Effects of Orientations, Aspect Ratios, Pavement Materials and Vegetation Elements on Thermal Stress inside Typical Urban Canyons

The analysis of local climate conditions to test artificial urban boundaries and related climate hazards through modelling tools should become a common practice to inform public authorities about the benefits of planning alternatives. Different finishing materials and sheltering objects within urban canyons (UCs) can be tested, predicted and compared through quantitative and qualitative understanding of the relationships between the microclimatic environment and subjective thermal assessment. This process can work as support planning instrument in the early design phases as has been done in this study that aims to analyze the thermal stress within typical UCs of Bilbao (Spain) in summertime through the evaluation of Physiologically Equivalent Temperature using ENVI-met. The UCs are characterized by different orientations, height-to-width aspect ratios, pavement materials, trees’ dimensions and planting pattern. Firstly, the current situation was analyzed; secondly, the effects of asphalt and red brick stones as streets’ pavement materials were compared; thirdly, the benefits of vegetation elements were tested. The analysis demonstrated that orientation and aspect ratio strongly affect the magnitude and duration of the thermal peaks at pedestrian level; while the vegetation elements improve the thermal comfort up to two thermophysiological assessment classes. The outcomes of this study, were transferred and visualized into green planning recommendations for new and consolidated urban areas in Bilbao.


Introduction
The current world population is expected to reach 8.5 billion by 2030, 9.7 billion in 2050 and 11.2 billion in 2100 [1]. This rapid growth of the world's population means that in the near future, more than half of all people will live in cities and this trend will inevitably have a strong impact on the sustainability and the energy costs of the built environment [2]. As a consequence of the global trend towards rapid and uncontrolled urbanization an increase in the magnitude of urban heat island (UHI) phenomena can be expected, together with an alteration of local patterns [3,4]. The natural and artificial morphology have influence also on the meteorological parameters such as air temperature (T a ), relative The Gran Bilbao metropolitan area hosts around 1 million inhabitants of which more than 340,000 live in the urban area of Bilbao municipality spread over an area of around 16 km 2 [68]. Bilbao (latitude 43.25 • N, longitude 2.96 • W), is characterized by medium-high urban density and it is surrounded by a complex topography which has always affected the urban development of the city and its climate. The climate is humid temperate, with the absence of a dry season and a moderate level of temperature and precipitation during the year (Cfb Oceanic climate according to the Köppen-Geiger climate classification [69]). The highest annual solar radiation is registered in July, when the global horizontal radiation reaches up to 930 W/m 2 as maximum hourly value during the day and more than 4790 W/m 2 as the highest average daily value in a month. The level of RH achieves values higher than 70% during the entire year (above 75% in winter). In summer, the air T a can surpass 30 • C from July until September [70]. The data registered by the Euskalmet [71] during the past decades show that the T a has exceeded repeatedly 40 • C during summertime. The increment of T a in Gran Bilbao area is also confirmed by González-Aparicio and Hidalgo [72]. Their statistical analysis has demonstrated that the heat wave events have affected significantly the people living in urban and sub-urban area of the Basque Country in the last two decades; and in the near future, the magnitude and the frequency of these events will significantly increase both in summer and in winter. According to the future projections of this study [72]. Regarding the warm season, the projections show an expected increment of T a up to 3.5 • C in comparison to previous period  and an increase from 15 days (1978-2000) to 24 days (2020-2050) of total number of days for a single heat wave event [73]. This data is aligned with the predictions in other European studies [74][75][76][77]. In Bilbao, UHI temporal and spatial variations have been also studied [78]. Among other aspects, the airflow patterns, the complexity of the orography and the characteristics of urban morphology (e.g., geometric average of building heights, ratio of building plan area to total plan area, etc.) have a relevant influence in the variations of UHI magnitudes. The hereby-presented study aims to address green planning recommendations for urban decision makers to mitigate thermal stress and impact of future heat wave events inside typical urban canyons in Bilbao.

The Urban Case Study Areas
In line with other studies [73,79], main parameters, such as the height of the building (i.e., geometric average of building heights), the presence of vegetation, the building surface fraction (i.e., ratio of building plan area to total plan area) were used to select the study areas/districts ( Table 1). The urban morphology analysis was conducted to extract the geometric dimensions of the selected districts and the characteristics of the local climate. Two main urban geometric aspects were considered: (1) the building surface area, i.e., the ratio between the surfaces covered by buildings (B) and the total surface (T), known as district's urban density (B/T), and (2) the average values of the buildings' height (H) and the streets' width (W), known as urban canyon's aspect ratio (H/W) ( Table 1). All the measurements were taken from the cadastral virtual office of Biscay [80]. Table 1. Analysis of the current situation of the selected urban case study areas. geometric average of building heights, ratio of building plan area to total plan area, etc.) have a relevant influence in the variations of UHI magnitudes. The hereby-presented study aims to address green planning recommendations for urban decision makers to mitigate thermal stress and impact of future heat wave events inside typical urban canyons in Bilbao.  [78]. a Ratio related to total area of the selected district in Bilbao and green spaces/squares presented in those areas. b Ratio of total land use category area to total urban area in Bilbao. geometric average of building heights, ratio of building plan area to total plan area, etc.) have a relevant influence in the variations of UHI magnitudes. The hereby-presented study aims to address green planning recommendations for urban decision makers to mitigate thermal stress and impact of future heat wave events inside typical urban canyons in Bilbao.  [78]. a Ratio related to total area of the selected district in Bilbao and green spaces/squares presented in those areas. b Ratio of total land use category area to total urban area in Bilbao. geometric average of building heights, ratio of building plan area to total plan area, etc.) have a relevant influence in the variations of UHI magnitudes. The hereby-presented study aims to address green planning recommendations for urban decision makers to mitigate thermal stress and impact of future heat wave events inside typical urban canyons in Bilbao.  [78]. a Ratio related to total area of the selected district in Bilbao and green spaces/squares presented in those areas. b Ratio of total land use category area to total urban area in Bilbao.  [78]. a Ratio related to total area of the selected district in Bilbao and green spaces/squares presented in those areas. b Ratio of total land use category area to total urban area in Bilbao.  [78]. a Ratio related to total area of the selected district in Bilbao and green spaces/squares presented in those areas. b Ratio of total land use category area to total urban area in Bilbao.  [78]. a Ratio related to total area of the selected district in Bilbao and green spaces/squares presented in those areas. b Ratio of total land use category area to total urban area in Bilbao.  [78]. a Ratio related to total area of the selected district in Bilbao and green spaces/squares presented in those areas. b Ratio of total land use category area to total urban area in Bilbao.  [78]. a Ratio related to total area of the selected district in Bilbao and green spaces/squares presented in those areas. b Ratio of total land use category area to total urban area in Bilbao.  [78]. a Ratio related to total area of the selected district in Bilbao and green spaces/squares presented in those areas. b Ratio of total land use category area to total urban area in Bilbao.

Microscale Numerical Modeling of ENVI-met
In this study, ENVI-met v4 [81][82][83], was used to evaluate the evolution of thermal comfort within the urban street canyon. ENVI-met is usually adopted to analyze the interactions surface-plant-air for microclimate analyses for typical horizontal resolution from 0.5 to 10 m, and a period of 24-48 h. For each time step, the atmospheric equations solved by ENVI-met produce output data of typical meteorological parameters, such as T a , RH, T mrt , T s , W s and wind direction (W d ), radiation fluxes (Grad). ENVI-met v4 allows forcing air temperature and relative humidity to consider their evolution along the day and consequently the evaluation of thermal comfort conditions along the diurnal cycle. In this study, the evolution of hourly meteorological data along the 7th of August was set as the background airflow characteristics forcing in the model with the aim of representing summertime conditions in Bilbao and the characteristics of thermal stress levels in the urban area. All scenarios were run with the same boundary conditions to allow an adequate comparison between them. Data was taken from the meteorological station of Deusto (a station of the Basque Meteorological Network), located at latitude 4.28 • N, longitude 2.93 • W in the northern urban area of Bilbao at 3 meters above sea level [71]. The other meteorological data (i.e., W s , cloud cover, etc.) were set constant and are described in Section 3.5.

The Thermal Comfort Index of PET
The PET is based on the Munich Energy-balance Model for Individuals (MEMI) [6], which simulates the thermal conditions of the human body ( Table 2). Table 2. PET level, thermal perception and grade of physiological stress according to [5,84]. On the right, a sample of heat balance calculation with the MEMI in summer (figure modified from [6].).  Table 2. PET level, thermal perception and grade of physiological stress according to [5,84]. On the right, a sample of heat balance calculation with the MEMI in summer (figure modified from [6].). PET is defined as the physiologically equivalent temperature and is equivalent to the air temperature at which, in a typical indoor setting, the heat balance of the human body is maintained with core and skin temperatures equal to those under the conditions being assessed. This way PET enables a layperson to compare the integral effects of complex thermal conditions outside with his or her own experience indoors [6]. It expresses the human thermal comfort in both indoor and outdoor environments using the international standard unit widely known as the Celsius degree (°C) [2]. This makes PET to be fully comprehensible by all the actors involved in the design process with or without technical background. PET gives the measure of thermal comfort considering the meteorological parameters: Ta, RH, Ws, Tmrt. It also takes into account the physics of the human body: gender, height, activity, and clothing resistance for heat transfer, short-wave albedo and long-wave radiation of the surface affected by the physical surface properties [6,20]. In order to classify cold, neutral and heat stress in the urban canopies, the calculated values have been referred to the evaluation scale of Matzarakis et al. [5], that allows estimating the level of human thermal perception based on seven classes defined by [85] for Central Europe.

Measurement Campaigns
Measurements to validate the model performance were carried out during the summertime period in 2011 in an E-W oriented street. The width of the street was 24 m and the aspect ratio was mostly close to 1.0 in all the street canyon. The area is classified as compact midrise [78] although the street aspect ratio is slightly lower than in the selected urban areas (Section 3.1). The figure in Table 3 shows the location (red dot) of the measurements inside the area that was modelled to validate ENVI-met for the purpose of this study. The selection of the point was done aiming to consider a representative value in the street (i.e., sufficiently separated from the building facade and way from street intersections). Ta, RH, WS, wind direction (WD) were measured at 8.5 m a.g.l. with a WXT520 Weather Transmitter (Vaisala,), PET is defined as the physiologically equivalent temperature and is equivalent to the air temperature at which, in a typical indoor setting, the heat balance of the human body is maintained with core and skin temperatures equal to those under the conditions being assessed. This way PET enables a layperson to compare the integral effects of complex thermal conditions outside with his or her own experience indoors [6]. It expresses the human thermal comfort in both indoor and outdoor environments using the international standard unit widely known as the Celsius degree ( • C) [2]. This makes PET to be fully comprehensible by all the actors involved in the design process with or without technical background. PET gives the measure of thermal comfort considering the meteorological parameters: T a , RH, W s , T mrt . It also takes into account the physics of the human body: gender, height, activity, and clothing resistance for heat transfer, short-wave albedo and long-wave radiation of the surface affected by the physical surface properties [6,20]. In order to classify cold, neutral and heat stress in the urban canopies, the calculated values have been referred to the evaluation scale of Matzarakis et al. [5], that allows estimating the level of human thermal perception based on seven classes defined by [85] for Central Europe.

Measurement Campaigns
Measurements to validate the model performance were carried out during the summertime period in 2011 in an E-W oriented street. The width of the street was 24 m and the aspect ratio was mostly close to 1.0 in all the street canyon. The area is classified as compact midrise [78] although the street aspect ratio is slightly lower than in the selected urban areas (Section 3.1). The figure in Table 3 shows the location (red dot) of the measurements inside the area that was modelled to validate ENVI-met for the purpose of this study. The selection of the point was done aiming to consider a representative value in the street (i.e., sufficiently separated from the building facade and way from street intersections). T a , RH, W S , wind direction (W D ) were measured at 8.5 m a.g.l. with a WXT520 Weather Transmitter (Vaisala,), mounted on a meteorological mast. Sampling was done every 5 seconds and 1-min average values were stored during the measuring period. The sensors provided the following accuracy for the measurements: ±0.3 • C for T a , ±3% for RH, and ±3% for W S and W D . Additionally, 10-min average data of W S , RH, T a and total incoming radiation measured in Deusto were used to evaluate meteorological boundary conditions during the measurement campaigns. The site is located in an open area surrounded by water 3 kilometres far from the measurements and validation area. More information can be found in Acero and Herranz-Pascual [86]. The domain used to validate ENVI-met covered an area of 300 × 300 m. The building characteristics were obtained from the Regional Government. The wall and roof material properties of the buildings are presented in Table 3. The vertical and horizontal resolution of all the model domain was 2 meters as has been set in previous studies [87]. The number of grid cells in the x, y, z directions were 150, 150 and 31, respectively. The size of the vertical grid cells was constant up to 26 m and then increased with height with a factor of 12%. Surface roughness was set to 0.2 m in correspondence with the surrounding area where the boundary conditions were measured. No specific vegetation elements were defined in the model due to the small size and low number of these. Their influence on the local microclimate is expected to be negligible. The surface materials were classified as light concrete for the pedestrian areas and asphalt for the traffic lanes. For the purpose of validating ENVI-met a few days of measurements (19th June, 1st July, 2nd July and 4th July) were selected with similar meteorological conditions to the ones used to analyse urban design scenarios in the current study (Section 3.5). This assures that the model was validated for the study's purpose. Hourly evolution of Ta and RH, in Deusto ( Figure 1) were used to force the model.

Description
Burned Brick Tile The domain used to validate ENVI-met covered an area of 300 × 300 m. The building characteristics were obtained from the Regional Government. The wall and roof material properties of the buildings are presented in Table 3. The vertical and horizontal resolution of all the model domain was 2 meters as has been set in previous studies [87]. The number of grid cells in the x, y, z directions were 150, 150 and 31, respectively. The size of the vertical grid cells was constant up to 26 m and then increased with height with a factor of 12%. Surface roughness was set to 0.2 m in correspondence with the surrounding area where the boundary conditions were measured. No specific vegetation elements were defined in the model due to the small size and low number of these. Their influence on the local microclimate is expected to be negligible. The surface materials were classified as light concrete for the pedestrian areas and asphalt for the traffic lanes. For the purpose of validating ENVI-met a few days of measurements (19th June, 1st July, 2nd July and 4th July) were selected with similar meteorological conditions to the ones used to analyse urban design scenarios in the current study (Section 3.5). This assures that the model was validated for the study's purpose. Hourly evolution of T a and RH, in Deusto ( Figure 1) were used to force the model.
Each day W S and W D was considered constant, and the median value of the hourly data between 10:00 and 20:00 (UTC) was used as input to the model. Between these hours air flow characteristics registered in Deusto (i.e., 30 m a.g.l.) were quite similar and corresponded to a well-established sea breeze. W S was adjusted to 10 m a.g.l. using the power law for wind profile [88,89]: where W S is the wind speed (ms −1 ) at the height of h and α is an empirical exponent which depends on the surface roughness. In this case α = 1/7 because the measurements were done in an open terrain with obstacles not closer than 75 m. The estimation of solar radiation made by the model was adjusted by comparing with data measured at the Deusto station. The solar factor selected to adjust the shortwave solar radiation simulated by the model for each modelling period (i.e., each day) was at the maximum solar radiation hour (i.e., 12.00 UTC). Due to lack of information of soil moisture and temperature initial conditions, model default values were used. The meteorological input parameters are shown in Table 4. Simulations were launched at 4:00 local time (i.e., UTC+2), approximately 3 h before sunrise. The total modelling time was 44 h to allow the spin-up of the model. The last 24 h output (complete daily cycle) were considered for the analysis. A discrete receptor was defined inside the model to specify the location of measurements.

Modelled Domain for Validation
The domain used to validate ENVI-met covered an area of 300 × 300 m. The building characteristics were obtained from the Regional Government. The wall and roof material properties of the buildings are presented in Table 3. The vertical and horizontal resolution of all the model domain was 2 meters as has been set in previous studies [87]. The number of grid cells in the x, y, z directions were 150, 150 and 31, respectively. The size of the vertical grid cells was constant up to 26 m and then increased with height with a factor of 12%. Surface roughness was set to 0.2 m in correspondence with the surrounding area where the boundary conditions were measured. No specific vegetation elements were defined in the model due to the small size and low number of these. Their influence on the local microclimate is expected to be negligible. The surface materials were classified as light concrete for the pedestrian areas and asphalt for the traffic lanes. For the purpose of validating ENVI-met a few days of measurements (19th June, 1st July, 2nd July and 4th July) were selected with similar meteorological conditions to the ones used to analyse urban design scenarios in the current study (Section 3.5). This assures that the model was validated for the study's purpose. Hourly evolution of Ta and RH, in Deusto ( Figure 1) were used to force the model. Each day WS and WD was considered constant, and the median value of the hourly data between 10:00 and 20:00 (UTC) was used as input to the model. Between these hours air flow characteristics registered in Deusto (i.e., 30 m a.g.l.) were quite similar and corresponded to a well-established sea breeze. WS was adjusted to 10 m a.g.l. using the power law for wind profile [88,89]: where WS is the wind speed (ms −1 ) at the height of h and α is an empirical exponent which depends on the surface roughness. In this case α = 1/7 because the measurements were done in an open terrain with obstacles not closer than 75 m. The estimation of solar radiation made by the model was adjusted by comparing with data measured at the Deusto station. The solar factor selected to adjust the shortwave solar radiation simulated by the model for each modelling period (i.e., each day) was at

Model Evaluation
The model validation was done by comparing hourly average measured data with ENVI-met results of T a , vapour pressure (e), and W S . Comparison was made with ENVI-met output values at 9 m height (the closest output to the measurements' height). Figure 2 shows the results for 19th June 2011. Both the modelled and measured T a and e describe a similar pattern. However, the modelled T a shows lower values during daytime and higher values during the nighttime period. These results are related to the accumulation of heat inside the urban area and the boundary conditions that are forcing the model. During nighttime water surrounding the area from where boundary conditions are taken allows higher T a (and thus modelled T a ) than inside the street canyon (measured T a ). Daytime results for e show a good agreement. However, model underestimation during nighttime can be associated with boundary soil moisture conditions. During the other days included in the model's validation (1st July, 2nd July and 4th July), similar diurnal patterns of modelled and measured values are encountered. Figure 2 shows the results for 19th June 2011. Both the modelled and measured Ta and e describe a similar pattern. However, the modelled Ta shows lower values during daytime and higher values during the nighttime period. These results are related to the accumulation of heat inside the urban area and the boundary conditions that are forcing the model. During nighttime water surrounding the area from where boundary conditions are taken allows higher Ta (and thus modelled Ta) than inside the street canyon (measured Ta). Daytime results for e show a good agreement. However, model underestimation during nighttime can be associated with boundary soil moisture conditions. During the other days included in the model's validation (1st July, 2nd July and 4th July), similar diurnal patterns of modelled and measured values are encountered. To evaluate the relationship between modelled and measured variables correlation analyses were performed for the four days considered in the model validation.
The best relationship between modelled and measured data is for Ta and e with Pearson correlation coefficients of 0.951 (p < 0.001) and 0.827 (p < 0.001) respectively. The accuracy of ENVImet simulations (i.e., the degree to which modelled values approach measured values) was evaluated using the root mean squared error (RMSE), the mean absolute error (MAE) to summarize the difference between modelled and measured variables, and the dimensionless index of agreement (d), a descriptive measure to make cross-comparisons between models outputs [90]. Results are shown on  To evaluate the relationship between modelled and measured variables correlation analyses were performed for the four days considered in the model validation.
The best relationship between modelled and measured data is for T a and e with Pearson correlation coefficients of 0.951 (p < 0.001) and 0.827 (p < 0.001) respectively. The accuracy of ENVI-met simulations (i.e., the degree to which modelled values approach measured values) was evaluated using the root mean squared error (RMSE), the mean absolute error (MAE) to summarize the difference between modelled and measured variables, and the dimensionless index of agreement (d), a descriptive measure to make cross-comparisons between models outputs [90]. Results are shown on Table 5. RMSE (1.87 • C) and MAE (1.64 • C) for T a are close to the median value of reported values of RMSE (1.51 • C) and MAE (1.34 • C) in other ENVI-met studies [91,92]. A study in a close by area of Bilbao provided similar RMSE values, between 1.00 and 2.07 • C, and MAE values, between 0.83 and 1.82 • C [93].
Another study in Phoenix (Arizona, USA) also showed a range of values between 1.41 • C and 2.00 • C for RMSE and between 1.18 • C and 1.74 • C for MAE depending on the land use [94]. In the case of e and W S , RMSE values correspond to 9.5% and 165.9% and MAE values correspond to 8.0% and 149.4% of the mean measured value during the 4 days. The results for W S are influenced by the limitations of the model to be forced with changing airflow conditions. Results improve when comparing the pairs (measured & modelled) under the same airflow condition [i.e., sea breeze; between 10:00 and 20:00 (UTC)], but still RMSE (1.28 m/s) and MAE (1.14 m/s) are high representing 104.6% and 89.6% respectively of the mean measured value. Similarly, the d for W S is low showing the limitations of the model to represent correctly the measured values. However, results are in accordance with previous studies [82,93,95,96].
On the contrary, T a and e present high d values similar or even higher than the ones reported in other studies with ENVI-met [91,93], representing a suitable performance of the model. From the results of the model validation, it is concluded that ENVI-met simulations have a deviation with respect to measurements taken in the area used for model validation (compact midrise). Although RMSE, MAE and d values obtained for T a and e can be consider reasonable and are similar to other works with ENVI-met available in literature [82,93,95,96], the model lacks a validation for T mrt (crucial parameter for thermal comfort evaluation) that is a limitation of the study (see Section 4.6).

Model Settings for Scenarios Analysis
Each spatial domain, modelled in ENVI-met environment, was constituted by four blocks and three UCs. The typical geometric proportions of each UC of the selected urban areas (Table 1) and the vegetation elements (i.e., tree, grass) have been reproduced, while the size of the grid cells (2-3 m, Table 6) was set according to the width of the street so as to provide reliable results. ENVI-met projects each of vegetation element to the corresponding grid cells in domain, varying their climatic properties (e.g., humidity and roughness). Table 6. Input configuration data applied in the models to run the ENVI-met simulations.

a) Initial Meteorological Conditions
Wind speed measured at 10 m height (m/s) The nesting grids were set equal to 7. As recommended by Bruse [97], the spatial domain along the z-direction was set at least twice the height of the tallest building. The length of the buildings' blocks was set equal to six times of their height [98] to avoid any perturbation from the borders.
In order to consider the city's surrounding, the roughness of the urban environment and the material of the soil outside the area of the model was set accordingly. The buildings blocks were modelled as completely straight volumes to avoid any obstacle, barrier and obstruction (i.e., shelters, balconies, decorative items on the façades and roofs).
The materials were set to reproduce the real materials of walls, roofs and soil, which are typically used in the analyzed urban areas (Table 7). For the green scenarios, the central part of the street canyon was with grass, covering 30% of the total street's width. Grass had a height of 0.1 m. The foliage of the trees' crown was set equal to 2/3 of full-fill density ( Table 8). The distance between the aligned trees (D1) was set to maintain a constant ratio of foliage coverage (D1/Wt) in all scenarios ( Figure 3). The nesting grids were set equal to 7. As recommended by Bruse [97], the spatial domain along the z-direction was set at least twice the height of the tallest building. The length of the buildings' blocks was set equal to six times of their height [98] to avoid any perturbation from the borders.
In order to consider the city's surrounding, the roughness of the urban environment and the material of the soil outside the area of the model was set accordingly. The buildings blocks were modelled as completely straight volumes to avoid any obstacle, barrier and obstruction (i.e., shelters, balconies, decorative items on the façades and roofs).
The materials were set to reproduce the real materials of walls, roofs and soil, which are typically used in the analyzed urban areas (Table 7). For the green scenarios, the central part of the street canyon was with grass, covering 30% of the total street's width. Grass had a height of 0.1 m. The foliage of the trees' crown was set equal to 2/3 of full-fill density ( Table 8). The distance between the aligned trees (D1) was set to maintain a constant ratio of foliage coverage (D1/Wt) in all scenarios (Figure 3).
perature, calculated at z-Grid = 0.    Fourteen specific receptors were positioned in the central part of the urban canyon at equidistant from each other (D2). The receptors allowed assessing the thermal comfort using PET index [15]. Modelled results were saved every 30 min. Data was analyzed at 1 m above the ground surface except for the surface temperature, calculated at z-Grid = 0.
The spatial distribution of the receptors was set to study the local benefits given by the presence of the trees: some receptors (e.g., R02, R09) were located between the trees, others (e.g., R04, R12) under the trees' crown. All model settings for building blocks were set to generalize the geometry of the analyzed urban areas and facilitate the replicability of the methodology in different parts of the city where geometric characteristics (H/W and B/T) and presence of vegetation elements are similar.
The start time of the simulations was stated at 4:00 a.m. of the 6th of August local time (i.e., UTC+2), approximately 3 h before sunrise, while the total modelling time was set as 44 h. The first 20 h are necessary to spin up the model [99], and therefore, only the outputs related to the last 24 h (i.e., from 0:00 to 24:00 of the 7th of August) were considered (Table 6). Fourteen specific receptors were positioned in the central part of the urban canyon at equidistant from each other (D2). The receptors allowed assessing the thermal comfort using PET index [15]. Modelled results were saved every 30 min. Data was analyzed at 1 m above the ground surface except for the surface tem The spatial distribution of the receptors was set to study the local benefits given by the presence of the trees: some receptors (e.g., R02, R09) were located between the trees, others (e.g., R04, R12) under the trees' crown. All model settings for building blocks were set to generalize the geometry of the analyzed urban areas and facilitate the replicability of the methodology in different parts of the city where geometric characteristics (H/W and B/T) and presence of vegetation elements are similar.
The start time of the simulations was stated at 4:00 a.m. of the 6th of August local time (i.e., UTC+2), approximately 3 h before sunrise, while the total modelling time was set as 44 h. The first 20 h are necessary to spin up the model [99], and therefore, only the outputs related to the last 24 h (i.e., from 0:00 to 24:00 of the 7th of August) were considered (Table 6).

Scenarios
The comparative analysis of human thermal comfort was conducted in four scenarios. In the first scenario, the typical orientation of the selected urban areas was considered. This part, named as scenario S0, focuses on the outdoor human thermal comfort at 1 m high from the ground level in the selected areas characterized by the typical orientations: 24 • North-South (N-S) in Casco Viejo; 17 • N-S in Abando Indautxu; and 9 • N-S in Txurdinaga/Miribilla (Table 9). In the second set of simulations, four standardized orientations such as N-S, East-West (E-W), North/East-South/West (NE-SW) and South /East-North/West (SE-NW) for all the selected urban areas and different pavement materials of the street were analyzed in all the analyzed urban areas. This analysis, named as scenario S1, aims to study the effect of street pavement material when changing from asphalt (albedo = 0.12), to pedestrian boulevards with red brick stone (albedo = 0.30). In Bilbao these interventions, which started during the last decade to promote new urban public spaces, are more frequent in compact mid-rise and open-set high-rise urban areas. Differently, in compact low-rise urban area, such as the historic center of Casco Viejo, the traffic was highly limited, therefore the decorative red brick stones were originally used as pavement material. In all scenarios no vegetation elements were set (Table 9). Table 9. Settings of the scenarios to study the effect of orientation (S0) and street's pavement (S1).

Scenario S0 Orientation in S0
Scenario S1 In the last set of simulations, the mitigation strategies using vegetation elements were studied. Two mitigation scenarios through urban green interventions without giving any obligation or disposition to dwellings' owners and designers to plan any intervention at building level were analyzed. The scenarios are characterized by a loamy ground soil positioned in the central part and covering 30% of the street's width, while the rest of the street was covered with red brick stones. The loamy soil included the presence of grass, while a tree-lined was set in the central part of the street. These features were set in all urban areas as follows:  (Table 10).

The Effect of the Orientation
In the compact low-rise urban areas, the level of PET varies substantially with the orientations. Considering all 14 receptors (Figure 4a), the PET level in Casco Viejo generally results quite high. N-S and E-W orientations are mostly below the hot thermal perception range (35 • C < PET < 41 • C), while for the NE-SW orientation, the PET value reaches the very hot thermal perception level (PET > 41 • C).
The NW-SE orientation shows slightly lower PET level than the N-S, E-W orientation. Regarding the diurnal cycle (Figure 4b), PET peaks are distributed in different moments of the day according to the streets' orientation: around midday for N-S orientation; early in the morning and evening for E-W orientation. Out of the peak periods, the PET level remains in the neutral level in all orientations, except for NE-SW orientation.
NE-SW orientation has the highest peak values, around 20 • C higher than the rest of the orientations (i.e., two thermo-physiological stress classes). In the compact mid-rise urban areas, the peak levels of PET reach the range of moderate heat st ress (29 • C < PET < 35 • C) for the N-S and W-E orientations, which is one thermophysiological class less than the existing typical orientation of Abando/Indautxu ( Figure 5a); while for NW-SE orientation, the peak levels are in the range of slightly warm thermal perception (23 • C < PET < 29 • C) (Figure 5b).   The diurnal cycle shows that the intensity of PET remains within the range of neutral thermal stress for N-S, NW-SE and E-W orientations, while the NE-SW orientation has the highest PET levels. In the open-set high-rise urban areas there is a relevant reduction of the PET peaks (Figure 6a) due to the geometry of the urban canyon (i.e., low aspect ratios H/W). The distribution and the duration of PET peak also change considerably (Figure 6b) (see Section 4.2).   The diurnal cycle shows that the intensity of PET remains within the range of neutral thermal stress for N-S, NW-SE and E-W orientations, while the NE-SW orientation has the highest PET levels. In the open-set high-rise urban areas there is a relevant reduction of the PET peaks (Figure 6a) due to the geometry of the urban canyon (i.e., low aspect ratios H/W). The distribution and the duration of PET peak also change considerably (Figure 6b) (see Section 4.2). The diurnal cycle shows that the intensity of PET remains within the range of neutral thermal stress for N-S, NW-SE and E-W orientations, while the NE-SW orientation has the highest PET levels. In the open-set high-rise urban areas there is a relevant reduction of the PET peaks (Figure 6a) due to the geometry of the urban canyon (i.e., low aspect ratios H/W). The distribution and the duration of PET peak also change considerably (Figure 6b)

The Effect of Streets' Orientation on the Duration of PET's Peak
The analyses for standard orientations have confirmed that the best performing orientation in terms of thermal comfort standards is the NW-SE one where the PET level remains in the range of neutral thermal stress for the major part of the day, except during the peak period that it remains within the warm thermal sensation in all urban areas (Figures 4a, 5a and 6a). The worst performing orientation in terms of thermal comfort standards is the NE-SW, where the average PET level stays under the warm thermal stress limit (PET < 29 °C) in the open-set high-rise and compact mid-rise urban area, while only in the compact low-rise that limit is overcome. The duration of the peak period varies from 1 h for compact low-rise urban canyon, where the PET level reaches almost 53 °C, to more than 2 h and 30 min in compact mid-rise and open-set high-rise urban areas, in which the PET value is more than 45 °C. For the best orientation (NW-SE) the duration of the peak period coincides with the duration of the thermal discomfort period (PET > 23 °C) while for the NE-SW orientation, the thermal discomfort persists for over 10 h in all urban canyons (Table 11). These aspects are affected by the orientation and the aspect ratio of each urban canyon, which impact on the presence of direct solar radiation and the level of WS. In fact, even if the short-wave (Sw) direct irradiation remains over 745 W/m 2 during the peak in all urban areas, the average wind speed presents significant differences with 3.74 m/s in open-set high-rise, 2.74 m/s in compact mid-rise against 1.3 m/s in the compact lowrise urban areas (Figure 7 and Table 11). Table 11. Comparison of the PET peaks' values, Sw, Ws, the duration and the intensity of PET peaks and the duration of thermal discomfort between the orientation with the highest (NE-SW) and the lowest (NW-SE) PET values (data corresponds to the receptor R04 with pavement in red brick stones).

The Effect of Streets' Orientation on the Duration of PET's Peak
The analyses for standard orientations have confirmed that the best performing orientation in terms of thermal comfort standards is the NW-SE one where the PET level remains in the range of neutral thermal stress for the major part of the day, except during the peak period that it remains within the warm thermal sensation in all urban areas (Figures 4a, 5a and 6a). The worst performing orientation in terms of thermal comfort standards is the NE-SW, where the average PET level stays under the warm thermal stress limit (PET < 29 • C) in the open-set high-rise and compact mid-rise urban area, while only in the compact low-rise that limit is overcome. The duration of the peak period varies from 1 h for compact low-rise urban canyon, where the PET level reaches almost 53 • C, to more than 2 h and 30 min in compact mid-rise and open-set high-rise urban areas, in which the PET value is more than 45 • C. For the best orientation (NW-SE) the duration of the peak period coincides with the duration of the thermal discomfort period (PET > 23 • C) while for the NE-SW orientation, the thermal discomfort persists for over 10 h in all urban canyons (Table 11). These aspects are affected by the orientation and the aspect ratio of each urban canyon, which impact on the presence of direct solar radiation and the level of W S . In fact, even if the short-wave (Sw) direct irradiation remains over 745 W/m 2 during the peak in all urban areas, the average wind speed presents significant differences

Impact of Pavement Materials
The reconversion of the street from vehicular traffic with asphalt pavement to pedestrian promenade with pavement in red brick stones, especially changes in Ts and Tmrt are observed. Overall, these cause an increase in PET levels in all orientations (Figure 8). The highest value of Ts with asphalt

Impact of Pavement Materials
The reconversion of the street from vehicular traffic with asphalt pavement to pedestrian promenade with pavement in red brick stones, especially changes in T s and T mrt are observed. Overall, these cause an increase in PET levels in all orientations ( Figure 8). The highest value of T s with asphalt pavement reaches 44.1 • C in the compact mid-rise and 40.1 • C in open-set high-rise areas, while with red brick stones it is reduced to 37.8 • C and 38.5 • C in the compact mid-rise and open-set high-rise areas, respectively. T mrt values surpass 60 • C in all orientations with both asphalt and red brick stones, while the highest values, over 70 • C, persist for NE-SW. In all urban typologies, T mrt is generally higher with red brick stone. In the case of T a the differences between asphalt and red brick stones are lower than 5 % and 3 % for compact mid-rise and open-set high-rise urban areas, respectively. W s and RH show the same values for both pavement types. Regarding PET, red brick stones increase the thermal stress at pedestrian level in all street orientations, although the thermophysiological class rises from slightly heat stress to moderate heat perception only in the compact mid-rise urban canyon for NW-SE orientation and in the open-set high-rise for N-S orientation (Figure 8 and Table 12).

Impact of Pavement Materials
The reconversion of the street from vehicular traffic with asphalt pavement to pedestrian promenade with pavement in red brick stones, especially changes in Ts and Tmrt are observed. Overall, these cause an increase in PET levels in all orientations ( Figure 8). The highest value of Ts with asphalt pavement reaches 44.1 °C in the compact mid-rise and 40.1 °C in open-set high-rise areas, while with red brick stones it is reduced to 37.8 °C and 38.5 °C in the compact mid-rise and open-set high-rise areas, respectively. Tmrt values surpass 60 °C in all orientations with both asphalt and red brick stones, while the highest values, over 70 °C, persist for NE-SW. In all urban typologies, Tmrt is generally higher with red brick stone. In the case of Ta the differences between asphalt and red brick stones are lower than 5 % and 3 % for compact mid-rise and open-set high-rise urban areas, respectively. Ws and RH show the same values for both pavement types. Regarding PET, red brick stones increase the thermal stress at pedestrian level in all street orientations, although the thermophysiological class rises from slightly heat stress to moderate heat perception only in the compact mid-rise urban canyon for NW-SE orientation and in the open-set high-rise for N-S orientation (Figure 8 and Table 12).

Spatial Differences Inside the Street Canyon
Another relevant effect is the PET evolution along urban canyons. In both compact mid-rise and open-set high-rise areas, the PET is affected locally: differences between 1 and 1.5 • C values of PET are registered at the beginning (R01), in the middle (R04) and at the end (R07) of the urban canyon. For example, for the W-E orientation, R01 is under the Sun for a longer time than for R04 because it is localized in a point that is not affected by the buildings' shadows in the early morning and in the late afternoon (Figure 9d). In addition, the daily profile of W S is different along the urban canyon. The highest spatial differences along the street canyon occur for NW-SE orientation ( Figure 10) which corresponds with the prevalent wind direction in Bilbao (used as boundary meteorological condition in the simulations). In all orientations, the highest W S values occur at the beginning and at the end of the canyon characterized by an airflow "channelling effect".

Spatial Differences Inside the Street Canyon
Another relevant effect is the PET evolution along urban canyons. In both compact mid-rise and open-set high-rise areas, the PET is affected locally: differences between 1 and 1.5 °C values of PET are registered at the beginning (R01), in the middle (R04) and at the end (R07) of the urban canyon. For example, for the W-E orientation, R01 is under the Sun for a longer time than for R04 because it is localized in a point that is not affected by the buildings' shadows in the early morning and in the late afternoon (Figure 9d). In addition, the daily profile of WS is different along the urban canyon. The highest spatial differences along the street canyon occur for NW-SE orientation ( Figure 10) which corresponds with the prevalent wind direction in Bilbao (used as boundary meteorological condition in the simulations). In all orientations, the highest WS values occur at the beginning and at the end of the canyon characterized by an airflow "channelling effect".

Impact of Vegetation Elements
The PET peak values in the mitigation scenarios M01 (Ht/H = 0.25, grass and trees) and M02 (Ht/H = 0.25, Wt/W = 0.3, grass and trees) show that the highest thermal stress is always for the NE-SW orientation. For this orientation, the heat stress remains at the extreme level (PET > 41 °C) in the compact low-rise and compact mid-rise urban areas, while in the open-set high-rise the PET results just below the 41 °C (i.e., in the strong heat stress range) (Table 13). The lowest PET peak values are reached in the NW-SE orientation in all urban areas in correspondence with the predominant wind conditions in Bilbao. For this orientation, the PET value varies from warm level in compact low-rise urban areas (31.6 °C in M01 and 30.6 °C in M02) to slightly

Impact of Vegetation Elements
The PET peak values in the mitigation scenarios M01 (Ht/H = 0.25, grass and trees) and M02 (Ht/H = 0.25, Wt/W = 0.3, grass and trees) show that the highest thermal stress is always for the NE-SW orientation. For this orientation, the heat stress remains at the extreme level (PET > 41 • C) in the compact low-rise and compact mid-rise urban areas, while in the open-set high-rise the PET results just below the 41 • C (i.e., in the strong heat stress range) ( Table 13).
The As has been demonstrated in previous studies [100][101][102][103][104][105], the presence of trees has a limited and localized benefit, highly dependent on the reduction of incoming solar direct radiation given by the shadow of the trees' crown. Relevant local benefits are due to the presence of shadow created by the geometry (i.e., height, shape and width) and the foliage density of the trees' crown, the buildings' geometry and the urban canyons' orientation. In all urban canyons and orientations, the highest PET values are registered at the beginning and the end of the urban canyon due to the low shadow of buildings and trees' crown during the day, as happens in R01 and R07, which are not localized in proximity or under a tree (Figure 3).

Impact of Mitigation Effect in the Compact Low-Rise Urban Areas
In the compact low-rise urban area, the presence of grass and trees improves the PET peak up to one thermophysiological level ( Figure 11).
Slightly higher reductions are expected in the NE-SW and E-W orientations. Along the urban canyon, relevant local differences in PET peak values can be encountered. In scenario M01, N-S and E-W orientations show the highest differences (5 • C) between R01, R04, R07, while scenario M02 is the NE-SW orientation that reaches 4.6 • C between the three points. In both M01 and M02, the highest reduction is registered in the NE-SW orientation for receptor R04 (middle of the street canyon). In all street orientations and locations inside the street canyons (receptors), reductions in PET peak is always higher in M02 (Table 14). The daily evolution of PET in R04 in the NE-SW orientation shows that, despite in the scenario M01 the value of PET decreases about 11.3 • C with respect to the scenario S1 without vegetation, the thermal stress remains above the limit of a very hot level (PET up to 41.7 • C). Table 13. Peak values of PET, Mean Radiant Temperature (T mrt ), Surface Temperature (T S ), Air temperature (T a ), Relative Humidity (RH) and Wind Speed (W S ). Data extracted by all 14 receptors.  [100][101][102][103][104][105], the presence of trees has a limited and localized benefit, highly dependent on the reduction of incoming solar direct radiation given by the shadow of the trees' crown. Relevant local benefits are due to the presence of shadow created by the geometry (i.e., height, shape and width) and the foliage density of the trees' crown, the buildings' geometry and the urban canyons' orientation. In all urban canyons and orientations, the highest PET values are registered at the beginning and the end of the urban canyon due to the low shadow of buildings and trees' crown during the day, as happens in R01 and R07, which are not localized in proximity or under a tree (Figure 3).

Impact of Mitigation Effect in the Compact Low-Rise Urban Areas
In the compact low-rise urban area, the presence of grass and trees improves the PET peak up to one thermophysiological level ( Figure 11). Slightly higher reductions are expected in the NE-SW and E-W orientations. Along the urban canyon, relevant local differences in PET peak values can be encountered. In scenario M01, N-S and E-W orientations show the highest differences (5 °C) between R01, R04, R07, while scenario M02 is the NE-SW orientation that reaches 4.6 °C between the three points. In both M01 and M02, the highest reduction is registered in the NE-SW orientation for receptor R04 (middle of the street canyon). In all street orientations and locations inside the street canyons (receptors), reductions in PET peak is always higher in M02 (Table  14). The daily evolution of PET in R04 in the NE-SW orientation shows that, despite in the scenario M01 the value of PET decreases about 11.3 °C with respect to the scenario S1 without vegetation, the thermal stress remains above the limit of a very hot level (PET up to 41.7 °C).  The cooling benefit given by the presence of trees and grass is more perceivable in M02, where the PET level reaches 37.7 • C ( Figure 12). Less improvement is reached in the NW-SE orientation, in which the PET level decrease about 9.2 • C for scenario M02 from warm to comfortable thermal stress sensation class. The PET peak lasts for one hour from 15:30 to 16:30, while the duration of the thermal discomfort (PET > 23 • C) lasts for 11 h in S1 and M01, while it is reduced by 20 min in M02. The cooling benefit given by the presence of trees and grass is more perceivable in M02, where the PET level reaches 37.7 °C (Figure 12). Less improvement is reached in the NW-SE orientation, in which the PET level decrease about 9.2°C for scenario M02 from warm to comfortable thermal stress sensation class. The PET peak lasts for one hour from 15:30 to 16:30, while the duration of the thermal discomfort (PET > 23 °C) lasts for 11 h in S1 and M01, while it is reduced by 20 min in M02.

Impact of Mitigation Effect in the Compact Mid-Rise Urban Areas
The data for the mid-rise urban areas confirms that the combined presence of grass and trees generally improve the human thermal comfort along the entire urban canyon in all orientations about one thermophysiological level of PET ( Figure 13). In the NE-SW orientation, the presence of trees in M01 and M02 provide a local benefit along the canyon by reducing PET peak from very hot thermal heat stress. The PET overcomes the threshold of 41 °C in S1 to the range of hot level of thermal stress with PET equal to 39.0 °C in M01 and to 37.0 °C in R04 and to 36.5 °C in R07 in M02 (Table 15). In NW-SE orientation, the cooling benefit is higher

Impact of Mitigation Effect in the Compact Mid-Rise Urban Areas
The data for the mid-rise urban areas confirms that the combined presence of grass and trees generally improve the human thermal comfort along the entire urban canyon in all orientations about one thermophysiological level of PET ( Figure 13).
In the NE-SW orientation, the presence of trees in M01 and M02 provide a local benefit along the canyon by reducing PET peak from very hot thermal heat stress. The PET overcomes the threshold of 41 • C in S1 to the range of hot level of thermal stress with PET equal to 39.0 • C in M01 and to 37.0 • C in R04 and to 36.5 • C in R07 in M02 (Table 15). In NW-SE orientation, the cooling benefit is higher by reaching two thermophysiological classes: passing from warm thermal heat stress with PET level of 29.1 • C in S1, to slightly warm in M01 and PET equal to 23.9 • C until comfortable thermal heat stress in M02 with a PET value of 21.4 in R04 and of 20.8 in R07 (Table 15). The daily evolution of PET level in R04 shows that intensity of the peak values lasts for more than two hours: from 14:40 to 17:00, which results to be more than double in comparison to the compact low-rise urban areas ( Figure 14).
Regarding the duration of intensity of peak value of PET, the different geometry of the trees in M02 contribute to reduce it up to 20 min. This effect is even stronger in relation to the duration of thermal discomfort (PET > 23 • C): in the M01, the reduction is about 20 min; while in the M02, the benefit given by the presence of the tree allows a decrease of more than one-hour compare to S1. M02 registered in R04, for the worst orientation NE-SW (a) and the best orientation NW-SE (b).

Impact of Mitigation Effect in the Compact Mid-Rise Urban Areas
The data for the mid-rise urban areas confirms that the combined presence of grass and trees generally improve the human thermal comfort along the entire urban canyon in all orientations about one thermophysiological level of PET ( Figure 13). Figure 13. PET values registered in all 14 receptors in the compact mid-rise urban areas for scenario S1, M01 and M02 in all orientations.
In the NE-SW orientation, the presence of trees in M01 and M02 provide a local benefit along the canyon by reducing PET peak from very hot thermal heat stress. The PET overcomes the threshold of 41 °C in S1 to the range of hot level of thermal stress with PET equal to 39.0 °C in M01 and to 37.0 °C in R04 and to 36.5 °C in R07 in M02 (Table 15). In NW-SE orientation, the cooling benefit is higher   (Table 15). The daily evolution of PET level in R04 shows that intensity of the peak values lasts for more than two hours: from 14:40 to 17:00, which results to be more than double in comparison to the compact low-rise urban areas ( Figure 14).
Regarding the duration of intensity of peak value of PET, the different geometry of the trees in M02 contribute to reduce it up to 20 min. This effect is even stronger in relation to the duration of thermal discomfort (PET > 23 °C): in the M01, the reduction is about 20 min; while in the M02, the benefit given by the presence of the tree allows a decrease of more than one-hour compare to S1.

Impact of Mitigation Effect in Open-Set High-Rise Urban Areas
The analyses conducted in the open-set high-rise urban areas confirm the effects observed in the other urban areas. However, in these urban areas the PET values are consistently lower (Figure 15) than in the other urban areas. The PET level in different points along the urban canyon with the combined presence of grass and trees can benefit up to two thermophysiological levels of PET in all orientations (Table 16). Figure 16 shows the daily evolution of the PET level in the middle of the urban canyon (R04) in the scenario S1 and in the mitigation scenarios M01 and M02 for the NE-SW orientation, for which the highest value of thermal heat stress has been registered and for the NW-SE orientation, characterized by the lowest thermal stress. than in the other urban areas. The PET level in different points along the urban canyon with the combined presence of grass and trees can benefit up to two thermophysiological levels of PET in all orientations (Table 16). Figure 16 shows the daily evolution of the PET level in the middle of the urban canyon (R04) in the scenario S1 and in the mitigation scenarios M01 and M02 for the NE-SW orientation, for which the highest value of thermal heat stress has been registered and for the NW-SE orientation, characterized by the lowest thermal stress. The highest intensity of PET reaches 48.2 °C and lasts for 2 h and 40 min, from 14:30 to 17:10. This is the longest heat stress period among the all analysed urban canyons. In terms of cooling effect given by the presence of vegetation elements, the highest PET reduction equal to 12.4 °C was observed in the scenario M01, while in M02 it reached 15.4 °C. These reductions allowed lowering the heat thermal stress from very hot thermophysiological class in scenario S1 to the hot and warm thermal stress levels in scenarios M01 and M02, respectively. Regarding the duration of the thermal discomfort (PET > 23 °C), there are relevant differences among the different scenarios. In the scenario S1, the thermal discomfort period lasts for more than 10 h from 10:00 until 20:10, while the mitigation scenario M01 only reduces this period about 10 min (it ends at 20:00) and finally the mitigation M02 decreases this period about half an hour. This is due to the geometry of the trees' crowns. The highest intensity of PET reaches 48.2 • C and lasts for 2 h and 40 min, from 14:30 to 17:10. This is the longest heat stress period among the all analysed urban canyons. In terms of cooling effect given by the presence of vegetation elements, the highest PET reduction equal to 12.4 • C was observed in the scenario M01, while in M02 it reached 15.4 • C. These reductions allowed lowering the heat thermal stress from very hot thermophysiological class in scenario S1 to the hot and warm thermal stress levels in scenarios M01 and M02, respectively.
Regarding the duration of the thermal discomfort (PET > 23 • C), there are relevant differences among the different scenarios. In the scenario S1, the thermal discomfort period lasts for more than 10 h from 10:00 until 20:10, while the mitigation scenario M01 only reduces this period about 10 min (it ends at 20:00) and finally the mitigation M02 decreases this period about half an hour. This is due to the geometry of the trees' crowns.

Limitation of the Study
It is worth describing some of the limitations of the current study. Firstly, thermal stress evaluation requires the measure of four physical (air temperature, mean radiant temperature, humidity and air velocity) and two subjective (metabolic rate and thermal clothing insulation) parameters [106]. Despite the fact that the climate model in ENVI-met has been validated for Ta, RH, no validation was carried out for Tmrt due to the lack of measurements of this variable which could be considered a relevant limitation of the study. However, other studies in Bilbao [93] already evaluated this parameter showing a limitation of the model to represent correctly variations of incoming radiation (and thus, Tmrt) during daytime. This constraint is less relevant for the current study where a comparison among different scenarios has been conducted under the same boundary conditions that represent a clear sky day in Bilbao. Also, the validation of the model was done for one reference point. Although further points could be checked/validated, the chosen one can be considered representative of a compact midrise street due to its location close to its middle and away from intersections. Moreover, the data are related to only one day at each UC.
Secondly, building facades have been modelled as opaque walls while most of them combine windows and concrete facades. In a subsequent study, detailed data on the real stratigraphy of the opaque walls and specific data on the transparent parts (windows) could be provided in order to improve the accuracy of the results and also evaluate the influence on outdoor thermal comfort of the window-to-wall ratio (WWR) [107].
Thirdly, the findings (although based on hypothetical/simple urban shapes) are valid for districts that have similar characteristics as the urban areas analyzed in this study. However, the same methodology could be replicated in different parts of the city with other geometric proportions such as a-symmetrical canyons, which effects have been treated in some studies [108,109].
Fourthly, the results have been analyzed for specific receptors and not for the whole space inside the street canyon where different outcomes, at some extent, can be expected. Despite the approach does not include a detailed spatial analysis, the results show clearly the effects of the different mitigation strategies in Bilbao. These could be included in the coming Master Plan of the city to improve thermal comfort levels and reduce the impact of heat waves.
Finally, the study has been carried out with the same boundary conditions for all the study areas which somehow guarantees consistency of the results, i.e., see the effect of the mitigation strategies in different urban developments. However, recent studies have found that better outcomes of ENVImet can be expected when the model is forced with measurements inside the modelled domain [110,111]. Thus, probably a better comparison could have been possible if measurements in the three types of urban developments would have been available.

Conclusions and Further Developments
This work presents the analysis conducted on a summer day in Bilbao to study the effect of different orientations (i.e., N-S, NE-SW, NW-SE, W-E), aspect ratio (H/W), pavement materials (i.e., Figure 16. Evolution of PET level in open-set high-rise urban area for scenarios S1, M01 and M02 in R04, located in the middle of the UC for the worst NE-SW (a) and the best NW-SE (b) orientations.

Limitation of the Study
It is worth describing some of the limitations of the current study. Firstly, thermal stress evaluation requires the measure of four physical (air temperature, mean radiant temperature, humidity and air velocity) and two subjective (metabolic rate and thermal clothing insulation) parameters [106]. Despite the fact that the climate model in ENVI-met has been validated for T a , RH, no validation was carried out for T mrt due to the lack of measurements of this variable which could be considered a relevant limitation of the study. However, other studies in Bilbao [93] already evaluated this parameter showing a limitation of the model to represent correctly variations of incoming radiation (and thus, T mrt ) during daytime. This constraint is less relevant for the current study where a comparison among different scenarios has been conducted under the same boundary conditions that represent a clear sky day in Bilbao. Also, the validation of the model was done for one reference point. Although further points could be checked/validated, the chosen one can be considered representative of a compact midrise street due to its location close to its middle and away from intersections. Moreover, the data are related to only one day at each UC.
Secondly, building facades have been modelled as opaque walls while most of them combine windows and concrete facades. In a subsequent study, detailed data on the real stratigraphy of the opaque walls and specific data on the transparent parts (windows) could be provided in order to improve the accuracy of the results and also evaluate the influence on outdoor thermal comfort of the window-to-wall ratio (WWR) [107].
Thirdly, the findings (although based on hypothetical/simple urban shapes) are valid for districts that have similar characteristics as the urban areas analyzed in this study. However, the same methodology could be replicated in different parts of the city with other geometric proportions such as a-symmetrical canyons, which effects have been treated in some studies [108,109].
Fourthly, the results have been analyzed for specific receptors and not for the whole space inside the street canyon where different outcomes, at some extent, can be expected. Despite the approach does not include a detailed spatial analysis, the results show clearly the effects of the different mitigation strategies in Bilbao. These could be included in the coming Master Plan of the city to improve thermal comfort levels and reduce the impact of heat waves.
Finally, the study has been carried out with the same boundary conditions for all the study areas which somehow guarantees consistency of the results, i.e., see the effect of the mitigation strategies in different urban developments. However, recent studies have found that better outcomes of ENVI-met can be expected when the model is forced with measurements inside the modelled domain [110,111]. Thus, probably a better comparison could have been possible if measurements in the three types of urban developments would have been available.

Conclusions and Further Developments
This work presents the analysis conducted on a summer day in Bilbao to study the effect of different orientations (i.e., N-S, NE-SW, NW-SE, W-E), aspect ratio (H/W), pavement materials (i.e., asphalt and red brick stones) and vegetation elements (i.e., grass and trees) on human thermal stress at pedestrian level inside typical urban canyons of low-rise, mid-rise and open-set high-rise urban areas. The findings are presented as planning recommendations visualized in (Figure 17): • Urban parameters such as aspect ratio and orientation were found to have a significant influence on the human thermal comfort at the pedestrian level. In all urban areas, for a NE-SW orientation the solar radiation has the highest impact on thermal discomfort. In open-set high-rise urban areas the presence of the trees could produce a relevant reduction in thermal stress at pedestrian level. Furthermore, orientation and aspect ratio have a considerable influence on the intensity of the PET peak, its duration and on the period of thermal discomfort (PET > 23 • C).

•
The reduction of the intensity of the thermal stress at the pedestrian level and its spatial extent highly depend on the vegetative measures applied inside typical urban canyons. In the analyzed scenarios, the highest PET peak reduction due to tree-lined streets reaches 15.3 • C. Tree-lined streets composed of species with tall and broad crowns are more effective because of the large vegetation volume and leaf biomass inside the canyon. Similar to previous studies in Bilbao [100] the results show that the cooling effect provided by this arrangement of trees is in general locally restricted to the immediate vicinity of the trees.

•
The benefit in terms of human thermal comfort created by the presence of the vegetation elements, is more significant in the proximity of the tree-lined streets. In that regard, for the R04, localized under a tree, the benefit reaches a reduction of up to two PET thermal perception classes (depending on the street orientation) in all urban areas.

•
Regarding the thermal effect of pavement materials, it was demonstrated that replacing asphalt with decorative red brick stones reduces the surface temperature value, but increases the T mrt and PET at the pedestrian level. Thus, materials used in pedestrian areas in Bilbao are not beneficial to reducing heat thermal stress.

•
This study has demonstrated that street orientation, aspect ratio and the presence of vegetation consistently influence the wind speed at the pedestrian level and the cooling effect provided by street ventilation. Therefore, municipalities should adequately choose the types of trees to plant in relation to the urban canyon geometry.
The methodology used for this study can be applied in the early planning stages to support urban planners and decision makers to reduce the risk of human thermal stress inside urban canyons and districts. In fact, conducting quantitative and qualitative analyses allow evaluating and considering which urban interventions should be prioritized in new and/or in consolidated urban areas in order to guarantee thermal comfort at the pedestrian level. Future development of the study could include an economic evaluation to estimate the financial impact of each specific intervention and its relation with indoor energy consumption. Furthermore, analyses on the use of novel cooling materials for pavement such as retro-reflective could be undertaken [112][113][114]. Finally, another potential intervention in terms of cooling effect can be the application of vegetation elements on the building envelope such as green façades on existing and new buildings [115,116].
asphalt and red brick stones) and vegetation elements (i.e., grass and trees) on human thermal stress at pedestrian level inside typical urban canyons of low-rise, mid-rise and open-set high-rise urban areas. The findings are presented as planning recommendations visualized in (Figure 17):


Urban parameters such as aspect ratio and orientation were found to have a significant influence on the human thermal comfort at the pedestrian level. In all urban areas, for a NE-SW orientation the solar radiation has the highest impact on thermal discomfort. In open-set highrise urban areas the presence of the trees could produce a relevant reduction in thermal stress at pedestrian level. Furthermore, orientation and aspect ratio have a considerable influence on the intensity of the PET peak, its duration and on the period of thermal discomfort (PET > 23 °C).