Assessing Urban Ventilation in Common Street Morphologies for Climate-Responsive Design toward Effective Outdoor Space Regeneration

Abstract

Urban heat islands are an environmental hazard which degrade people’s lives worldwide, reducing social life and increasing health problems, forcing scientists to design innovative acclimatization methods in public places, such as sheltering. This paper focuses on providing quantitative indicators about airflow rates and qualitative information about airflow patterns in street canyons for typical street canyon morphologies, which is essential when designing outdoor acclimatization strategies to mitigate urban overheating. This is based on CFD simulations using an enhanced numerical domain model, which can reduce computational cost and simulation time. The study is performed for different ARs, from wide (AR = 0.75) to narrow (AR = 4), and wind speed to characterize their effect on street ventilation The results show that air renewal decreases while the AR increases. The reduction is faster for a low AR and then comes to a standstill for a high AR. In addition, the study shows that inside narrow streets, the pattern of airflow is affected by the wind velocity magnitude. These findings provide numerical values of air ventilation for a wide range of typical street canyon configurations, which represent essential data for designing effective climate control strategies, mitigating urban heat islands and conducting outdoor thermal comfort studies. This work contributes valuable knowledge to the multidisciplinary efforts aimed at enhancing urban living environments.

1. Introduction

1.1. Background and Street Canyon Characterization

Urban climate is a current topic for most of the scientific community due to its reflection on every inhabitant’s life quality. Different factors such as climate change, massive urbanization, city expansion and very high population density in some urban areas have led to the development of a very worrying phenomenon called urban heat island (UHI) described by authors such as Kim [1] in the 1990s. Because of this, many researchers have found evidence of a local rise in temperature around cities of up to 8 °C above the average of surrounding rural areas during the periods when maximum temperatures are reached [2]. This has further implications in urban climates, developing the urban overheating effect [3,4], which is related to the local rise of temperatures inside urban layouts of streets and public places due to changes in urban morphology relating to less vegetation and the use of less adequate materials. This is highly disturbing especially in public places such as streets, where this temperature rise is more severe and affects people directly [5]. In addition, some authors [6] have highlighted the important role that urban morphology has on the UHI effect, addressing the importance of studying street canyons. Therefore, different methods or strategies are required in order to recover the former life of streets among other benefits such as an improvement in life quality and the social interaction between people. This is currently being studied by a major part of the environmental scientific community, not only in relation to its characterization and causes [7] but also the strategies for its mitigation [8,9]. Another goal of these studies is to mitigate the bad effects on general health related to continuous heat exposure in humans [10] and also the effects on sleep quality [11]. In order to elaborate strategies which aim to protect citizens from extreme heat and reduce the effect of climate change in cities, it is vital to gather more information and gain knowledge about the flow patterns of air in cities. For this reason, it is necessary to conduct an intensive study of the airflow inside street canyons, which is one of the main ventilation mechanisms occurring in cities, as some authors have done via simulation [12] and experimentation [13]. This can have a useful application to outdoor acclimatization methods, as the results of the characterization of street canyon ventilation can be applied to depression confinement techniques where the entry of unwanted air into a street or outdoor area is reduced, and together with the natural conditioning of the area, thermal comfort can be improved.

In the characterization of the airflow inside street canyons, it is common to use several parameters to set a classification of the case studies. The most important ones are the aspect ratio of the street and the Reynolds number. The first one measures the ratio between the height of the street (H) and its width (W) as $AR={\displaystyle \frac{H}{W}}$, and it gives an idea of its narrowness, as some authors have pointed out [6,14]. On the other side, the Reynolds number (Re) represents the turbulent behavior of the flow and is defined as $Re={\displaystyle \frac{V_{ref}\times H}{\nu }}$ and has also been used in different research [15,16,17], where H is the reference length of the scale (the street height), V_ref is the reference wind speed and ν is the kinematic viscosity of the fluid. Both parameters are crucial because they determine the flow pattern as well as the number and position of the vortices inside the street.

There is much research on this topic, and the majority can be classified into two categories. The first one is a qualitative description of street canyon phenomena [18] with no numerical results. The second one provides numeric results focused on the determination of pollutant concentration and dispersion inside streets, which is the main research topic in this field [19,20,21,22]. However, there is a significant lack of information related to the quantification and evaluation of street ventilation with clear results for a wide range of realistic street morphologies, which this paper aims to supply.

All these studies are based on the investigation of the urban boundary layer (UBL), which is the nearest layer of the atmosphere to the surface, which is highly disturbed by the presence of different rural and urban landscapes. The topic has been deeply studied by some scientists such as Oke [23,24,25] and many others whose results have led to the gain of useful knowledge for the characterization of the UBL and its dependence on the landscape [26,27,28].

There are two main procedures to obtain results for this phenomenon intrinsically related to fluid dynamics: real experimentation or numeric simulation. On the one hand, the experiments can be split into those conducted outdoors [29,30], which results in more realistic conditions but cannot be controlled freely. For instance, Chen Guanwen et al. [29,30] first studied the thermal behavior of a specific street canyon on a reduced scale and then studied the ventilation of a specific reduced-scale street canyon outdoors. In both, they obtained contours of different magnitudes, such as temperatures or airflow patterns, but did not obtain numerical ventilation results.

As such, the experiments conducted indoors [31,32] can adjust their conditions to specific problems. Garau et al. used a wind tunnel to study the removal of pollutants in a street canyon for low Re values, and Baratian-Ghorghi and Kaye studied the effect of the shape of a roof and AR on the airflow pattern but again without providing numerical ventilation results.

In addition, the main handicap of experiments is that they usually use scaled models, which can lead to some issues if boundary conditions and air-inflow speed are not adequate, as results can be affected by the Reynolds Number. Chew et al. [33] in their work studied a wide range of Re numbers for a fixed AR, determining that airflow pattern depends on the Re number, but did not provide ventilation results for other AR configurations. This shows that the Reynolds number of the calculation has to resemble reality, using realistic values of building height (H) and wind speed, as Zhang et al. [34] did for a specific street canyon. In order to reflect this in this work, deep research about the influence of Re varying V_ref has been conducted. This was achieved by comparing the main ventilation results for three different ARs, changing the wind speed from very low to high.

On the other hand, simulations are carried out by CFD (computational fluid dynamics) with different software. The most widespread are OpenFoam and Fluent, with the latter being the one used in this work. There are different options for modeling the street canyon. Nosek et al. [13] used a 3D model and obtained ventilation results along the length of the street, while Mei et al. [35] compared the ventilation results for two specific AR values using a 3D and 2D model, determining that the results were similar when the street was long. That is why other authors use the simplification of an infinitely long street that was proposed in that work, therefore using a 2D model. For instance, Hang et al. [36] used a bidimensional model to study the effects of different wall heating on street ventilation, focusing on airflow pattern and flow characteristics such as turbulence but again missing numerical information about air renewal.

Authors use different approaches to solve the turbulent nature of the flow across the urban landscape. For instance, Chung and Liu and Li et al. [37,38] opted for using a large eddy simulation, which provided good results compared to reality. However, the research using this approach is usually used for specific studies as it requires more time and computational cost, lacking street ventilation data in a wide range of cases. On the other hand, different authors have proposed Reynolds-averaged Navier–Stokes (RANS) equations, more specifically the k-ε RNG turbulent model, to be the best solution for this problem [39,40,41], as it can provide good results compared to LES or wind tunnel experimental data with low computational cost. For example, Salim et al. [42] compared flow magnitudes in a street canyon using both methods, concluding that the use of RANS equations is acceptable, although LES provides more accurate results.

Traditionally, studies in this regard have used two main types of geometrical models. On one hand, considering the urban canyon as a canyon on the roof level, for instance, Yang Chen et al. [43] studied the temperature reached within a street for different ARs. On the other hand, some authors consider the street canyon as an elevation above the floor level, such as Nazarian and Kleissl [44], who obtained numerical ventilation results for only one street canyon configuration. Both simplify the street layout by considering the windward wall and the leeward wall of the building and the ground in between, which provides very good results without having to design a more complex geometry. However, this simplification implies that an appropriate numerical domain must be determined because it affects fluid behavior. In geometrical models where a single canyon is considered (as in this work), upstream length or roughness elements are crucial in order to have a fully turbulent developed fluid, so a bad geometrical model leads to a bad characterization of the fluid and not accurate results. That is the reason why in this work an optimization procedure is performed in order to reduce the dimensions of the model and the meshing elements in it while maintaining good-quality results with a low simulation time.

Other studies have used more complex street models and have obtained illustrative results on how different features of the street layout affect the fluid pattern inside the canyon. Some of them have included trees [45], vehicles [46], different shapes of gables [32] or different street morphology [47]. All of them have in common that they study the effect of a specific feature on the flow characteristics in a street canyon, but they do not provide the air ventilation results for the base case.

Another important aspect of CFD simulations is the type and size of the mesh. The quality of it highly determines the accuracy of the results, as shown in some papers, for instance Sukri and Ali [48], who deeply studied the influence of mesh quality when obtaining street ventilation results. For this reason, it is advisable to compare results obtained when using different meshes to know which one provides accurate enough solutions within a fairly low simulation time. It is desirable to reach a balance between both factors to obtain optimal results.

In summary, all previous work from different authors has studied the different features of ventilation and pollutant dispersion in street canyons. However, they have studied the different magnitudes of flow within the street, such as its velocity, its turbulence, or the temperature field, without giving numerical results for air ventilation or discussing the effect that air speed (and thus Reynolds number) has on the airflow pattern changing the number of vortices formed. In addition, in those papers where air ventilation results are provided, they usually focus on a few specific street configurations, with a low applicability to acclimatization methods, as information is needed for a wide range of morphologies. This paper aims to fill the lack of information regarding the ventilation results in this phenomenon for a wide range of cases of typical street canyon configurations, in the form of the numerical values of the mass flow rate entering the street canyon and the flow pattern inside it, in addition to different graphs which help to understand the relation of these magnitudes with the outlined parameters as AR or Re. This will be conducted with a CFD model, using H = 10 m as the height reference value, assembling a typical three-floor building in a city. Furthermore, an innovative method has been carried out in order to distinguish the proportion of air inflow which comes from outside to that propelled back from the street vortex itself, which is not renewed air.

1.2. Objectives

The main goal of this study is to provide a sort of catalog where different indicators of interest are obtained for each case studied. These indicators are the streamline patterns inside the urban canyon, the airflow rate, the air renovation rate and, eventually, an innovative quality indicator named equivalent airflow rate, which is defined in this paper and provides information about the portion of the flow rate which comes from outside the street canyon and not from the vortex itself, with the temperature of the external air. Therefore, the main interest of this study is obtaining full quantitative results about airflow rates and other indicators in a wide variety of cases for their potential use in subsequent engineering acclimatation projects.

For this aim, a study is conducted varying both main parameters (AR and Re) to obtain those indicators and see how they are affected by the variation in them. First, a specific air velocity reference is fixed, and the AR is changed in a range from AR = 0.75 to AR = 4 for including different street morphologies (wide, regular and narrow). Then, the influence of air speed is tested by means of fixing the aspect ratio of the street and varying the inlet air speed from very low (0.15 m/s) to high (4 m/s).

This study will provide very valuable information about the effect of AR and Re on the ventilation results and streamline patterns of a street canyon, leading to a full characterization of this process so a catalog can be collated with the resulting indicators. This is valuable for the design of climatic-adaptable cities, as it forecasts the ventilation efficiency and heat dissipation of different street morphologies, so urban layouts with less ventilation and therefore more heat accumulation can be avoided to increase inhabitants’ wellness. In addition, this can be useful when designing strategies for the adaptation of cities to climate change, as many of the indicators provided (for instance, the incoming flow rate) are usually inputs for outdoor acclimatization techniques where the focus is to reduce the air entering a specific zone or street for cooling the air with different methods.

2. Materials and Methods

To provide useful quantitative and qualitative information about street ventilation, it is necessary to define all the features affecting this CFD study. Firstly, a discussion about the configuration of Fluent is given, highlighting the most appropriate option set among all the possibilities Fluent provides. Secondly, the different parameters related to flow within street canyons are introduced to establish a characterization among all the cases studied. Then, the indicators of interest in this paper are introduced, as obtaining them for all the cases is the main interest of this work. Eventually, the general script of the simplification and optimization process of the numerical domain used in this work is introduced, but its results are discussed in the Section 3. The general framework of the methodology is represented in Figure 1.

2.1. Numerical Model

Conducting studies and research using CFD simulations involves choosing a number of parameters which directly affect obtaining a result resembling reality. Throughout all the previous research by different scientists, several configuration sets have been proposed to solve efficiently the turbulent constitution of this problem.

The chosen CFD software in this research is Fluent 2023 R1 [49], widely spread in the scientific community because of its calculation capacity and its versatility. To solve the turbulent nature of the problem, RANS equations have been selected, more specifically the k-ε RNG closure scheme model. According to previous studies from the scientific literature [42], this model has been proven to be the one whose results best resemble the experimental data in these specific problems. Its use has been compared to more complex models like large eddy simulation (LES) which usually obtains better results. However, researchers have exposed that both results are very alike, while RANS simulations have great savings in time and computational cost, making RANS k-ε RNG a good choice for studying flow in street canyons [50].

The incompressible Navier–Stokes equations in isothermal conditions are employed, which consist of continuity Equation (1):

$${\displaystyle \frac{\partial {\overline{u}}_i}{\partial x_i}}=0$$

(1)

and momentum conservation Equation (2):

$${\overline{u}}_j{\displaystyle \frac{\partial {\overline{u}}_i}{\partial x_j}}=-{\displaystyle \frac{\partial \overline{p}}{\partial x_i}}+{\displaystyle \frac{\partial }{\partial x_j}}\left(\nu {\displaystyle \frac{\partial {\overline{u}}_i}{\partial x_j}}\right)+{\displaystyle \frac{\partial R_{ij}}{\partial x_j}}$$

(2)

Both equations are written in tensor notation in which indices have value i, j = 1, 2 due to the bidimensional nature of the problem. Because of this, x₁ and x₂ stands for the x-y cartesian coordinates, ${\overline{u}}_i$ are the average air velocities in both directions of space, $\overline{p}$ is the average kinematic pressure and ν is the kinematic viscosity of the fluid. In addition, R_ij is the Reynolds stress tensor which is defined in Equation (3):

$$R_{ij}={\nu }_t\left({\displaystyle \frac{\partial {\overline{u}}_i}{\partial x_j}}+{\displaystyle \frac{\partial {\overline{u}}_j}{\partial x_i}}\right)-{\displaystyle \frac{2}{3}}{\delta }_{ij}k$$

(3)

where ν_t is the turbulent viscosity of the fluid, δ_ij is delta Kronecker function and k is the turbulent kinetic energy (TKE). In addition to these equations, the k-ε renormalization group model RNG adds two more equations, one for TKE transport Equation (4):

$${\displaystyle \frac{\partial \left(k{\overline{u}}_i\right)}{\partial x_i}}={\displaystyle \frac{\partial }{\partial x_i}}\left({\alpha }_k{\nu }_{eff}{\displaystyle \frac{\partial k}{\partial x_i}}\right)+P_k-\mathsf{\epsilon }$$

(4)

and the other for TKE dissipation rate Equation (5):

$${\displaystyle \frac{\partial \left(\epsilon {\overline{u}}_i\right)}{\partial x_i}}={\displaystyle \frac{\partial }{\partial x_i}}\left({\alpha }_{\epsilon }{\nu }_{eff}{\displaystyle \frac{\partial \epsilon }{\partial x_i}}\right)+C_{\epsilon 1}{P}_k{\displaystyle \frac{\epsilon }{k}}-\left[C_{\epsilon 2}+{\displaystyle \frac{C_{\nu }{\mu }^3\left(1-\frac{\eta }{{\eta }_0}\right)}{1+\beta {\eta }^3}}\right]{\displaystyle \frac{{\epsilon }^2}{k}}$$

(5)

where P_k is the kinematic turbulent energy production, both α_k, α_ε are the inverse Prandtl numbers and ν_eff is the effective kinematic viscosity expressed as ${\nu }_{eff}=\nu +{\nu }_t$ with turbulent viscosity obtained from ${\nu }_t=C_{\nu }{\displaystyle \frac{k^2}{\epsilon }}$ and $\eta =Sk/\epsilon$, where S is the scalar measure of the deformation tensor. Eventually, the turbulent production expression can be expressed in Equation (6):

$$P_k={\nu }_t\left({\displaystyle \frac{\partial {\overline{u}}_i}{\partial x_j}}+{\displaystyle \frac{\partial {\overline{u}}_j}{\partial x_i}}\right){\displaystyle \frac{\partial {\overline{u}}_i}{\partial x_j}}$$

(6)

In order to solve these equations, a few more parameters must be defined in the Fluent environment: $\left\{C_{\epsilon 1},C_{\epsilon 2},C_{\nu },{\eta }_0,\beta \right\}=\left\{1.42;1.68,0.0845;4.38;0.012\right\}$ [51].

Regarding the rest of the configuration of the Fluent environment, the SIMPLE scheme has been chosen to solve the coupled equations of pressure and velocity, using a least squares cell-based discretization scheme for the gradient. Furthermore, a second-order discretization is used for the rest of the magnitudes to achieve better accuracy, using the hybrid initialization to start the simulation. Finally, when all the residuals are below the value of 10 × 10⁻⁵, convergence is considered to have been achieved, and the simulation is stopped.

2.2. Parameters

There are several parameters influencing this type of simulation, but the most important one for the characterization of this phenomenon is the street height-to-width aspect ratio (AR) expressed as $AR={\displaystyle \frac{H}{W}}$. According to that definition, low-AR streets represent wide streets or avenues, while high ARs mean narrow streets. The classification of flow regimes can be established according to the AR [24], distinguishing four behaviors of the flow in a street canyon: isolated roughness flow (AR < 0.3), wake interference flow (0.3 < AR < 0.7), skimming flow (0.7 < AR < 1.5), multi-vortex flow (AR > 1.5). Nevertheless, in this study, only the last two of them will be of interest since they are the most common configurations of streets in a city center. The main effect of the AR is changing the mass flow rate into the canyon and the number of vortices that originate. According to the previous classification, in the skimming flow regime, only one vortex may appear, whereas in the multi-vortex flow regime, more than one does. However, some authors outline that this is not only dependent on AR but also on Re, so streets with an AR > 1.5 might show only one vortex [33].

The Reynolds number is the other main parameter regarding this topic. In former investigations, it was assumed that the results obtained from scale-reduced experiments were easily scalable to other measurements or air speeds. However, the latest research has shown that Re greatly affects the flow pattern mainly by altering the number and position of vortices for a fixed AR. Having described Re as a function of the reference air velocity and the characteristic length of the model, there are two parameters which can change the Re number. Due to this, some authors have found that by changing the inflow velocity or the scale of the model, the result was not the expected one according to the base classification of the regimes. For instance, refs. [33,34] obtained a lower number of vortices than expected in high-AR streets both in real experiments and simulations because they were carried out at a high Re. This means that in order to fully characterize the flow inside a street canyon, there must be studies conducted not only for different ARs but also for different Re regimes. Nevertheless, some other studies have pointed out the existence of a critical Re. In theory, for Re > Re_c, results do follow the same pattern, so they do not change with other parameters. That limit is commonly set as Re_c = 11,000 [52], but some authors [33] have proven that this value is highly dependent on the geometry of the model, scale and AR, showing that for a high AR, Re_c increases. For this reason, a detailed study of this concern is given later.

These are the most basic parameters affecting this phenomenon. Firstly, an AR sweep is performed to characterize the airflow inside the street canyon with respect to its morphology, starting from wide streets (AR = 0.75), typically in avenues or less-crowded urban areas with low buildings, and ending with very narrow ones (AR = 4), as is usual in city centers with tall buildings. After this, new cases are studied where the AR is fixed, and Re is changed by means of the reference air speed velocity. This reference speed is defined as the measured wind speed of the flow which is not disturbed by city buildings, which can be obtained from meteorological information. The range of V_ref varies from low speed (V_ref = 0.15 m/s) to high (V_ref = 4 m/s). Moreover, to enrich the study and check the behavior of different street morphologies, the study is conducted on three different AR values: low (AR = 0.75, wide streets), intermediate (AR = 1.5, regular streets) and high (AR = 3, narrow streets).

A summary of the cases studied is in

Table 1. List of cases studied.

	AR	Vref [m/s]	Re
AR characterization	0.75–4	1	6.7·105 (typical value of Re for a three-floor building)
Reynolds number characterization	0.75	0.15–4
	1.5		3.4 × 105–2.8 × 106
	3

.

2.3. Indicators

The first indicator is the flow streamlines within the canyon, which represent the pattern describing the air as it moves inside it. In addition, it is useful to easily visualize air recirculation and where its origin is located. The analysis of the number and position of these vortices is fundamental to obtain full knowledge about this phenomenon.

Another indicator is the air mass flow rate (ACH) in units of kg/s. This indicator is an illustrative measure of the amount of air which enters the canyon and consequently the air renewal inside the street. Theoretically, the expression of the inlet air mass rate for a steady state is Equation (7):

$$ACH=\rho {{\int }_0^{\mathsf{\Gamma }}\left.w^{-}\right|}_{roof}d\mathsf{\Gamma }$$

(7)

where ρ is the density of the fluid, ${\left.w^{-}\right|}_{roof}$ is the vertical velocity of the fluid at the entrance surface of the canyon and ᴦ is the area of that surface. Nevertheless, due to the bidimensional discrete nature of the CFD calculations, this expression must be transformed into another one for allowing the use of the discrete data extracted directly from Fluent. The expression for ACH is in Equation (8):

$$ACH=1[\mathrm{m}]·\rho {\sum }_0^{k=N}∆{\overline{w_k}}_{roof}·∆x_k$$

(8)

This expression is multiplied by 1 m because it is the virtual depth Fluent uses to define fluxes in bidimensional geometries. The integral is also changed to a summation because of the discrete nature of the problem. Here, $∆x_k$is the size of the k cell located in the entrance line of the canyon (this means the distance between two adjacent nodes), and $∆{\overline{w_k}}_{roof}$ is the mean vertical velocity between the two nodes from one cell.

After this definition, this indicator can be changed to other useful units for later studies and applications. Firstly, it can be changed to dimensionless by eliminating its dependence on scale, fluid and velocity. This is named the dimensionless ventilation (ACH_ad), and its expression is in Equation (9):

$$AC{H}_{ad}={\displaystyle \frac{ACH}{1\left[\mathrm{m}\right]·\rho ·(H/AR)·V_{ref}}}$$

(9)

This form of presenting the result is very useful when it comes to comparing it with different values of the parameters, making it easier.

Another unit of interest is the renewal/hour, which is specially used when describing indoor air renovation, but its use can be extended to outdoor ventilation [53]. This can be used as an approximation of the number of times the air within the street is completely renewed in a given period of time. This change in units is named ren/h, and its expression is in Equation (10) with units of h⁻¹:

$$ren/h={\displaystyle \frac{ACH·3600\left[\frac{s}{h}\right]}{\rho ·A·1\left[\mathrm{m}\right]}}$$

(10)

where A is the area that encloses the canyon of the urban canyon. Due to the bidimensional constitution of this research, it can be obtained from $A=H·W$.

The last indicator of interest is an innovative approach to calculate the portion of the air inflow which comes from outside the street canyon and not from the vortex itself. It is called the equivalent air mass flow rate (ACH_eq) together with its variations, ACH_ad,eq and Renh/h_eq. To obtain its expression, an energy balance must be performed. This approach is similar to other authors [54,55], but they used a mass balance equation as they studied pollutant dispersion. Assuming a uniform mean temperature in the street, the energy balance in a bidimensional and discrete formulation is expressed in Equation (11):

$${ACH}_{eq}·C_p·{(T}_{ext}-T_{area})=\rho ·C_p·{\sum }_0^{k=N}∆{\overline{w_k}}_{roof}·∆x_k·(\overline{T_k}-T_{area})$$

(11)

where T_k is the mean temperature of the air in cell k, and T_area is the mean temperature in the street, which are obtained directly from the simulation solving the energy equation of Fluent. The first temperature is caused by the movement of the air itself, which is responsible for the mixture of the temperature of the external air (far from the street) and the mean temperature in it. Now, ACH_eq is defined as the flow rate at the same temperature as the undisturbed flow (T_ext) obtained from Equation (12), which is the result of the energy balance in the urban canyon. This is a quality indicator because it measures the effective net air renewal in the street. This is highly valuable for street ventilation, as it provides the real airflow which enters the street with external properties. For instance, when studying heat accumulation in streets, values of ACH_eq closer to ACH mean that the air is being more effectively renewed, dissipating more heat.

$${ACH}_{eq}(\mathrm{k}\mathrm{g}/\mathrm{s})=1[\mathrm{m}]·\rho {\displaystyle \frac{\sum _0^{k=N}∆{\overline{w_k}}_{roof}·∆x_k·(\overline{T_k}-T_{area})}{T_{ext}-T_{area}}}$$

(12)

Although the phenomenon of ventilation in street canyons has been studied before, in this paper, the quantitative values of the air renewal and streamline flow pattern are provided for a wide variety of cases so they can be used in subsequent real projects. Furthermore, this paper introduces a strategy for obtaining the part of air renewal caused by external air and not by the vortex itself. This is very important not only for outdoor acclimatization strategies but also for pollutant dispersion studies.

2.4. Computational Domain

A geometrical model is needed in order to solve the concerning CFD problem. Traditionally in this field, the most common street canyon geometry model is the one seen in Figure 2, where there are several canyons before and after the one of interest to simulate a virtual roughness in the upstream and downstream, increasing the turbulent intensity of the fluid [56]. It is an acceptable way for scientific research, but it tends to be quite big with a very high number of cells in it, increasing the computational cost and the simulation time. For research whose aim is to obtain illustrative results for later application in projects, it is desirable to have an enhanced model which is easy to replicate and computationally efficient, sacrificing some accuracy in the results.

Because of this, the first part of this paper focuses on designing an approach to optimize the numeric domain for this study. Some authors compared the effects on the results when using different geometrical models but did not make any optimization method [57]. Another important part of the computational domain is the definition of the mesh. A comparative study of different mesh sizes is also included, in order to reach an enhanced cell size value so the result is accurate enough, but the simulation time is not excessive.

By eliminating the consecutive canyons, the number of cells of the meshing will be significantly reduced, as well as the simulation time. At first, the geometry model has been simplified to a single-canyon model in which the upstream length is long (25 H). This, along with the roughness, allows the flow enough path for it to fully develop to resemble the chaotic nature of flows in cities. The downstream length is also long (10 H) but not as long as the other one, as the fluid recovers from the effect of the canyon faster. For the optimization process, a street with reference height H = 10 m is used with AR = 1 and a uniform velocity profile of V_ref = 1. The geometry used is in Figure 3. Different options have been tested to prove which one is the best solution for this aim. The first one includes roughness in the upstream wall; the second one gives the inflow velocity profile a certain expression to simulate the UBL.

To ensure the development of the flow and increase its turbulence, a certain roughness value has been added to the upstream and downstream walls. This is called aerodynamic roughness (z₀), which is an empirical value obtained from some authors [58] who sought to assign a numerical value of roughness to typical landscape configurations for modeling them in mathematical applications. With this, the urban city skyline can be modeled, and the roughness can be simulated in CFD simulations. For this research, an aerodynamic roughness value of z₀ = 0.1 m is considered as an initial guess. However, this parameter must be changed to another one that Fluent can read, known as equivalent sand-grain roughness. The relation between both is in Equation (13) [58]. There, C_s is the roughness uniformity value. When its value is one, it means that the roughness is chaotic and does not have a pattern, which is the case of the urban landscape. Other authors [59] have also used this parameter in their studies.

$$k_s={\displaystyle \frac{9.793z_0}{C_s}}$$

(13)

Once this model is fully characterized with roughness parameters, the reference simulation is performed, and its results are compared with the multiple-canyon model. Then, the next step is to shorten the upstream and downstream wall lengths. For this purpose, an optimization of these lengths is carried out, similar to the study conducted by Abu Zidan [60], reducing progressively from 25 H for the upstream wall and 10 H for the downstream wall to a value of 3 H for both. Then, the airflow rates and streamline pattern results are checked to figure out which geometry is the most appropriate for obtaining good-quality results and a low simulation time.

The last thing to do is change the inlet profile velocity of the fluid. In this research, two expressions for the velocity profile are considered; both are recommended to simulate the UBL [23] and depend on the aerodynamic roughness. The first one is the exponential law in Equation (14), where the exponent directly depends on the aerodynamic roughness [23]; in this study, its value is 0.16.

$$U=U_{ref}{\left({\displaystyle \frac{y}{H}}\right)}^{0.16}$$

(14)

The second one is the logarithmic law in Equation (15). These expressions recreate the roughness sublayer of the UBL, as shown in Figure 4.

$$U=\frac{u_{\ast }}{K}\mathrm{l}\mathrm{n}\left({\displaystyle \frac{y+z_0}{z_0}}\right)$$

(15)

where $u_{\ast }={\displaystyle \frac{K\times U_{ref}}{\mathrm{l}\mathrm{n}(\frac{H+z_0}{z_0})}}$, and K = 0.4 is a constant.

Eventually, a comparison of the results with both profile velocity expressions is performed, and a decision about which one best fits the reference result is made. Once this is done, the definitive enhanced geometry model is achieved, reducing the size of the numerical domain.

3. Results

In this section, the results of the optimization process are stated, with an eventual enhanced model of a single canyon. Before starting the process, several cases are tested with a multiple-canyon geometrical model in order to ensure that the Fluent environment configuration is correct. For this aim, the simulation cases included in the research of W. C. Cheng et al. [62] (which are based on the experimental research of J. Baik [63]) have been recreated, using a model with Re = 12,000. When comparing the results of the recreation (Figure 5) and theirs, it is noticeable that both are almost identical, so it can be assured that the configuration set of Fluent is correct, and the optimization can begin.

3.1. Numerical Domain Optimization

The first thing to do is checking if the approximation of the single-canyon model with a long upstream wall length with roughness is appropriate, so two simulations have been conducted, one with consecutive canyons and the other one with one canyon with roughness, both with AR = 1, H = 10 m and V_ref = 1 m/s. The features that are compared are the ACH value and the flow streamlines. For the reference multiple-canyon model, ACH = 0.129 kg/s, whereas for the proposed single-canyon model, ACH = 0.133 kg/s. Both results are very similar, as well as the streamline patterns. In addition, the vertical velocity profile of the fluid on the entrance line of the canyon (Figure 6) is similar in both cases. Because of this, the approximation without consecutive canyons but with roughness is appropriate for recreating this phenomenon.

The next step is shortening the upstream wall length as mentioned before, changing the upstream wall length from 25 H to 3 H and the downstream wall length from 10 H to 3 H, and comparing them with the reference solution (25 H) maintaining the uniform inlet velocity profile. Because all the airflow rate indicators are related, comparing one of them is enough to know the behavior of the others. In this case, the results of ACH_ad are compared in

Table 2. Comparison of ACH_ad when reducing the upstream wall length.

	25 H (ref)	10 H	9 H	8 H	7 H	6 H	5 H	4 H	3 H
ACHad (×10−2)	1.09	1.19	1.21	1.22	1.25	1.27	1.30	1.34	1.39

.

It is noticeable that the shorter the length, the higher the ACH_ad diverged from the reference. However, the tendency is not the same. For very short lengths (less than 5 H), when one step is taken, the error sharply increases, whereas for relatively long lengths (more than 7 H), an incremental step does not involve a great reduction in the ACH_ad. However, the value of 6 H provides an accurate enough solution while maintaining a low value of the upstream length, because from that value, increasing the length does not significantly improve the solution. Because of this, this value is fixed for both upstream and downstream wall lengths. This reduction leads to a great reduction in computational cost and simulation time.

After this reduction, the following step is testing which one of the two possible inlet velocity profiles is the most appropriate for comparing each solution. With both methods, the air ventilation is underestimated, ACH = 0.1041 kg/s for the logarithmic law, and ACH = 0.1184 kg/s for the exponential law. However, it is remarkable that with the second one, the values obtained are closer to the reference one (deviation of ~10%); in fact, the error committed when using this model is less than half of the error when using the logarithmic expression (deviation of ~22%). Due to these results, the exponential profile is used in the following study.

Furthermore, the meshing of the numeric zone plays a crucial role in CFD simulations, as the results are highly dependent on a good-quality mesh. Different strategies have been used in this paper to achieve this:

(1)

Quadrilateral cells. When the geometry is not very complex, this is the preferred one because it usually provides better mesh quality.

(2)

The cell size is dependent on the height reference. This means that the mesh cell size is proportional to the scale of the model, preventing excessive growth.

(3)

The cell size is dependent on the AR. If it was constant for every AR, the number of cells in the entrance line of the canyon would be very high for wide streets but very low for narrow streets, which could lead to accuracy problems in the latter.

(4)

The numeric domain is split into two areas. The “inner zone” is the one related to the street canyon itself and the surrounding area. Here, the mesh cell is finer because it is the area where the most accuracy is needed to obtain good results. On the other hand, the “outer zone” is the rest of the domain where the meshing is coarser because no results are obtained there.

Once the meshing method is explained, in this research, four different mesh refinement have been tested in order to obtain a grid with low computational cost but with good results. This optimization process has been conducted by some authors but for other configuration sets, for example, using LES instead of RANS [48]. This method is applied to the geometry in Figure 3. The results obtained from each of them are listed in

Table 3. Mesh refinement ventilation results.

	Element Size (Face Meshing “Outer Zone”)	Element Size (Face Meshing “Inner Zone”)	ACH (kg/s)
Very coarse grid	H/10	H/(25 × AR)	0.1274
Coarse grid	H/13.33	H/(50 ×AR)	0.1298
Fine grid	H/20	H/(100 ×AR)	0.1335
Very fine grid	H/40	H/(200 ×AR)	0.1346

. The ventilation results from the first two meshes do vary from the fine grid. Furthermore, both fine meshes have almost identical results, so in this case, increasing the number of cells does not improve the solution accuracy. The meshing used in the following study is the fine meshing because it is the optimal cell size as the solution does not change when decreasing the size.

This optimization process has eventually come to an end. The enhanced numeric and boundary conditions are shown in Figure 7. The final meshing corresponds to the “fine grid” in Table 3. With this mesh, the number of elements for AR = 1 is 81,402 cells. This model is used in the rest of this work.

3.2. AR Characterization

The goal of this paper is to provide useful air renovation results for very different street morphologies, which can be used in subsequent acclimatization projects. For this, an AR characterization study of the street canyon phenomenon is conducted, changing this parameter from low values (AR = 0.75) to high (AR = 4) so a wide range of cases are included. The height building is fixed at H = 10 m for every case, so the upstream and downstream wall length and numeric domain height are also fixed. The inflow velocity is considered V_ref = 1 m/s (Re = 6.7·10⁵).

The first indicator of interest is the streamline pattern of the flow, which is represented for each AR in Figure 8. For low and intermediate ARs from AR = 0.75 to AR = 1.5, there is a single vortex in the centroid of the canyon. Then, from AR = 2 to AR = 2.5, the center of the vortex moves toward the roof level. Eventually, for narrow streets (AR > 3), multiple air recirculations are formed. First, the size of the bottom vortex is very small compared to the other one, and its center is very close to the floor. However, when increasing the AR, its size grows, and its center displaces upward, while the center of the upper vortex remains fixed. Eventually, for AR = 4, a new vortex at the bottom appears. This result is fundamental when understanding the air movement inside street canyons because it allows to split them according to their recirculations. These results may not be coincident with the previous results in the literature. This is because of the Reynolds number range used. In this work, a typical street configuration is used, using a three-floor building as the reference. This leads to a Reynolds number of approximately 6.7·10⁵, far from the typical value of approximately Re = 12,000 used in the literature [20,31,50]. However, the results in this work are similar to those of other authors who have tested the dependence of flow with the Reynolds number, using values similar to the one in this study [17,30,33]. They also concluded that for an AR which traditionally was supposed to have multiple vortices (AR = 2), only one vortex is formed for typical Reynolds number values.

The next step is checking the air ventilation results, as shown in

Table 4. Ventilation results for the AR study.

AR	ACH (×10−1 kg/s)	ACHeq (×10−1 kg/s)	ACHad (×10−2)	ACHad,eq (×10−2)	Ren/h (h−1)	Ren/heq (h−1)
0.75	1.876	0.961	1.15	0.59	4.14	2.12
0.85	1.639	0.833	1.14	0.58	4.09	2.08
1	1.184	0.637	0.97	0.52	3.42	1.87
1.25	0.682	0.444	0.72	0.45	2.51	1.63
1.5	0.412	0.280	0.50	0.34	1.83	1.23
1.75	0.277	0.211	0.40	0.30	1.43	1.08
2	0.238	0.162	0.39	0.27	1.4	0.95
2.5	0.185	0.136	0.39	0.28	1.39	1.01
3	0.154	0.110	0.38	0.27	1.36	0.97
3.5	0.132	0.094	0.38	0.27	1.36	0.97
4	0.116	0.087	0.38	0.28	1.36	1.01

. There is a general tendency in the results: ACH decreases when AR increases. This agrees with the result obtained by other authors qualitatively [38]. From AR = 0.75 to AR = 1.75, the ACH reduction is more noticeable, up to the point of reducing its value by one order of magnitude, from ACH~0.2 kg/s to ACH~0.02 kg/s approximately. Nevertheless, from AR = 1.75 to AR = 4, the tendency changes with a smoother diminution maintaining the same order of magnitude, varying from ACH~0.02 kg/s to ACH~0.01 kg/s. This implies an air renovation standstill in narrow streets, which means that in practice, it is roughly independent of AR, providing AR > 1.75. This phenomenon is easily perceptible when looking at ACH_ad, whose value is practically identical in those cases.

Furthermore, when looking at the ACH_eq, it is noticeable that its value is approximately half of ACH for low-AR streets. This means that the air inflow is roughly equally formed by external air and by the vortex itself. However, for high-aspect-ratio streets, this value is around 75% of ACH. This implies that in narrow streets proportionally more air is coming from outside the canyon. This effect can be seen in Figure 9, where for a low AR up to AR = 1.75, the difference between ACH and ACH_eq is significantly bigger than for a high AR. It is also remarkable that both indicators have similar trends, lessening faster in wide streets and then coming to a standstill for a high AR.

Additionally, in the multi-vortex flow regime, the general volume of the street canyon can be fragmented, taking as a reference the volume occupied by each one of them. Once this is done, an illustrative value for the air renovation in each zone can be obtained by applying the same expression as for the previous cases, but this time over the entry line to the respective volume, as shown in Figure 10. This finding is very important as it provides an illustrative measure of the air renovation in the bottom part of a street where pedestrians and citizens stand. As it shows, the more vortices found, the lower this value is, which can cause severe heat or pollutant accumulation, threatening people’s health. The value of the air exchange between different vortices is significantly lower than the air exchange of the upper vortex with the exterior, as they are two orders of magnitude lower.

3.3. Wind Speed Characterization

Eventually, in order to provide the full results for its use in practice, information for air ventilation with different wind speed conditions is obtained. Three representative values of AR have been selected: AR = 0.75 for streets whose width is longer than the height, AR = 1.5 for others whose width is shorter but roughly the same as the height and AR = 3 for narrow ones whose height is much longer than the width. In each case, V_ref varies from 0.15 m/s to 4 m/s.

Firstly, the streamline flow patterns are discussed. Both results for AR = 0.75 and AR = 1.5 have the same trend (Figure 11). For the entire range studied, only one vortex is formed, and it is fixed at the centroid of the street canyon. The only effect wind speed has is smoothing and eliminating corner vortices, which can be found for low V_ref and then disappear. This means that the airflow pattern in streets with an AR < 1.5 does not change when the Reynolds number does, which is beneficial as it prevents inaccuracy or different behaviors of the flow due to the forming of unexpected vortices.

Nevertheless, the results for AR = 3 show that narrow streets have a different trend (Figure 12). It is noticeable that the streamline pattern is highly dependent on wind speed. At first, several vortices are formed, three for very low speed and then two. When increasing V_ref, the bottom vortex reduces until it is transformed into a corner vortex. Then, the same effect happens to the former intermediate vortex, reducing its size, until for V_ref =2 m/s, only one vortex is left whose center has remained constant through the wind speed range. This center is located near the roof level and not at the canyon centroid. This implies that a little uncertainty in wind speed magnitude may trigger crucial changes in airflow pattern because of the formation of unexpected vortices.

Finally, all three air ventilation results can be grouped because they have the same trend. The ACH value linearly increases as wind speed does (

Table 5. Air ventilation results for different ARs and Reynolds numbers.

	${\mathbf{V}}_{\mathbf{ref}}$	0.15	0.5	0.75	1	2	3	4
$\mathrm{AR}=$ 0.75	ACH (×10−1 kg/s)	0.26	0.90	1.37	1.91	3.82	5.80	7.85
	ACHad (×10−2)	1.02	1.10	1.12	1.17	1.17	1.18	1.20
	Ren/h (h−1)	0.56	1.98	3.02	4.21	8.41	12.79	17.30
$\mathrm{AR}=$ 1.5	ACH (×10−1 kg/s)	0.06	0.20	0.30	0.41	0.86	1.31	1.70
	ACHad (×10−2)	0.47	0.50	0.50	0.50	0.52	0.53	0.54
	Ren/h (h−1)	0.26	0.89	1.32	1.81	3.79	5.77	7.49
$\mathrm{AR}=$ 3	ACH (×10−2 kg/s)	0.24	0.81	1.15	1.54	3.06	4.67	6.03
	ACHad (×10−2)	0.41	0.38	0.38	0.38	0.37	0.38	0.37
	Ren/h (h−1)	0.20	0.67	1.02	1.36	2.69	4.11	5.31

). This effect can be easily seen by looking at ACH_ad because when eliminating the velocity term its variation is very low; in fact for AR = 3, it remains practically constant. This means that for a first approximation, the air renovation results for a given wind speed can be rounded as in Equation (16):

$$ACH\left(V_{ref}\right)=ACH\left(V_{ref}=1\right)\times$$

(16)

This finding is crucial because it saves time as it is not necessary to have different simulations for each wind speed of interest, with the solution for V_ref =1 being enough to obtain an approximate result.

4. Discussion

The movement of air in spaces due to depression in the external areas of architectural environments is one of the most complex points. The inherent complexity of air movement in open spaces is even greater due to the diversity of geometric shapes that characterize the environment. In the present study, the main objective of the depression confinement technique is to reduce the entry of unwanted air and, together with the natural conditioning of the area, to achieve its thermal comfort. All of this is in order to achieve the recovery of life on the street through the creation of comfortable areas. Therefore, in this section, we will investigate the typical pattern of changes in the airflow caused by different geometric typologies, and a simplified characterization of the technique will be carried out in order to develop the knowledge base of it and allow its integration into the decision-making processes in the design phase of the rooms.

For the study of air movement in confined spaces due to depression when the flow is transverse or oblique to it, it is necessary to look for the parameters that directly influence the problem, so that the results can be extrapolated to as many cases of interest as possible. In order to determine the influence of all the parameters that characterize the variation in the behavior of the flow pattern, a bank of simulations is carried out using CFD (computational fluid dynamics) numerical methods. Regarding the design of an enhanced numerical domain, a simplified model has been obtained whose results are highly similar to those of the multiple-canyon model used by many authors in this field. Thanks to his model, computational cost and simulation time are highly reduced, which is very valuable for project applications where lots of cases must be simulated or where fast illustrative results are needed.

When conducting all the cases in order to establish a catalog where the main indicators are obtained from a specific AR and Reynolds number, the AR characterization has provided lots of useful information. Firstly, having represented streamline patterns for each case has helped identify the number and position of vortices formed in a street canyon. Different behavior is shown depending on the narrowness of the street. In those where the width is the same order of magnitude as the height, only one vortex is formed in the centroid of the area. However, in narrow streets (approximately for AR > 1.75), a new vortex-formation process starts, displacing the existing one toward the roof level, and eventually a multi-vortex regime occurs for an AR > 3. This allows the segregation of the space in different volumes coincident with each vortex to create different acclimatization strategies. These results may not be the same as previous research where the Reynolds numbers are considerably lower [20,50], as in this work, a building of 10 m height and a reference speed of 1 m/s is tested, giving more appropriate values of the Reynolds number (approximately 6.7·10⁵), which are more similar to the study conducted by Chew et al. [33]. In consequence, the number of vortices found in narrow streets is higher than expected. This information is crucial when understanding this phenomenon as it means that narrower streets lead to poorer ventilation. This is negative for heat and pollutant dissipation, but it is beneficial for acclimatization techniques because less air enters and increases the temperature of the cooled area.

Furthermore, numerical ventilation results provide an illustrative value for ACH for each street configuration. Two different trends are shown in Figure 9. For 0.75 < AR < 1.75, the incoming air mass flow rate decreases rapidly with AR, from ACH~0.2 kg/s to ACH~0.02 kg/s, which is one order of magnitude. However, for AR > 1.75, a ventilation standstill occurs, and the ACH variation is much smoother than before, reducing from ACH~0.02 kg/s for AR = 1.75 to ACH~0.01 kg/s for AR = 4. This effect can be easily seen when looking at ACH_ad and Ren/h as they remain roughly constant at ACH_ad = 0.0040 and Ren/h = 1.00. Another important result is that this last parameter is five-times lower than AR = 0.75, which means that in fact the air renewal in narrow streets is much smaller than in wide streets. When designing acclimatization methods, this is valuable, as it reduces unwanted hot air entering the street, enabling the opportunity to use different techniques to cool the space and improve thermal comfort. This information is very valuable not only when talking about outdoor acclimatization strategies as the fact that more than one recirculation is formed makes it difficult to eliminate pollutants concentrated at the street level. According to the results, the more vortices there are in a street, the less ventilation performance the bottom layer of the street has, which contributes to the effect mentioned before.

In addition, the definition of the equivalent air mass flow rate has allowed for a distinction between the air inflow due to the vortex itself and that from the outside, which is extremely important because the temperature of air from the recirculation is like the one found within the canyon, which is not desired when designing strategies for tempering an outdoor space. This study provides an order of magnitude result for ACH_eq, which is proven to be higher in proportion for an AR > 1 (nearly 75%), whereas in wide streets (AR < 1), approximately half of the air comes from outside the street canyon. This is vital when it is applied to acclimatization methods, as a higher value of ACH_eq means that more hot air is entering the cooled zone, thus decreasing the cooling effectiveness.

On the other hand, the Re variation study has allowed to obtain the dependence of the air ventilation with wind speed. It is proved that the relation between both is almost linear for each AR tested, which makes it easy to scale the results to any air velocity of interest within the range studied. Regarding the streamline pattern, for high-AR streets, the number and position of recirculations are highly dependent on the wind speed. For low V_ref, new ones which were not expected may appear, and while V_ref increases, they change in size and displace toward the bottom of the street canyon. This can cause only one vortex to be found in high-AR streets. This finding is fundamental because according to other authors [17,33], the airflow pattern does depend on the Reynolds number for deep street canyons. This means that the previous work from the literature that was carried out with reduced-scale models might be insufficient to explain the real behavior of flow across streets. This work supplies that lack of information and present information about the number of recirculations formed in a narrow street when the exterior wind speed changes.

With these findings, the street canyon ventilation phenomenon is completely characterized. The ventilation performance of a street is provided for each street morphology (from narrow to wide ones) and for different external wind speeds. This can be used when studying heat dissipation because it gives the amount of air entering the street and then renovated. On the other hand, this study is extremely useful for outdoor acclimatization methods. The results show that in order to isolate a street or an outdoor area, narrower morphologies are better as they have a lower air renovation, and then the techniques used to cool the area are more effective because less hot external air is going to disturb the space. This helps in the decision-making process when designing these strategies, providing all the quantitative information necessary for their implementation.

Limitations

The simplification to a bidimensional model of a street canyon induces inaccuracy to the results when using it in real-life applications. This is because of the complex nature of the airflow within three-dimensional streets, causing corner recirculations and zones where unexpected recirculations are formed, impulsing air with more force in or outside the street. The effect of the 2D simplification was studied by some authors [19,35] who concluded that the results for the 2D simulations were acceptable for long streets, obtaining very low errors. That is why this simplification is widespread among the scientific community, providing good illustrative results. In addition, the wind profile is supposed to be perpendicular to the street axis, so the effect of crosswinds is uncertain. However, this hypothesis is the most unfavorable for street ventilation, obtaining the worst ventilation results. For this reason, this is the case commonly tested.

In addition, only street configurations with an AR between 0.75 and 4 are tested, but there are in fact streets with an AR value outside that range. However, if other morphologies or conditions must be studied, it can be easily achieved using the optimal numerical domain model also introduced in this paper, saving resources and time.

5. Conclusions

To summarize, this paper provides valuable information referring to air ventilation and streamlines patterns within street canyons, for a wide range of cases. Not only have different morphologies been studied but also a fixed one, the variation in its solution a result of changing wind speed conditions. This enables the creation of a database where if numerical information about the air exchange in street canyons is needed for heat dissipation or pollutant removal, it can be obtained knowing the morphology of the street (AR) and the Reynolds number regime (depending on H and V_ref) without further simulations. In addition, the streamline pattern results for different ARs and Reynolds numbers can be used for determining critical urban layouts, as it has been found that narrower streets have poorer ventilation, especially in the bottom vortices formed, causing a more severe heat accumulation. With this information, more resilient cities can be designed to reduce climate change effects (as UHIs) and hazards, as many authors suggest [14,43].

These results can be now used in different real projects related to the acclimatization of outdoor spaces as the air renovation is fully detailed. Different techniques try to reduce the amount of air entering a specific place or street in order to isolate it and cool the air to improve thermal comfort. This is very valuable as streets and public places are where people spend most time when they are outside, so it is needed to address the overheating issue in streets. With the quantitative information about air renewal in this work, lot of time is saved in the initial steps of designing these methods because air exchange is usually one of the main inputs in acclimatization projects. There are some examples of this in real life, for instance in Seville [64,65], where with the combination of several acclimatization techniques, one of them the depression confinement, the desired public space is cooled.

Future Research

This research can still be expanded. Firstly, the Reynolds number characterization has been performed for three representative ARs of each street morphology (wide, regular and narrow). Although the quantitative information about street ventilation can be extended to any AR, it would be very valuable to obtain the airflow pattern dependence with Re for each AR from the range studied in this work. This way, the information would be more valuable for real acclimatization strategies as more information about the number and position of vortices for a wide range of street morphologies would be provided.

In addition, this study can be performed with even more AR values to include different morphologies excluded in this first work, for example, very low ARs representing avenues or open spaces such as squares and very high ARs representing very urbanized areas with tall buildings. This would increase the value of this work.