**2. Methods**

an aerodynamic diameter less than 2.5 μm) [2]. This demonstrates the widespread need to moderate the rate of motorization by introducing more sustainable ways of transportation. In addition, motorization and technological advances result in diminished physical activity among the world's population, with 23% of adults and 80% of adolescents being insufficiently physically active [3]. This also enhances the risks of cardiovascular diseases, cancer, and diabetes, summing up to one of the leading risk factors for death worldwide [4, 5]. Sadly, obesity levels caused by food and physical inactivity are expected to grow worldwide, and developing countries are projected to exceed the obesity levels of developed countries in the near future [6]. In accordance with the recommendations of World Health Organization (WHO), numerous countries are setting goals on reducing air pollution and insufficient physical activity [3]. It is suspected that both issues of deteriorating air quality and physical activity could be solved by promoting active modes of transportation. In comparison with motorized traffic, cycling alternative requires very little space, and in addition, it is economic, clean, and

100 Air Pollution - Monitoring, Quantification and Removal of Gases and Particles

Transport infrastructure is influenced by historical, political, cultural, structural, and economic aspects of urban development [9]. In the era of motorization, the choice of active way of travel over motorized mode depends on a vast range of factors such as proximity, connectivity, convenience, safety, population density, costs, environmental quality, existence of infrastructure and land use mix [10]. Planning urban cycling network is, therefore, a complex task as the existing and potential uses have to be anticipated in order to assure the efficiency of the infrastructure. Simplifying, urban bicycle path planning approaches are based on space availability and route demand by using GIS, GPS, remote sensing and even artificial intelligence techniques [11, 12]. Any new infrastructure, previously unaccounted for in the city's traffic model, has to be incorporated in an existing urban design [8]. In developed countries, efficient regulations are prioritized (traffic calmed residential neighborhoods, car-free city centers, special bicycle streets, etc.) in order to assure the safety of cyclists [9, 13]. At the same time, the route demand requires analyzing the necessity for the bicycle paths. This is done by computational analysis of urban infrastructure and population (census) or user surveys, and recently by mobile phone user applications registering the cycling activity to study the prioritized pathways (e.g., Strava Metro, Kappo, Chaquiñan Urbano, etc.). These applications analyze mobility patterns through voluntary geo-information (crowdsourcing) [14, 15] and interact with the user through games, implemented levels, missions, and rewards to attract

Common approaches, however, often locate the bicycle lanes, tracks, and paths on the roads or in close proximity to motorized traffic. To improve travel safety and accident prevention on the road, a pioneer strategy applies a clear priority rules limited to two types of transport (e.g., bicycle and bus, or bicycle and cars, etc.) [8]. While it helps to reduce the cases of transit accidents, it does little in terms of cardiopulmonary health, especially in developing countries. While earlier studies on cycling focused on engineering, safety, and promoting cycling [7, 16], a number of recent studies discuss health benefits of cycling. For example, research indicates that if using the roads, the personal exposure of cyclists to

promotes health [4, 7, 8].

new users (Kappo, Capos SpA).

The historical Mariscal (area 1.5 km<sup>2</sup> ) district of Quito, Ecuador, was chosen for this study due to the existing network of bicycle lanes and a low variation in traffic intensity. A motorized traffic count was performed to determine the influence of traffic load on the concentration of PM2.5. It was evaluated by manual observations [29] based on a week's (February 20–26, 2017) traffic counts, performed in a number of one- and two-way streets of the district. In case of two-way streets, the vehicle count was performed both ways. All heavy (trucks, buses, minibuses ) and light (light vehicles and taxis) vehicles were counted from 7 to 8 pm.

At the same time of the experiment, PM2.5 pollution scans were performed. To account for different vertical pollution mixing conditions, a multiple coverage of the main streets was performed. Street level PM2.5 concentrations were measured using a portable real-time CEL-712 Microdust Pro™ monitor [30] coupled with a GPS [31]. The Microdust Pro was calibrated before the experiment using zero-air and a known concentration filter (164 mg/m<sup>3</sup> ). The performance of the Microdust Pro sensor based on near-forward angle light scattering technique was validated collocating it with Thermo Scientific 5014i Beta Continuous Ambient Particulate Monitor (R2 = 0.74). The particle sensor and the GPS were both functioning at a synchronized step of 10 s (particle sensor at 10 s average), held at the height of 1.5 m facing the particle inlet forward while walking at an approximate speed of 2 km/h on the side of the street or an existing bicycle path. The gathered data were then combined with the GPS data to elaborate a pollution map in Qgis. All the collected points were used for the data processing, but atypical data were eliminated. This was done to reduce the impact of these values, which may be the result of some equipment failure [32]. The running average of 1 min was used to represent the points on the map. Geostatistical interpolation of PM2.5 concentrations for the district was performed in Qgis using ordinary kriging method. Ordinary kriging is used in pollution dispersion models to estimate an unmeasured region, assuming a constant linear mean over space [33].

r =

and *N* is the total number of pairs of cells.

stands for the color value of the traffic, *p*<sup>i</sup>

where *t* i 2N − ∑i=1 <sup>n</sup> <sup>∣</sup> ti <sup>−</sup> pi <sup>∣</sup> \_\_\_\_\_\_\_\_\_\_\_\_\_\_\_

**Figure 1.** Flowchart to process traffic and pollution data from maps, and compare the values between each matching cell.

2N (2)

Urban Air Pollution Mapping and Traffic Intensity: Active Transport Application

stands for the color value of the pollution,

http://dx.doi.org/10.5772/intechopen.79570

103

Particulate matter PM2.5 is mainly produced by the vehicular traffic, especially by diesel vehicles. Thus, in order to assess the correlation between traffic and contamination, pollution maps are compared with traffic maps provided by Google Maps Traffic. Google Maps sets four levels going from fast to slow represented by colors green, orange, red, and brown, respectively. For our analysis, slow traffic representing red and brown were combined together. In the case of PM2.5 concentrations, the maximum permissible limits of 24-h international and Ecuadorian regulations were used, three ranges were established: <25, 25–50, and > 50 μg/m<sup>3</sup> ). These ranges of pollution were also assigned by the same colors: green, orange, and, red respectively; in this case, brown is also combined with red. Once the ranges were defined in the GIS software, in this case Qgis, we proceeded to establish the format of the points for the pixel analysis. It was defined that each spatial point has a size of 6 pixels. In addition, the primary colors of green, blue (accounting for yellow), and red were used for the different levels, because it improved the subsequent computation of the spatial representation of the pollution. To create the traffic layers, the PM2.5 pollution map was taken as a basis, and depending on the usual traffic, the color was modified point by point to add the traffic information to each cell. This process was done with the usual traffic of labor days at hours 09h00, 11h00, and 13h00 to cover the hours of the sampling for every studies. A complete street sampling took over 6 h (8h00–14h00) per day.

The original maps produced with Qgis are processed with a program written in Java (**Figure 1**). The first step consists of formatting the maps as a grid. Thus, it is possible to compare the color (level of pollution/traffic) between a cell of the traffic grid and its corresponding cell in the pollution grid. In the second step, the coordinates and the value (1 for green, 2 for yellow, and 3 for red) of each colored cell are recorded in a table, in which the correlation analysis is performed. Two methods are used to carry out the correlation. The first method is based on a strict matching between the colors of the peer cells. Thus, the assessment is Boolean. If the color of both cells is the same, the matching is evaluated as 1, otherwise the value is 0. Then, the value of the correlation (*r*) is calculated by Eq. (1).

$$\mathbf{r} = \frac{\sum\_{i=1}^{n} (\mathbf{m}\_i)}{\mathbf{N}} \tag{1}$$

where *m*<sup>i</sup> stands for a true matching and *N* is the total number of pairs of cells.

The second method is based on a weighted correlation. A weight is calculated according to the amplitude of the difference between the peer cells. For instance, if one cell is red (high) and the other is green (low), the value will be 2. But, if one cell is red and the other is yellow (medium), the value will be 1, only. And if there is no difference between the cells, the score will be 0. For this method, the calculation of the correlation (*r*) is provided by Eq. (2).

Urban Air Pollution Mapping and Traffic Intensity: Active Transport Application http://dx.doi.org/10.5772/intechopen.79570 103

$$\mathbf{r} = \frac{2\mathbf{N} - \sum\_{i=1}^{n} |\mathbf{t}\_i - \mathbf{p}\_i|}{2\mathbf{N}} \tag{2}$$

where *t* i stands for the color value of the traffic, *p*<sup>i</sup> stands for the color value of the pollution, and *N* is the total number of pairs of cells.

elaborate a pollution map in Qgis. All the collected points were used for the data processing, but atypical data were eliminated. This was done to reduce the impact of these values, which may be the result of some equipment failure [32]. The running average of 1 min was used to represent the points on the map. Geostatistical interpolation of PM2.5 concentrations for the district was performed in Qgis using ordinary kriging method. Ordinary kriging is used in pollution dispersion models to estimate an unmeasured region, assuming a constant linear

102 Air Pollution - Monitoring, Quantification and Removal of Gases and Particles

Particulate matter PM2.5 is mainly produced by the vehicular traffic, especially by diesel vehicles. Thus, in order to assess the correlation between traffic and contamination, pollution maps are compared with traffic maps provided by Google Maps Traffic. Google Maps sets four levels going from fast to slow represented by colors green, orange, red, and brown, respectively. For our analysis, slow traffic representing red and brown were combined together. In the case of PM2.5 concentrations, the maximum permissible limits of 24-h international and Ecuadorian regulations were used, three ranges were established: <25, 25–50, and > 50 μg/m<sup>3</sup>

These ranges of pollution were also assigned by the same colors: green, orange, and, red respectively; in this case, brown is also combined with red. Once the ranges were defined in the GIS software, in this case Qgis, we proceeded to establish the format of the points for the pixel analysis. It was defined that each spatial point has a size of 6 pixels. In addition, the primary colors of green, blue (accounting for yellow), and red were used for the different levels, because it improved the subsequent computation of the spatial representation of the pollution. To create the traffic layers, the PM2.5 pollution map was taken as a basis, and depending on the usual traffic, the color was modified point by point to add the traffic information to each cell. This process was done with the usual traffic of labor days at hours 09h00, 11h00, and 13h00 to cover the hours of the sampling for every studies. A complete street sampling took

The original maps produced with Qgis are processed with a program written in Java (**Figure 1**). The first step consists of formatting the maps as a grid. Thus, it is possible to compare the color (level of pollution/traffic) between a cell of the traffic grid and its corresponding cell in the pollution grid. In the second step, the coordinates and the value (1 for green, 2 for yellow, and 3 for red) of each colored cell are recorded in a table, in which the correlation analysis is performed. Two methods are used to carry out the correlation. The first method is based on a strict matching between the colors of the peer cells. Thus, the assessment is Boolean. If the color of both cells is the same, the matching is evaluated as 1, otherwise the value is 0. Then,

> ∑i=1 n (mi ) \_\_\_\_\_\_

The second method is based on a weighted correlation. A weight is calculated according to the amplitude of the difference between the peer cells. For instance, if one cell is red (high) and the other is green (low), the value will be 2. But, if one cell is red and the other is yellow (medium), the value will be 1, only. And if there is no difference between the cells, the score

stands for a true matching and *N* is the total number of pairs of cells.

will be 0. For this method, the calculation of the correlation (*r*) is provided by Eq. (2).

<sup>N</sup> (1)

).

mean over space [33].

over 6 h (8h00–14h00) per day.

the value of the correlation (*r*) is calculated by Eq. (1).

r =

where *m*<sup>i</sup>

**Figure 1.** Flowchart to process traffic and pollution data from maps, and compare the values between each matching cell.
