhood.

The current work shows that, for the assessment of the implementation of specific measures for air quality improvement, it is important to consider the spatial variability of pollutant concentrations in an urban environment. The numerical methodology employed in the present study (WA CFD-RANS) can be applied for this purpose. In this case, the study focuses on the use of electric vehicles, which is expected to grow in the near future.