Using smart city tools to evaluate the effectiveness of a low emissions zone in Spain: Madrid Central

: Population concentration in cities brings new risks as an increase in pollution, which causes urban health problems. In order to address this problem, trafﬁc reduction measures are being implemented, as pedestrianization areas and the deﬁnition of Low Emissions Zones (LEZ). When the effectiveness of these types of measures is in doubt, smart city tools provide data that can be used to scientiﬁcally asses their impact. This article analyzes the situation of Madrid Central (Spain), a LEZ subject to controversy. We apply statistical and regression analyses to evaluate the effectiveness of this measure to reduce air pollution and outdoor noise. According to the results, this LEZ was able to signiﬁcantly reduce NO 2 , PM 2.5 , and PM 10 concentration locally, having the same positive impact in the rest of the city. In terms of noise, this measure is able to mitigate background noise levels generated by the road trafﬁc. these results, we answer RQ2: Is the deﬁnition of a LEZ an effective measure to reduce environmental noise levels? . MC slightly reduces the outdoor levels of noise, mainly the background noise produced by road trafﬁc. However, this decrease is not enough to keep the noise in the range of healthy levels.

life, and diminishes the quality of health [3]. Air pollution, which causes more than 400,000 premature deaths is considered the top health hazard in the European Union (EU) [17,18]. It has proved to be associated with heart disease and strokes, lung diseases, and lung cancer, besides reducing lung capacity and aggravating asthma. Air pollution has been also pointed out as carcinogenic and causes infertility and diabetes type 2 [19]. Other studies link air pollution to obesity, systemic inflammation, aging, Alzheimer's disease, and dementia [20]. It also affects the brain in the same way that Alzheimer's does as it causes changes in the structure of the brain [21]. While pollution affects all ages, some population groups are more vulnerable to pollution problems, as pregnant women, newborns, children, and the elderly. It also can exacerbate preexisting conditions at all ages.
Environmental noise is the major preventable cause of hearing loss [4]. It can also cause a range of non-auditory problems. To begin with, the evidence for the effects of environmental noise on health is strongest for annoyance, sleep disturbance, and cognitive performance in both adults and children [22]. Sound pollution also affects the cardiovascular system and causes hypertension [4,22]. Among children, it generates cognition problems as communication difficulties, impaired attention, increased arousal, frustration, and worst performance [23,24]. Last but not least, noise pollution causes annually at least 16,600 cases of premature death in Europe [25].
These risks could be reduced by limiting car use. Among different approaches to reduce car use, pedestrianization and LEZ have been proved as one of the most effective strategies against emissions. Absolute pedestrianization is difficult to be implemented (and expensive) but LEZ in different European Member States have proved to be successful. These are the cases for the LEZ ("Umweltzone") for trucks and cars in the center of Berlin inside the S-Bahn ring (Germany); the LEZ in the Lombardy Region for motorcycles, buses (whole year), and vehicles during wintertime, e.g. in Milan (Italy); or the LEZ for vans and lorries in Greater London (United Kingdom). All of these experiences have registered good results in terms of improving air quality.

Reducing pollution: international directives and the Low Emissions Zone in Madrid
While cities are related to social and economic progress, the increase of air pollution and noise can be considered a modern plague [26] decreasing the quality of life in cities [27]. In Europe (where 74% of the population lives in urban areas) the concern on this matter has led to the creation of a common set of environmental rules. These directives could not just safeguard EU citizens from environment-related pressures and risks to health and well being, but also reduce significantly different expenses: an adequate implementation of environmental legislation could save 50 billion euro every year in health and environment costs to the EU economy [20].
Because of this, a Clean Air Policy Package is adopted in 2013 by the the European Commission. This air quality policy, based on Directive 2008/50/EC [28] and 2004/107/EC [29] rests on three pillars: i) air quality standards; ii) national emission reduction targets established in the National Emissions Ceiling Directive; and iii) emissions standards for key sources of pollution (as cars).
Being noise another harmful hazard, the EU regulation (Directive 2002/49/EC) [30] establishes for every Member State: i) the creation of noise mapping to determinate exposure levels to environmental noise; ii) made environmental noise information available to the citizen; iii) to adopt action plans, based upon noise-mapping results, to prevent and reduce environmental noise where necessary.
The referred Directive points road traffic are major noise sources, which has been also noted as the largest contributor to nitrogen oxide emissions [31]. For the reasons alluded previously, NO 2 is one of the main concerns of the EU: exposure to air pollution from road transport costs about 137 billion per year in Europe [32] but, more important, it produced around 76,000 premature deaths in 2015 [33]. Despite this, a number of countries keep exceeding the maximum established PM 10 and the NO 2 levels. Spain is one of them, with several urban areas under surveillance by the EU because of their poor air quality. The European Commission required the reduction of these air pollutants under the threat of penalties. Madrid City Council decided to reduce transit traffic in a delimited area of the downtown replicating other European experiences. Madrid Central is a LEZ in Madrid, consisting of a number of car access restrictions for non-residents, independent from the current pollution level of the air.
The clarity in the perimeter of MC is one of their virtues, facilitating so the understanding of zonal delimitation and helping to introduce a behavioral change in the city.
MC is created by the Ordenanza de Movilidad Sostenible (October 5th, 2018) and it covers almost the entire Centro district, which is formed by the neighborhoods of Palacio, Embajadores, Cortes, Justicia, Universidad, and Sol. In this area, of 4,720,000 m 2 , Centro inhabit 134,881 people, of which 21.6% are over 65 years old and 9.2% are less than 17 years old. Children and the the elderly are the age groups which can benefit more from the reduction of noise and pollutants.
The goal of MC is to improve air quality, but also respond to the idea of changing the uses of spaces in the city center, prioritizing the pedestrian one and reducing noise pollution. But as we said, its conformation mainly responds to ensure the objectives demanded by the EU. The traffic restriction started on November 30, 2018, and the fines for noncompliance did not start until March 16, 2019. Thanks to this straightforward measure, Spain avoided being brought to the European Court of Justice and so, the economic fine.
However, the measure raised strong opposition from some political parties. After the elections (held on May, 26th 2019) and qualifying the MC measures as inefficient and even unnecessary, the newly elected government approved art. 247 of the Ordenanza de Movilidad Sostenible, applying with it a moratorium on fines from July 1st to September 30th, 2019. Besides a warning from the EU, the decision raised social protests, especially from environmental groups. This suspension led to the emergence of social movements claiming the paralyzation of this reversal based on the negative effects over health and environment, and a warning from the EU. Finally, after a contentious-administrative appeal filed by the Platform in Defense of Central Madrid, a judge reactivated MC suspending the moratorium in fines.

Related works
The proliferation of pedestrianization measures and policies in different cities has led to a number of studies to evaluate their efficacy. A brief review of the related literature is presented next.
The impact of the rapid growth of car ownership in Beijing, China was analyzed in terms of transportation, energy efficiency, and environmental pollution [34,35] A set of measures were applied in Beijing in 2010 to mitigate the effects of traffic congestion and reduce air pollution. Liguang et al. [34] analyzed data from Beijing Municipal Committee of Transport to evaluate the implementation of car use restriction measures. Fairly good effects on improving urban transportation and air quality were achieved according to the results reported. Liu et al. [35] proposed an indirect approach to evaluate the impact of car restrictions and air quality, by applying a generalized additive model to explore the association of driving restrictions and daily hospital admissions for respiratory diseases. Several interest facts were obtained from the analysis, including higher daily hospital admissions for respiratory disease for some days, and the stronger effect on cold season. Female and people older than 65 years benefited more from the applied environmental policy. Overall, authors found positive effects on the improvement of public health. More research had been performed addressing the LEZ analysis in China [36][37][38].
Regarding noise pollution, there are is a body of work on analyzing urban noise and its impact on the population's health [50][51][52][53][54][55]. There is just a few research done on this regard [12,56]. However, most of the studies presented above studied also some variables related to noise pollution.
Focusing on Madrid, air quality has been been a health issue for the last decades. Thus, there are several studies dealing with air pollution in this city. Borge et al. [57] reported the modeling activities carried out in Madrid emphasizing the atmospheric emission inventory development which comprises the combination of models and real data. They showed that Madrid required to reduce NO x emissions to meet the NO 2 European standards, which was the main motivation to implement MC. Different models were used to simulate and evaluate a short-term action plan to mitigate pollution emissions [58,59]. More recently, after the application of MC, Lebrusán and Toutouh [12] evaluated the effectiveness of this measure in reducing pollution inside the LEZ. Another study, applied computational intelligence (deep neural networks and regression analysis) to evaluate the evolution of NO 2 concentration in the air, but again only locally (inside the MC area) [60]. Both reported a decrease in the NO 2 concentration. Our article contributes in this line of research by extending the set of pollutants studied by adding PM among others, taking into account a longer time frame, and including 24 areas of study in order to get a more general idea of the impact of this type of measures in whole Madrid.

Materials and methods
The primary goal of this study is to take advantage of smart city tools, such as a sensor network and open data, to evaluate the effects of a LEZ on environmental pollution and noise. We address such an analysis, first locally, evaluating the impact of this measure at the MC area, and globally, by assessing the same in different areas of the city (Madrid).
Madrid City Council installed different sensors (see Figure 1) that gather data on ambient concentration of different pollutants and measure the output noise levels.  The analysis is performed considering a temporal frame of six years (72 months), from December 2013 to November 2019. Two periods are distinguished: Pre-MC, i.e., the period of five years before implementing MC (from December 2013 to November 2018), and Post-MC, i.e., the period of one year after implementing the mobility measure (from December 2018 to November 2019). The main idea is to compare both periods to asses the effect of the measures applied in MC. 1 Madrid Open Data Portal URL: https://datos.madrid.es/ Following, we introduce the air pollutants analyzed, the outdoor noise metrics studied, and the methodology applied in the evaluation.

Air quality evaluation
The Open Data Portal (OPD) provides the hourly mean concentration of several air pollutants. In this study we focus on six: sulfur dioxide (SO 2 ), nitrogen dioxide (NO 2 ), ozone (O 3 ), carbon monoxide (CO), particulate matter 10 micrometers or less in diameter (PM 10 ), and particulate matter 2.5 micrometers or less in diameter (PM 2.5 ). Table 1 summarizes the maximum concentration allowed of the studied pollutants established by WHO and EU directives.

Noise evaluation
It is not trivial to chose the right parameters to evaluate noise pollution and its impact on the people. The sound-meters return measures that describe the physical attributes of the sound, but not the subjective response and the physiological and psycho-social harm extent to the public. The noise pollution data provided by the OPD includes the daily mean of the equivalent sound pressure levels, the percentile noise levels [61], and the noise pollution level (NPL) [62]: • Equivalent sound pressure levels (L eq ) can be described as the average sound level for the measurement. The data analyzed include the L eq24 that corresponds to the noise measured during the whole day (24 hours). In addition, three L eq are returned by taking into account the period of the day: L eqD , L eqE , and L eqN , which are evaluated during the day (from 7:00h to 19:00h), evening (from 19:00h to 23:00h), and night (from 23:00h to 7:00h), respectively. • Percentile noise levels (L x ) are the levels exceeded for x percent of the time, where x is between 0.1% and 99.9%. L x is calculated by applying statistical analysis. We evaluate the L 10 , L 50 , and L 90 . The L 10 and L 90 are extensively used for rating any annoying traffic noise and background noise, respectively. • NPL was developed to estimate the dissatisfaction caused by road traffic noise comprising the continuous noise level (L eq ) and the annoyance caused by fluctuations in that level. NPL is equal to L eq plus 2.56 times the standard deviation of the noise distribution and it is generally approximated by Equation 1.
All sound levels referred in this article are measured in terms of A-weighted decibel (dBA), which corresponds to the A-weighted sound level readings to replicate the response of the human ear to the annoyance caused by road traffic noise.
The Guideline Development Group (GDG) of WHO strongly recommends reducing L eq noise levels produced by road traffic below 53 dBA. Road traffic noise above this level is associated with , the L eq levels should be between 50 and 65 dBA during the day time and between 40 and 55 dBA during the night.

Methodology
In order to asses the effectiveness of the LEZ(see Section 2.2), we evaluate the air pollution and the noise in the LEZ and in different areas of the city. Thus, we asses the local effect on the LEZ area and whether if it is possible that a border effect is occurring as a result of potential traffic diversion. • The percentage of the time the population is exposed to air pollutant concentration or noise levels bellow to the thresholds defined by EU denoted by τ. This thresholds are shown in Table 1.
The value of τ allows the assessment of the effectiveness of MC to potentially improve urban health, because there may be situations where the pollution or noise is reduced but the situation is still unhealthy according to the EU regulations (e.g., NO 2 concentration > 40 µg/m 3 ). • Polynomial regression is applied to evaluate the general trend in the air pollution concentration or levels of noise with and without the implementation of the road traffic restrictions in Madrid Central. Although is one of the simplest methods for analysis and estimation of time series, yet it is frequently used in the related literature [12,35,38] In this article, two polynomial regression methods are studied: linear and polynomial of grade 10.
In order to determine the statistical significance of the results obtained, Shapiro-Wilks statistical test is applied to check the normality of the distributions and, as the results are not normally distributed, Mann-Whitney U rank test is used to asses if the pollutant is statistically reduced during Post-MC.

Pollution evaluation at the LEZ area
This section analyzes the information gathered by the sensor located in Pza. del Carmen (id. 35), which is the one inside of the LEZ. When evaluating the pollution in MC, we face two main drawbacks: first, the sensor does not gather information about PM 2.5 and PM 10 , and second, the noise data shared through ODP does not provide complete data for 2013. Thus, we do not include in this section the analysis of these two air pollutants and Pre-MC period is defined from December 2014 to November 2018 when the noise is evaluated. Table 2 reports the minimum (Min), the maximum (Max), the median (Median), and the interquartile range (Iqr) for the concentration of the pollutants sensed in MC during the two periods analyzed here. The last column includes the value of ∆ and the check-mark ( ) indicates that there is a statistical difference according to Mann-Whitney U test in such a pollutant for the periods analyzed (i.e., p-value<0.01). Table 3 presents the percentage of time τ that the air pollutant concentration is lower than the threshold defined by the EU. The evaluated measures are grouped by seasons because the meteorological conditions (i.e., wind direction and speed, atmospheric pressure, temperature, and relative humidity) affect the chemical behavior of the evaluated pollutants [65]. Figures 2, 3, 5, and 6 show the mean concentration of the pollutants by months. Notice that Pre-MCcovers a wider amount of time. These figures also illustrate the boxplot of the concentration of the air pollutants for Pre-MC and Post-MC periods and the probability density function (PDF) of the whole data grouped by periods, i.e., Pre-MCand Post-MC. The concentration of SO 2 in the air is higher during Post-MC than during Pre-MC for all the seasons (i.e., ∆ > 0). Figures 2 illustrates this increase. This is principally due to the largest sources of SO 2 emissions are from fossil fuel combustion at power plants and other industrial facilities [65]. Thus, mobility-related measures or policies are not appropriate for reducing the SO 2 concentration in the air.

Air quality
For both periods, the mean concentration does not exceed the threshold defined by the EU (20 µg/m 3 ), however, the maxima values do (see Table 2). Table 3 shows that the population is under a SO 2 concentration lower than the UE threshold during 95.81% of the time for Pre-MC and 94.48% for Post-MC. Thus, the excess of this pollutant is exceptional and negligible, so it is not considered problematic in Spain.
Focusing on NO 2 , which is the pollutant that almost lead Spain to the European Court and whose excess is a public health concern, its concentration is significantly reduced for all the seasons, but winter (i.e., Mann-Whitney U statistical test p-value<0.01). The decrease of NO 2 concentration is lower during winter because of the heavier use of combustion power plants for wintertime home heating (therefore, the road traffic may not be the main source of NO 2 ), as well as the fact that NO 2 stays in the air longer in the winter [65]. As it can be seen in Figure 3.a, the concentration of NO 2 exceeds during several months the maximum one allowed by the EU for both periods (Pre-MC and Post-MC) but with important differences. Thus, Table 3 results indicates that after MC measures the population is under healthier air during 15.29% longer than during Pre-MC(62.17%-46.88%). Figure 3.c confirms that MC is under lower concentration of NO 2 in the air during Post-MC.
It is noticeable that there is a clear downward trend in the concentration of NO 2 after the application of the car restrictions (see Figure 3.a). We apply regression analysis to evaluate the general trend of this pollutant.  In Figure 4.a, the polynomial regression of grade 10 (black line) shows, first, how the NO 2 concentration increases during colder seasons and decreases in warmer ones and, second, that the pollution values during Post-MC are lower than Pre-MC. In turn, the linear regression (black dashed line) displays a declined trend over time for this air pollutant. In Figure 4.b, the black dashed line that represents the linear regression of the NO 2 concentration before applying MC (five years) has a positive slope (i.e., the NO 2 concentration tends to increase). The solid black line that represents the general trend after applying MC measures has a negative slope, which indicates that NO 2 concentration in the air tends to be reduced. Thus, the behavior of the concentration of NO 2 in the air under the application of MC measures points out that the traffic restriction has a positive effect on air quality.
The concentration of O 3 shows a similar behavior for both periods (see Figure 5 and Table 3). The results in Table 2 shows that the concentration of this pollutant has increased after the application of MC during spring, autumn, and winter, but it has decreased during summer. All the monthly averages O 3 values are lower than the maximum defined by the EU (120 µg/m 3 ). According to Table 3  The increment of O 3 , that we illustrate in our analysis, can be due by the oxidation of NO, i.e., the chemical reaction of O 3 and NO that forms NO 2 and O 2 , which occurs in urban areas [66]. As the road traffic limitation reduces the concentration NO, the portion of O 3 that reacts with NO is lower. Therefore, the levels of O 3 do not decrease, and subsequently, the concentration of NO 2 produced by the oxidation of NO is lower. In short, this upturn can be a chemical consequence of the reduction in the air of other components concentration.
Regarding CO, as it occurs with O 3 , the concentration of this pollutant decrease during summer, but increase during the other seasons (see Table 2 and Figure 6). Although one of the major sources of this pollutant to outdoor air is road traffic vehicles or machinery that burn fossil fuels, it seems that the reduction of road traffic does not lead to a decrease of CO in this case. However, according to the EU regulations, there is not a need of reducing CO since during the time frame analyzed in this article there is not any measurement over the threshold stipulated by the EU (10 mg/m 3 ). motivations for making this move. Therefore the answer to RQ1: Is the deployment of MC effective to reduce the concentration of NO 2 and so improving the air quality in this area? is yes. The restriction to the road traffic applied in MC has significantly reduced the NO 2 concentration. Table 4 reports the minimum (Min), the maximum (Max), the median (Median), and the interquartile range (Iqr) of the levels of noise sensed in MC during the two periods analyzed here. The last column includes the value of ∆ and the check-mark ( ) indicates that there is a statistical reduction in such a noise level between Pre-MC and Post-MC according to Mann-Whitney U test(i.e., p-value<0.01). Table 5   Regarding the equivalent sound pressure levels (L eq24 , L eqD , L eqE , and L eqN ), the level the noise is higher during the hours between 7:00h and 23:00h than during the night time (see Table 4 and Figure 8). This is mainly due to MC neighborhood is located in a commercial area and the business hours in Madrid usually end at 22:00h, thus there is road traffic until that late hours. There is a reduction in the median of all these noise levels during Post-MC. Therefore, in general, the noise is lower during this period. The highest differences between the two evaluated periods are given by the evening (L eqE ) and at night (L eqN ) noise levels (see Table 4).

Outdoor noise in MC area
Focusing on Post-MC time, the levels of noise decrease during the first months but they experience an increase after June, i.e., when MC suffered from the reversal (see Figure 7). If we focus on L eq24 and L eqD , this increment is important. Taking into account the noise levels before the reversal, the reduction ∆ for these two levels of noise is significantly higher, i.e., ∆(L eq24 )=-0.52 dBA and ∆(L eqD )=-0.45 dBA. Thus, we can observe a negative impact on the noise pollution of the reversal, that is mainly suffered at the time between 7:00h and 19:00h.
Evaluating τ during Pre-MC and Post-MC, Table 5 shows that the outdoor levels of noise practically always surpass the thresholds of the city council. During the nights the levels of noise never are lower than 55 dBA, which is the threshold for these hours. Notice that we are evaluating equivalent sound pressure levels averaged for every day. Figure 7 also illustrates how the noise levels generally exceed the thresholds, i.e., the monthly average levels of noise are higher than 65 dBA for L eqD and L eqE and higher than 55 dBA for L eqN . Therefore, other measures must be applied to further reduce outdoor noise in this area. Focusing on the percentile noise levels (L 10 , L 50 , and L 90 ), Pre-MC and Post-MC differences are statistically significant. The best improvement ∆ is shown by L 90 (see Figure 4), which represents the residual background levels of noise of the urban area analyzed. As the continuous road traffic flow is one of the main sources of the background noise, the reduction of traffic transit provokes a decrease in this type of noise.  Regarding L 50 , which statistically represents the median of the fluctuating levels of noise, is also reduced, i.e., ∆(L 50 )=-0.71. The reduction of the annoying peaks in noise (i.e. L 10 ) is lower than for the other two percentile levels (∆(L 10 )=-0.37). This represents a limited decrease of 0.5% regarding the median value of this noise during Pre-MC (70.70 dBA).
There is not a statistically significant reduction of NPL (see Table 4) and the average improvement is 0.00. Besides, as the computation of this metric depends on L eq24 , it suffers from the same increase during the last months of Post-MC after the reversal of MC (see Figure 10).  In the bottom of Figures 11, 12, and 13, the black dashed line that represents the linear regression of the levels of noise before applying MC (from 2014 to 2018) have slopes close to zero or even positive in the case of L eqD , i.e., there is an increase of outdoor noise in the area. The solid black line that represents the general trend after applying MC measures (last four years) has a steeper negative slope, which is higher in the case of L 90 . Thus, the car restrictions tend to improve the background noise generated by road traffic.
Finally, we have computed the PDF of the L eq24 , L eqD , and L 90 noise values to confirm that there is a slight reduction in the outdoor levels of noise in the area of MC. Figure 14 illustrates that the distributions of Post-MC sensed values are more likely to be lower than the Pre-MC ones (the Post-MC distribution is lightly shifted to the left). However, although there is such a reduction, it is notorious that other types of measures are needed to mitigate this source of health problems and discomfort because the levels of noise exceed the thresholds set by the institutions nearly all the time during both evaluated periods, Pre-MC and Post-MC.
According to these results, we answer RQ2: Is the definition of a LEZ an effective measure to reduce environmental noise levels?. MC slightly reduces the outdoor levels of noise, mainly the background noise produced by road traffic. However, this decrease is not enough to keep the noise in the range of healthy levels.

Indirect repercussion on the pollution of the whole city
According to WHO and EU, NO 2 and particulate matter (PM 2 .5 and PM 1 0) are the main culprits in public health problems due to pollution. This section aims at taking advantage of the sensors network installed in Madrid to: • first, confirm that the important reduction of NO2 emissions at MC area does not lead to an increase of such a pollutant in other zones due to a hypothetical redirection of traffic to other areas of the city, i.e., investigate the possible border-effect of MC, and • second, analyze the concentration of the particulate matter in the areas where the sensors gather such an information in order to asses the possible impact on this aspect of the city due to MC measures.
The data provided by the ODP is incomplete and contains errors for several of the areas covered by the sensors. Only the data coming from sensors that correctly registered values for the whole time frame analyzed here are used in this analysis. Therefore, this section discusses data about NO 2 from 23 sensors, PM 2.5 from six, and PM 10 from 12. This limits the areas of Madrid analyzed in our article. But it ensures that the data reflect reliably the concentration in the air of the pollutants analyzed.

Temporal variation in the air pollutants
In order to better understand the impact of NO 2 , PM 2 .5, and PM 1 0 in Madrid, we evaluate their monthly trends that are shown in Figure 15. As seen in Section 4.1, NO 2 exhibits seasonal variations, with the highest concentrations in winter and the lowest in summer. During warmer months (from March to September) the median NO 2 concentration is lower than the EU threshold (40 µg/m 3 ). The highest NO 2 concentration is suffered during December and the lowest one during August.
This seasonal variation is mainly due to two different factors: the meteorological conditions and the emissions patterns. For example, temperature inversions and lower boundary layer heights in winter can avoid NO 2 to be ventilated from the boundary layer, leading to higher concentrations in European cities. In contrast, the increase of photochemical activity, solar radiation, etc. during summer lower NO 2 concentration [67]. In addition, fossil fuel combustion sources such as residential coal and and biomass combustion for heating also contributed to the formation of a high NO 2 concentration in wintertime. Focusing on PM 2.5 , the highest concentration of this pollutant occurs in December (as it happens with NO 2 ). This pollutant presents a decreasing trend from December to May, being springtime the least polluted season. March, April, and May median values are the only ones that are lower than the threshold marked by the EU. For the months between June to November, the PM 2.5 concentrations tend to be slightly higher than the EU threshold and similar to each other. PM 10 shows a similar seasonal variation than PM 2.5 , from December to May there is a decrease in the concentration of this pollutant. July is the month with the highest PM 10

Effect of mobility restrictions on the NO 2 concentration at other areas of the city
The application of restrictions to vehicles to access a street or area affects the mobility patterns of the inhabitants in the whole city [3], and therefore, it impacts on the pollution in different areas of the same city. One of the main motivations of the deployment of MC was the reduction of this pollutant (see Section 2.2). In order to asses the effect of MC on the rest of the city of Madrid, we study the NO 2 concentration measured by sensors located through Madrid that gather the ODP data (see Figure 1). Thus, we evaluate the concentration of this pollutant during Pre-MC and Post-MC. Thus, we compute the difference between these two periods for each one of the sensed areas. Table 6 shows the median (Med) and interquartile range (Iqr) of the daily sensed NO 2 concentration for each sensor grouped by years (these values are selected because the distributions are not normally distributed). The last column presents the ∆ values. The minimum median concentration for each sensor is marked in bold. Finally, Figure 16.a illustrates the boxplots of the concentration for all the sensors for each year.  According to the results in Table 6, the median NO 2 concentration in the air during the year 2019 (i.e., Post-MC period) is the minimum for 14 of the 23 areas sensed. It is noticeable that 2014 is the year that has been sensed the minimum median concentration of this pollutant (for 11 of the 23 sensors). Thus, the years with lowest NO 2 concentration are fist 2019 and second 2014 (see Figure 16.a).
During the period from 2015 to 2018, several areas suffer from median NO 2 concentrations higher than 40 µg/m 3 (the EU threshold), which motivates the EU action to fine Spain and the following measures to define MC to avoid it. In 2019 and 2014 there are only the Pza. Elíptica sensor exceed this threshold. The main problem in this area is that it suffers from heavy traffic and congestion because its main road, the A-41 highway, connects Madrid with the southern towns.
As shown in the last column of Table 6, the highest reduction of the NO 2 concentration is sensed in Pza. del Carmen, which is in MC area. As expected, another important reduction is given at the area sensed by sensor installed in Pza. España (sensor id 4), which is the closest sensor to the MC area. It experiences a reduction in the NO 2 concentration of 5.48 µg/m 3 . In addition, there is a reduction in the average NO 2 concentration for all the sensed areas but three exceptions (Barajas Pueblo, Ens. de Vallecas, and Juan Carlos I, which are suburb areas far from the center of Madrid).
The results in Table 6 and Figure 16.a indicate that, in general, the deployment of the LEZ has a positive impact on the whole city because after that the air in Madrid generally is healthier (contains lower NO 2 ). This results are in line with the study that acknowledged that NO 2 concentration levels in Madrid are dominated by local traffic (up to 90%) [57]. Thus, reducing the road traffic leads to reduce NO 2 concentration in this city.

Repercussion on the particulate matter concentration in other areas of Madrid
As the sensor located at MC does not gather particulate matter concentration data, we analyze the effect of MC on this type of pollutants in different areas of the city. The number of sensors that gather trustworthy data during the time frame of our study is only six for PM 2.5 and 12 for PM 10 . Figure 17 shows the location of these sensors. This limits the outcomes of this section about the concentration of these pollutants in the whole city. a) Sensors that gather PM 2.5 data a) Sensors that gather PM 10 data Figure 17. Location of the sensors that gather particulate matter concentration data.
Tables 7 and 8 present the median (Med) and interquartile range (Iqr) of the daily sensed concentration of PM 2.5 and PM 10 , respectively, for each sensor grouped by years. The last column presents the ∆ values (difference between Pre-MC and Post-MC). The minimum median concentration for each sensor is marked in bold. Finally, Figures 16.b and 16.c show the boxplots of the concentration for all the sensors for each year of PM 2.5 and PM 10 , respectively. There are only six sensors able to provide trustfully data about PM 2.5 during the time frame of our study and they are located in the downtown of the city (see Figure 17.a). Five of these six sensed areas show the minimum concentration of of this pollutant during 2019 (see Table 7). The sensor that does not presents its minimum during 2019 (Casa de Campo sensor id 24) is located in the center of a large Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 5 May 2020 doi:10.20944/preprints202005.0086.v1  park (green area). In addition, Casa de Campo presents the lowest PM 2.5 concentration in comparison with the others. Among the analyzed years, the median of the PM 2.5 concentration during 2019 is the lowest one (see Figure 16.b). For all the evaluated areas, there is a decrease on the concentration of this pollutant, which is statistically significant for five of them. This is expected because the main source of PM 2.5 in Madrid is the road traffic according to the Screening for High Emission Reduction Potentials for Air quality tool (SHERPA) developed by the Joint Research Centre to quantify the origins of air pollution in cities and regions [68].   Table 8). The concentration of PM 1 0 during these two years (2018 and 2019) show similar lower distributions than the other periods of time (see Figure 16.c). When comparing Pre-MC and Post-MC, eight areas present a decrease on PM 10 after the deployment of MC, five of which are statistically lower.
According to the results about the NO 2 , PM 2.5 , and PM 10 concentration in the whole city of Madrid, the answer to RQ3: Do pedestrianization policies in a given area of the city produce a pollution displacement to other zones of the city? is that they do not produce pollution displacement. These pedestrianization policies positively impact on the whole city because there is a general reduction on the concentration of these three pollutants.
Finally, after the whole study we answer the RQ4: Are smart city tools effective for evaluating urban health policies and other measures implemented in the city?. As we have seen throughout the article, the application of smart city tools proves to be an effective way of assessing the effectiveness of measures against urban pollution. In this sense, the evolution towards the smart city would improve the local capacity to deal with the risks arising from rapid urbanization. The application of Internet Of Things to policy implementation allows us to avoid wrong, subjective or biased appraisal, offering an objective assessment of their effectiveness. Furthermore, the smart city paradigm proves to be the best way to monitor compliance with international emission requirements. However, we have faced the issue of dealing with non-homogeneous and incomplete data for our study. This may limit the outcomes of the analysis carried out: data analyses only can be as trustworthy as the data source. Thus, it is mandatory to provide a platform able to gather and share data complete and accurate.

Conclusions
The growth of car-oriented cities is raising new urban health problems resulting from the pollution increase. This requires quick responses to create sustainable environments from an environmental point of view. Several initiatives are being taken into account to address this challenge but some have been questioned in terms of their effectiveness. Smart city related technologies provide invaluable tools of analysis, helping decision making, and leading to the best outcome for the city. In this article, we evaluate the LEZ deployed in Madrid (Spain) applying smart city tools in order to objectively asses the reduction of the pollution of this measure and potential side-effects.
Real data provided by the Madrid City Council was processed to get time series of air pollutant concentrations and levels of noise in different areas of the city. According to the statistical and regression analyses, MC was able to significantly reduce NO 2 concentration locally, having the same positive impact in the rest of the city. In addition, it has been experienced a decrease in PM 2.5 and PM 10 in most of the analyzed zones of the city. Thus, this LEZ effectively improves the air quality and it does not provoke border-effect. In terms of noise, this measure is able to slightly reduce outdoor noise levels, mainly the background ones generated by road traffic.
We found difficulties in terms of the quantity, quality, and reliability of the open data shared by the city council. Despite these limitations, smart city tools in Madrid have proved to be an invaluable resource to evaluate the effectiveness of this type of environmental measures.
The main lines for future work include extending the analysis performing a multivariate analysis by taking into account related data (e.g., wind speed, temperature, etc.); evaluating the impact on other relevant indicators (e.g., economical impact, mobility behavior, citizens' health, etc.); and applying other time series analysis methods and models (e.g., Markov Chains and recurrent neural networks) to characterize the pollution. The appendix is an optional section that can contain details and data supplemental to the main text. For example, explanations of experimental details that would disrupt the flow of the main text, but nonetheless remain crucial to understanding and reproducing the research shown; figures of replicates for experiments of which representative data is shown in the main text can be added here if brief, or as Supplementary data. Mathematical proofs of results not central to the paper can be added as an appendix.