Using Climate-Sensitive 3D City Modeling to Analyze Outdoor Thermal Comfort in Urban Areas

International Journal of Geo-Information

Article Using Climate-Sensitive 3D City Modeling to Analyze Outdoor Thermal Comfort in Urban Areas

1, 1, 2, SeyedehRabeeh HosseiniHaghighi †, Fatemeh Izadi †, Rushikesh Padsala † and Ursula Eicker 1,*

1 Department of Building, Civil, and Environmental Engineering, Concordia University, Montréal, QC H3G1M8, Canada; [email protected] (S.H.); [email protected] (F.I.) 2 Faculty of Geomatics, Computer science, and Mathematics, Hochschule für Technik Stuttgart, Schellingstr. 24, 70174 Stuttgart, Germany; [email protected] * Correspondence: [email protected] These authors contributed equally to this work. † Received: 29 September 2020; Accepted: 13 November 2020; Published: 19 November 2020

Abstract: With increasing urbanization, climate change poses an unprecedented threat, and climate- sensitive urban management is highly demanded. Mitigating climate change undoubtedly requires smarter urban design tools and techniques than ever before. With the continuous evolution of geospatial technologies and an added benefit of analyzing and virtually visualizing our world in three dimensions, the focus is now shifting from a traditional 2D to a more complicated 3D spatial design and assessment with increasing potential of supporting climate-responsive urban decisions. This paper focuses on using 3D city models to calculate the mean radiant temperature (Tmrt) as an outdoor thermal comfort indicator in terms of assessing the spatiotemporal distribution of heat stress on the district scale. The analysis is done to evaluate planning scenarios for a district transformation in Montreal/Canada. The research identifies a systematic workflow to assess and upgrade the outdoor thermal comfort using the contribution of ArcGIS CityEngine for 3D city modeling and the open-source model of solar longwave environmental irradiance geometry (SOLWEIG) as the climate assessment model. A statistically downscaled weather profile for the warmest year predicted before 2050 (2047) is used for climate data. The outcome shows the workflow capacity for the structured recognition of area under heat stress alongside supporting the efficient intervention, the tree placement as a passive strategy of heat mitigation. The adaptability of workflow with the various urban scale makes it an effective response to the technical challenges of urban designers for decision-making and action planning. However, the discovered technical issues in data conversion and wall surface albedo processing call for the climate assessment model improvement as future demand.

Keywords: 3D city models; spatial design; outdoor thermal comfort; mean radiant temperature

1. Introduction Canada’s Changing Climate Report has announced a doubled rate of warming for Canada compared to the rest of the world [1]. The Crowther Lab at ETH Zurich University has predicted that some Canadian cities will experience an average temperature increased by more than three degrees Celsius by 2050. It is estimated at 3.1 ◦C for Montreal [2]. Figure1 shows the predicted increase of stress days across Canadian cities for the projected years starting in 1961. The impact of heat stress is already being felt by increasing the number of hot days, more than 30 ◦C[3]. The rising frequency of heatwaves in many Canadian cities resulted in the death of 70 people in 2018 in Quebec alone [4]. Health Canada (2011) reported that when the daily average temperature goes higher than 20 ◦C in

ISPRS Int. J. Geo-Inf. 2020, 9, 688; doi:10.3390/ijgi9110688 www.mdpi.com/journal/ijgi ISPRSISPRS Int. Int. J. J. Geo-Inf. Geo-Inf. 20202020, ,99, ,x 688 22 of of 21 20

Health Canada (2011) reported that when the daily average temperature goes higher than 20 °C in sevenseven Canadian Canadian cities, cities, the the relative relative mortality mortality rate rate rises rises by by 2.3% 2.3% for for every every degree degree increase. increase. This This means means thethe intensity intensity of of 2–3 2–3 °C◦C would would lead lead to to a a 4–7% 4–7% increa increasese in in the the mortality mortality rate rate [5]. [5]. Environment Environment Canada Canada (Ontario(Ontario Region) introducesintroduces the the heatwave heatwave as aas period a period in which, in which, for three for successive three successive days, the days, maximum the maximumtemperatures temperatures stay or go stay over or 32 go◦C[ over6]. 32 However, °C [6]. However, the air temperature the air temperature is not the is onlynot the factor only for factor heat. forThe heat. daily The thermal daily experiencethermal experience is quite complex; is quite forcomp thelex; same for day the andsame air day temperature, and air temperature, there is a di ﬀthereerent isfeeling a different on a difeelingﬀerent on pathway. a different pathway.

FigureFigure 1. 1. CurrentCurrent and and projected projected number number of of days days exceedin exceedingg 30 30 °C◦C for for Canadian Canadian cities cities [3] [3 ]based based on on the the observedobserved temperature temperature data data between between 1961 1961 and and 1990. 1990. Human thermal comfort in the urban space is associated with four environmental parameters: Human thermal comfort in the urban space is associated with four environmental parameters: air temperature, relative humidity, air velocity, and thermal radiation, which is the most complicated air temperature, relative humidity, air velocity, and thermal radiation, which is the most complicated parameter. The thermal radiation within an urban open space is usually described by mean radiant parameter. The thermal radiation within an urban open space is usually described by mean radiant temperature (Tmrt), which is defined as a precise measurement of urban sites prone to heat [7]. temperature (Tmrt), which is defined as a precise measurement of urban sites prone to heat [7]. Thorsson et al. (2014) identified it as a better indicator of the harmful heat stress condition [8]. Thorsson et al. (2014) identified it as a better indicator of the harmful heat stress condition [8]. Tmrt Tmrt depends on the temperature difference and the radiant net exchange between the object and the depends on the temperature difference and the radiant net exchange between the object and the surrounded environment [9]. Thus, the potential of a surface to absorb or negate heat and the amount surrounded environment [9]. Thus, the potential of a surface to absorb or negate heat and the amount of object seen by surrounding radiation sources has a crucial influence on Tmrt variation. The local of object seen by surrounding radiation sources has a crucial influence on Tmrt variation. The local convection coefficient is another practical factor for increasing Tmrt in the dormant area. If the wind convection coefficient is another practical factor for increasing Tmrt in the dormant area. If the wind speed decreases, the portion of radiation in the global temperature increases; in turn, Tmrt grows speed decreases, the portion of radiation in the global temperature increases; in turn, Tmrt grows more [10]. more [10]. In the urban area, the physical characteristics of the built environment affect the distribution of Tmrt In the urban area, the physical characteristics of the built environment affect the distribution of by changing the exposure to sunlight and the material coverage of the surfaces. When the intra-urban Tmrt by changing the exposure to sunlight and the material coverage of the surfaces. When the intra- temperature shows a slight variation during the day, the Tmrt represents a considerable variation urban temperature shows a slight variation during the day, the Tmrt represents a considerable over short distances resulting in a noticeable difference in thermal perception [11,12]. The effect of variation over short distances resulting in a noticeable difference in thermal perception [11,12]. The urban form on thermal comfort focused on the Tmrt indicator has been the subject of much research for effect of urban form on thermal comfort focused on the Tmrt indicator has been the subject of much different climate zones using simulation method or real data observation analysis [11–15]. Two principle research for different climate zones using simulation method or real data observation analysis [11– geometry factors in evaluating the heat stress of urban spaces include the aspect ratio, the ratio of 15]. Two principle geometry factors in evaluating the heat stress of urban spaces include the aspect building height to the street width and, sky view factor, representing the surface exposure to the ratio, the ratio of building height to the street width and, sky view factor, representing the surface skydome, ranging from zero to one [12–14]. The finding highlighted the open structure design with exposure to the skydome, ranging from zero to one [12–14]. The finding highlighted the open low-rise buildings, and the lower aspect ratio is prone to increase the number of extreme heat stress structure design with low-rise buildings, and the lower aspect ratio is prone to increase the number hours. In contrast, the street canyon with a smaller sky view factor, shaded by surrounding buildings, of extreme heat stress hours. In contrast, the street canyon with a smaller sky view factor, shaded by provides more comfortable with lower Tmrt during summer days [12,15,16]. For canyons of the surrounding buildings, provides more comfortable with lower Tmrt during summer days [12,15,16]. same aspect ratio, their orientation has an influential role in managing the Tmrt [13]. However, it is For canyons of the same aspect ratio, their orientation has an influential role in managing the Tmrt recommended to consider the simultaneous impacts of all critical factors such as the aspect ratio, [13]. However, it is recommended to consider the simultaneous impacts of all critical factors such as canyon orientation, level of urban density and, latitude of canyon location [12,13,16,17]. the aspect ratio, canyon orientation, level of urban density and, latitude of canyon location The management of geometry parameters is costing financially and needs to be considered in the [12,13,16,17]. first place since they will not change over a long period of time. Although a dense structure maintains The management of geometry parameters is costing financially and needs to be considered in the extreme swings in Tmrt and daytime thermal stress during seasons [12,13], without respecting the the first place since they will not change over a long period of time. Although a dense structure flexible mitigators such as vegetations or adequate ventilation, it could increase nighttime temperature maintains the extreme swings in Tmrt and daytime thermal stress during seasons [12,13], without

ISPRS Int. J. Geo-Inf. 2020, 9, 688 3 of 20 due to the urban heat island effects during summer [12]. The contribution of urban heat island on the local scale and the increase of temperature at the regional level intensify the heat stress in the outdoor spaces [12,18]. Many studies showed the potential of Tmrt reduction through urban vegetation [11,14–16,19]. Planting trees is an effective measure to improve the urban microclimate by reducing summer day temperature within the tree’s boundaries and beyond the leeward side [19–21]. In addition to their ecological, aesthetic, health and, physiological benefits, the distribution and density of trees also allow flexible strategies to manage sun exposure in urban areas. Trees help reduce the transmission of solar radiation [15,16] and enhance the evapotranspiration and improve the shading pattern on surfaces underlying heat stress conditions [21]. A more extensive canopy means a more significant contribution in reducing temperature and decreasing Tmrt in both daytime and nighttime. The effect of clustered tree planting is more than the isolated placement of trees in Tmrt reduction at low density and low height structure [15,19]. However, dense construction must prioritize tree placement in areas not currently shaded by other sources like buildings, particularly in areas with heavy pedestrian traffic [14,15]. The effect of surrounded coverage is another concern of designers to manage Tmrt through the albedo of surface materials. Albedo is a physical property of the material to change the thermal behavior of the surrounded environment. Albedo refers to the fraction of solar radiation reflected by a projected surface [22]. Materials with low albedo, dark-colored, tend to absorb radiation and contribute to the urban heat island, while surfaces with a high albedo, white-colored, are prone to reflect it. Although using high albedo material is recommended for buildings to save energy during summer time [20], it is a less desirable solution on the surfaces near the ground where the reflected beams bounce back to urban spaces where human activities are present [23]. Alchapar, E.N. Correa (2015) demonstrated albedo’s effect on the temperature variation of surrounded surfaces, particularly in the high-density area, higher aspect ratio and, larger wall surfaces. When albedo increases by 10%, the air temperature increases by 0.5 K to 0.7 K, whereas it does not significantly change the low-density area [24]. It is essential to highlight that the impact of albedo is not constant during the day. Hui Li (2016) examined that the effect of albedo is at the highest level during the early morning and late afternoon. While during midday, it tends to a low and constant value, and cloudy days intensify it as well [25]. The summation of shortwave and longwave fluxes outgoing from the surrounded surfaces is comparatively small to incoming fluxes originating from the cardinal points [14,23]. Hence, the simulated albedo effect is expected to be small, given the different turbulence evened out in local air temperature. The complexity of outdoor thermal conditions in response to Tmrt, as a significant indicator of thermal comfort, demands the combination of 3D city modeling and climate assessment program to support climate sensitive spatial design. Traditional urban design methods were mostly based on 2D spatial design. However, with the continuous evolution of geospatial technologies and the added benefit of analyzing and virtually visualizing our world in three dimensions, it is possible to model and evaluate the reflection of the urban decisions before implementation and reduce the cost of trial errors. Thus, this research aims first to make use of 3D city modeling tools to design and create the study area and then uses these 3D city models to calculate the mean radiant temperature (Tmrt) in terms of assessing the spatiotemporal distribution of the heat stress for an upcoming transformed district in the city of Montreal. Alongside that, the study introduces a systematic workflow to evaluate and improve outdoor thermal comfort through the accurate placement of vegetation (trees), convenient in city scale.

2. Materials and Methods

2.1. Study Area and Scenario Development

Montreal is situated (45.50◦ N and 73.56◦ W) at the conﬂux of several climatic regions and is classiﬁed as humid continental [26]. The coldest month of the year is January, with a daily average temperature of 9.7 C, although due to the wind chill, the feeling temperature decreases more than − ◦ the actual temperature. The warm months start from mid-May to mid-September, with an average ISPRSISPRS Int.Int. J.J. Geo-Inf.Geo-Inf. 20202020,, 99,, 688x 44 ofof 2021

the actual temperature. The warm months start from mid-May to mid-September, with an average daily temperature of 20 C. The hottest month is July, with an average daily high of 26.3 C and lower daily temperature of 20 ◦°C. The hottest month is July, with an average daily high of 26.3 ◦°C and lower nighttime temperatures of 21.2 C. High humidity resulting from surrounded waterbodies turns the nighttime temperatures of 21.2 ◦°C. High humidity resulting from surrounded waterbodies turns the feeling temperature to higher than the summer’s actual temperature, and some nights temperature feeling temperature to higher than the summer’s actual temperature, and some nights temperature remains at an uncomfortable degree. The clear days last for 4.5 months, beginning in June and ending remains at an uncomfortable degree. The clear days last for 4.5 months, beginning in June and ending in October. Almost 64% of the year, Montreal’s sky is clear, mostly clear or partly cloudy. During in October. Almost 64% of the year, Montreal’s sky is clear, mostly clear or partly cloudy. During the the bright days, an average daily incident shortwave energy per square meter goes above 5.6 kWh bright days, an average daily incident shortwave energy per square meter goes above 5.6 kWh and and reaches 6.7 kWh on June 30. Variation of wind for cold days versus warm days is less than 1 m/s, reaches 6.7 kWh on June 30. Variation of wind for cold days versus warm days is less than 1 m/s, and and during calm and warm days of summer, the average speed is almost 4 m/s from the south direction. during calm and warm days of summer, the average speed is almost 4 m/s from the south direction. The study area, Dominion Bridge within the borough of Lachine East, is a low-rise industrial The study area, Dominion Bridge within the borough of Lachine East, is a low-rise industrial site sitelying lying on land on land with with a 3 am 3 mdifference difference between between north north and and south south of of the the site site that that covers covers around around 10 ha, locatedlocated inin thethe southsouth ofof MontrealMontreal island.island. TheThe areaarea isis surroundedsurrounded byby residentialresidential landland usesuses plannedplanned forfor midmid toto high-risehigh-rise developmentdevelopment fromfrom north,north, south,south, andand east,east, whilewhile thethe westwest sideside isis fullyfully developeddeveloped withwith low-riselow-rise residentialresidential buildings,buildings, seesee FigureFigure2 2.. The The developers developers are are interested interested in in turning turning Dominion Dominion Bridge Bridge intointo aa densedense andand high-risehigh-rise areaarea withwith residentialresidential buildings.buildings. TheThe municipalitymunicipality andand itsits mayormayor firmlyfirmly intendintend toto developdevelop anan eco-quartiereco-quartier withwith aa resilientresilient design, low levels of building energy consumption, usingusing onsiteonsite energyenergy generation,generation, andand improvingimproving humanhuman thermal comfort in outdoor urban spaces. RegardingRegarding thethe district’s district’s zoning zoning bylaw, bylaw, the the floor floor area area ratio ratio equals equals 3, and 3, and the lotthe coverage lot coverage ratio ratio equals equals 40%. The40%. site The provides site provides an above an 250,000 above sqm 250,000 floor sqm area distributedfloor area indistributed the different in buildingthe different configuration building layoutconfiguration and height. layout Since and theheight. site carriesSince the valuable site carries history valuable and heritage history and and tends heritage to accommodate and tends to moreaccommodate than 2700 more households, than 2700 the households, combination the of communitycombination and of urbancommunity spaces and were urban designed spaces to servewere didesignedfferent purposes to serve different and activities purposes in the and site. activities in the site.

FigureFigure 2.2. LocationLocation andand spatialspatial designdesign andand distributiondistribution ofof urbanurban spacesspaces inin thethe studystudy area:area: ((aa)) StudyStudy areaarea locationlocation inin LachineLachine neighborhoodneighborhood onon MontrealMontreal Island;Island; ((bb)) patrimonialpatrimonial buildingbuilding preservedpreserved inin thethe studystudy area;area; ((cc)) thethe alternativealternative mastermaster planplan andand focalfocal urbanurban spacesspaces withinwithin thethe studystudy area;area; ((dd)) birdbird viewview ofof buildingsbuildings bulk bulk distribution distribution and and pathways pathways connecting connecting urban urban spaces spaces and buildings and buildings generated generated through ArcGISthrough CityEngine. ArcGIS CityEngine.

BasedBased onon thethe aboveabove setssets ofof boundaryboundary conditionsconditions andand locallocal zoningzoning bylaws,bylaws, 3D city models,models, includingincluding buildings,buildings, streetstreet networks,networks, landscape,landscape, and trees, were modeled using ArcGIS CityEngine. CityEngineCityEngine isis partpart ofof ESRIESRI softwaresoftware suitessuites enabledenabled forfor interoperabilityinteroperability withwith GISGIS softwaresoftware suchsuch asas ArcGISArcGIS ProPro [ 27[27]] and and partially partially with with the open-sourcethe open-sou softwarerce software QGIS. QGIS. Whereas Whereas the traditional the traditional and manual and 3Dmanual modeling 3D modeling software suchsoftware as SketchUp, such as Rhinoceros3D, SketchUp, Rhinoceros3D, or Autodesk 3Dor software,Autodesk CityEngine 3D software, can generateCityEngine 3D citycan modelsgenerate leveraging 3D city models its underlying leveraging shape its grammar underlying technique shape togrammar program technique procedures to andprogram set parameters procedures to and generate set parame and iterateters large-scaleto generate 3D and city iterate models large-scale in considerably 3D city less models amount in considerably less amount of time. Such a parametric procedural modeling technique using shape

ISPRS Int. J. Geo-Inf. 2020, 9, 688 5 of 20 of time. Such a parametric procedural modeling technique using shape grammar offers a powerful 3D city modeling and visualization toolset, allowing quick visualization of complex city models, scenario development, and iterative urban design workflows [28]. Additional integration of a 3D vegetation library and inbuilt support to 3D objects for street furniture helps to make ArcGIS CityEngine a specialized 3D city modeling tool. For data conversion to other data formats, software such as feature manipulation engine (FME) provides a complete read and write support to CityEngine. The CityEngine derived 3D city models can be converted to various other 3D data formats such as COLLADA, gITF, FBX, OBJ, 3D multipatch shapefile, and file geodatabase (FGDB). Additionally, using FME, 3D city models from CityEngine can also be converted to the Open Geospatial Consortium (OGC) standardized open data model of CityGML in its different levels of details [29]. Although CityEngine is not open-source software, its application in the present work is highly motivated by its adaptability by city designers [30] as a specialized “3D urban modeling software” and creating all 3D city model elements. One such software is “Random3Dcity” [31]. Random3Dcity is an open-source procedural 3D city modeling software. However, due to its certain limitations, such as first, it is an experimental engine developed only as a prototype to be a first of its kind open-source procedural 3D city modeling tool. The second, its main priority is to procedurally generate 3D building geometries, giving the other elements of 3D city models such as vegetation, landcover, streets, and street furniture. That means the least amount of support and functionality. Thus, for the present work, the use of Random3Dcity was deemed not suitable. Another commercially available alternative is Rhinoceros3D [32], a computer-aided design (CAD) software widely used by city designers worldwide. Rhinoceros3D, when plugged in with Grasshopper, makes the 3D modeling parametric. The combination of Rhinoceros3D with Grasshopper and many different available plugins makes it a design environment that facilitates designing and analysis, all under one workspace. However, for analysis in the present work, QGIS plugin SOLWEIG needs to be used. Due to various technical shortcomings such as lack of direct support to coordinate systems, data interoperability issues with GIS software, no inbuilt and direct support to model other elements of 3D city models such as streets, landcover, and vegetation, Rhinoceros3D was also neglected. The proposed design provides an urban scenario with a combination of urban spaces (USp) and walking pathways in the various spatial scales and orientation, see Figure2. Following the designed function and active timing for each urban space, they are divided into four categories, the urban space with the intention of public gathering (USp_1), the local community spaces (USp_2–USp_4), and the public-heritage park as an accessible space for public use (USp_5). The final group of spaces as the medium spaces (USp_medium) located between principal urban spaces (USp_1 to USp_5) and connected them to buildings is divided into two groups according to the measurement of sky view factor, below 0.45 and above 0.45. From the energy-efficient design perspective for building construction, different setbacks have been considered from the street line to prevent the mutual shading of buildings and increase the efficient solar access in terms of heating demand reduction. This strategy has shaped wide canyons with a non-uniform and detached urban edge at 15 m above the street line, which affects the sky view factor of urban spaces unequally and results in the variation of shadow pattern on the ground. Regarding the study area’s complicated geometry in three dimensions, the 2D measurement of aspect ratio is not reliable and does not reflect the openness of urban spaces to the sky. While each point’s sky view factor in the outdoor spaces clearly demonstrates the quality of sun exposure and radiation accessibility for that spot. The average sky view factor of points fall into each urban space is measured through the mean radiant temperature simulation process and listed in Table1. ISPRS Int. J. Geo-Inf. 2020, 9, 688 6 of 20

Table 1. Characteristics of principle urban spaces.

Urban Parameters USp_1 USp_2 USp_3 USp_4 USp_5 ISPRSAvg. Int. J. Sky Geo-Inf. view 2020 factor, 9, x 0.70 0.56 0.56 0.71 0.97 6 of 21 Orientation N_S N_S N_S NE_SW Open Community ActivityActivity purposePublic Public squareCommunity Community purposes purposes_private _ Community purposes_Public ParkPublic purposes_public purpose square private public Park

2.2. Spatial Modeling and Simulation of Mean Radiant Temperature (Tmrt)(Tmrt) Following the aim of research for amelioratingameliorating dense building construction concerning the outdoor thermal condition and retrieving the effect effect of regional temperature increment for a time ahead, we leveraged a procedure using two modelin modelingg tools: SOLWEIG SOLWEIG as a climate assessment model for measuring the mean radiant temperature (Tmr (Tmrt)t) in complicated urban geometry, and ArcGIS CityEngine as a spatial modeling tool to generate and convert 3D city models to adopt the format required by SOLWEIG consistingconsisting ofof the digital elevationelevation model (DEM) and the digital surface model (DSM). DEM also is known as the bare earth model or also the digital digital terrain terrain model (DTM), the digital representation ofof elevation elevation data data depicting depicting real-world real-world terrain terrain profiles. profiles. A DSM A DSM is a digital is a digital surface surface model modelthat, in that, addition in addition to DEM, to also DEM, represents also represents elevation el dataevation of physical data of objects physical present objects on present top of the on terrain,top of thesuch terrain, as buildings, such as streets, buildings, and streets, landcover. and The landcove workflowr. The presented workflow in presented Figure3 intends in Figure to evaluate 3 intends the to evaluatespatial distribution the spatial of distribution the mean radiant of the temperaturemean radiant in temperature the outdoor areain the resulting outdoor from area building resulting layout. from Thebuilding output layout. helps The discover output the helps effect discover of building the geometryeffect of building in the distribution geometry in of shadowthe distribution and sunlit of shadowarea translated and sunlit to the area Tmrt translated map. Comparing to the Tmrt the resultmap. withComparing the physiological the result equivalentwith the physiological temperature (PET)equivalent index temperature allows clustering (PET) areasindex aboveallows the clusteri thresholdng areas and above prepare the the threshold hot-spots and map prepare representing the hot- spotsthe zones map with representing a higher demand the zones for with heat a mitigation higher dema practices.nd for This heat stage’s mitigation outcome practices. will be This used stage’s as the outcomeinput for thewill remodeling be used as process the input within for CityEnginethe remodeling and applying process systematicwithin CityEngine tree generation and applying process systematicas a response tree action generation to the eprocessfficient as heat a respon reduction.se action to the efficient heat reduction.

Figure 3.3.The The ﬂowchart flowchart of outdoorof outdoor thermal thermal assessment assessme andnt improvement and improvement via systematic via systematic tree generation tree generationas the passive as the heat passive mitigation heat strategy.mitigation strategy. 2.3. 3D City Modeling 2.3. 3D City Modeling To better produce and manage 3D city models, ESRI has developed the 3D City Information Model To better produce and manage 3D city models, ESRI has developed the 3D City Information (3DCIM). CityEngine, the 3D city modeling tool used in the present work, allows us to model the Model (3DCIM). CityEngine, the 3D city modeling tool used in the present work, allows us to model 3DCIM thematic structure shown in Figure4. The use of the 3DCIM thematic structure also satisﬁes the the 3DCIM thematic structure shown in Figure 4. The use of the 3DCIM thematic structure also data inputs required to run the SOLWEIG model. Although SOLWEIG does not take 3D city models as satisfies the data inputs required to run the SOLWEIG model. Although SOLWEIG does not take 3D a direct input for its execution, DEMs and DSMs can be easily derived from the 3D city models using city models as a direct input for its execution, DEMs and DSMs can be easily derived from the 3D ESRI’s ArcGIS Pro or FME. city models using ESRI’s ArcGIS Pro or FME.

ISPRS Int. J. Geo-Inf. 2020, 9, 688 7 of 20 ISPRS Int. J. Geo-Inf. 2020, 9, x 7 of 21

FigureFigure 4. The4. The 3D 3D City City Information Information Model Model (3DCIM) (3DCIM thematic) thematic structure structure by by ESRI ESRI [33 [33].].

TheThe present present work, work, the the district district of of Dominion Dominion Bridge Bridge in in Montreal, Montreal, is modeledis modeled based based on on the the 3DCIM 3DCIM thematicthematic structure structure using using CityEngine. CityEngine. First, First, the the terrain terrain data data (base (base map) map) is extractedis extracted and and converted converted to to a geotaggeda geotagged image image file file format format (GeoTi (GeoTiff)ff) from from the the available available CityGML CityGML terrain terrain (TINRelief) (TINRelief) data data of of MontrealMontreal [34 [34].]. Further, Further, taking taking the extractedthe extracted terrain terrain data asdata the as elevation the elevation source, source, 3D city 3D models city inmodels terms in ofterms buildings, of buildings, pathways pathways (built environment) (built environment) and land cover,and land trees cover, (natural trees environment) (natural environment) were modeled were accordingmodeled to according the mentioned to the zoningmentioned using zoning the CGA using rule the file CGA in ArcGIS rule file CityEngine. in ArcGIS ForCityEngine. modeling For buildings,modeling the buildings, CityEngine the inbuilt CityEngine CGA shape inbuilt grammar CGA shape rulefile grammar named “Genericrule file named Modern “Generic Buildings.cga” Modern wasBuildings.cga” used and tweaked was used to fit and the tweaked study area’s to fit zoning the stud bylawsy area’s and zoning boundary bylaws conditions, and boundary as explained conditions, in Sectionas explained 2.1, see Figurein Section5a. For 2.1, modeling see Figure landcover 5a. For and modeling its internal landcover pathways, and a publiclyits internal available pathways, CGA a shapepublicly grammar available rule file CGA for landscapeshape grammar design named rule “greenspacefile for landscape construction.cga” design named was tweaked “greenspace and used;construction.cga” see Figure5b. Then,was tweaked for modeling, and used; structural see Figure pathways, 5b. Then, and connectedfor modeling, sidewalks, structural an updated pathways, versionand connected of the publicly sidewalks, available an updated ESRI’s complete version of street the publicly CGA shape available grammar ESRI’s rule complete file published street CGA by D.shape Wasserman grammar (2020) rule [35 file] was published tweaked by to fitD.Wasserma the boundaryn (2020) conditions [35] was of the tweaked study area,to fit see the Figure boundary5c. Sinceconditions the first of purpose the study of the area, presented see Figure work 5c. is Since to use th 3De first city modelspurpose to of mitigate the presented the heat work stress is of to urban use 3D spacecity duringmodels heat to mitigate days of summer,the heat 3Dstress vegetation of urban objects space wereduring placed heat baseddays of on summer, the urban 3D heat vegetation stress hotspotobjects map were produced placed bybased the SOLWEIGon the urban model. heat Commonlystress hotspot found map deciduous produced tree byspecies the SOLWEIG in Montreal’s model. cityCommonly from the maplefound treedeciduous family “Norwaytree species maple in Montre (Acer platanoidesal’s city from)” is the used maple as it tree suits family our intention “Norway tomaple investigate (Acer theplatanoides shadow)” pattern is used in as the it suits urban our and intention open spaces to investigate [19]. A realistic the shadow representation pattern in of the Norwayurban mapleand open available spaces in [19]. the ESRI A realistic 3D vegetation representa librarytion [35 of] isNorway used to maple symbolize available trees inin CityEngine. the ESRI 3D Sincevegetation the proposed library design [35] wasis used focused to symbolize on being used trees as in an CityEngine. input to SOLWEIG, Since the buildings proposed were design modeled was infocused the level on of detailbeing (LoD)used as 2, an and input landscape, to SOLWEIG, trees, and bu pathwaysildings were were modeled modeled in in the LoD level 3, see of detail Figure (LoD)5d. 2, andAs an landscape, input to thetrees, SOLWEIG and pathways model, were all the modeled above 3Din LoD city models3, see Figure were rasterized5d. using ArcGIS Pro. For data consistency and to satisfy the SOLWEIG model’s prerequisite, all the raster data were produced with the same spatial resolution (1 m) and coordinate system NAD_1983_CSRS_MTM_8. This coordinate system is chosen because Montreal’s available open data are also modeled with the same coordinate system. The output from the conversion of CityEngine 3D city modeling consists of the DEM representing the terrain, DSM representing buildings with terrain, landcover representing the distribution of pathways, buildings, and the landscape for the study area, and canopy digital surface model (CDSM) representing trees.

ISPRS Int. J. Geo-Inf. 2020, 9, x 8 of 21

ISPRS Int. J. Geo-Inf. 2020, 9, 688 8 of 20 ISPRS Int. J. Geo-Inf. 2020, 9, x 8 of 21

Figure 5. The process of Dominion Bridge modeling through ArcGIS CityEngine: (a) The textured 3D building modeling; (b) the landcover modeling (landscape + internal pathways): (c) the pathway modeling: (d) the realistic (LoD3) 3D tree modeling using the 3D Norway maple tree model in ArcGIS CityEngine.

As an input to the SOLWEIG model, all the above 3D city models were rasterized using ArcGIS Pro. For data consistency and to satisfy the SOLWEIG model’s prerequisite, all the raster data were produced with the same spatial resolution (1 m) and coordinate system NAD_1983_CSRS_MTM_8. This coordinate system is chosen because Montreal’s available open data are also modeled with the same coordinate system. The output from the conversion of CityEngine 3D city modeling consists of FigureFigure 5. 5. TheThe process of Dominion Bridge Bridge mo modelingdeling through through ArcGIS ArcGIS CityEngine: CityEngine: (a) ( aThe) The textured textured 3D the DEM3Dbuilding building representing modeling; modeling; the(b ()b theterrain,) the landcover landcover DSM modelingrepresenting modeling (landscape (landscape buildings + internalinternal with terrain, pathways): landcover ((c)) thethe pathway pathwayrepresenting the modeling:distributionmodeling: ( d ))of the the pathways, realistic realistic (LoD3) (LoD3) buildings, 3D 3D tree tree andmodeling modeling the landscape using using the 3D the for Norway 3D the Norway study maple maplearea, tree modeland tree canopy modelin ArcGIS in digital ArcGIS CityEngine. surfaceCityEngine. model (CDSM) representing trees. However, the above-modeled buildings’ rasterization, particularly one 3D building geometry, However, the above-modeled buildings’ rasterization, particularly one 3D building geometry, as shownAs an ininput Figure to the 6, facedSOLWEIG a technical model, constrai all the abovnt. Thee 3D output city models DSM weregenerated rasterized using using 3D building ArcGIS as shown in Figure6, faced a technical constraint. The output DSM generated using 3D building Pro.geometry, For data and consistency its underlying and terrainto satisfy resulted the SOLWEI in inteGrpolated model’s elevation prerequisite, values al linstead the raster of preserving data were geometry, and its underlying terrain resulted in interpolated elevation values instead of preserving its producedits no data with cell thevalues same for spatial the open resolution space (1between m) and the coordinate horizontal system bridge NAD_1983_CSRS_MTM_8. building and the terrain no data cell values for the open space between the horizontal bridge building and the terrain surface. Thissurface. coordinate system is chosen because Montreal’s available open data are also modeled with the same coordinate system. The output from the conversion of CityEngine 3D city modeling consists of the DEM representing the terrain, DSM representing buildings with terrain, landcover representing the distribution of pathways, buildings, and the landscape for the study area, and canopy digital surface model (CDSM) representing trees. However, the above-modeled buildings’ rasterization, particularly one 3D building geometry, as shown in Figure 6, faced a technical constraint. The output DSM generated using 3D building geometry, and its underlying terrain resulted in interpolated elevation values instead of preserving its no data cell values for the open space between the horizontal bridge building and the terrain surface. FigureFigure 6. 6. OpenOpen spacespace as a part of 3D 3D building building geometry geometry in interpolatedterpolated with with elevation elevation cell cell values values in inits itsDSM. DSM.

OneOne possible possible explanationexplanation forfor suchsuch anan errorerror in thethe DSM can be that that since since the the building building DSM DSM is is a asummation summation of of the the building building and and terrain elevation cellcell values,values, thethe openopen spacespace belowbelowthe the horizontal horizontal bridge-buildingbridge-building isis alsoalso byby default default considered considered a a part part of of the the building. building. TheThe elevationelevation cellcell valuesvalues areare interpolatedinterpolated based based on on the the elevation elevation cell cell values values of theof the underlying underlying terrain terrain and and the bottomthe bottom surface surface of the of horizontal bridge-building. Such complications lead to an incorrect analysis for building geometries such as in Figure6 in simulation models such as SOLWEIG for which primary inputs are raster datasets. For aFigure proper 6. analysis Open space of theseas a part complex of 3D building building geometry geometries interpolated that are with easy elevation to ﬁnd cell in values the real in world,its the SOLWEIGDSM. model and other similar models need to be further developed to directly take 3D city models as input instead of converting them into a DSM, which leads to a loss of data/information. One possible explanation for such an error in the DSM can be that since the building DSM is a summation of the building and terrain elevation cell values, the open space below the horizontal bridge-building is also by default considered a part of the building. The elevation cell values are interpolated based on the elevation cell values of the underlying terrain and the bottom surface of

ISPRS Int. J. Geo-Inf. 2020, 9, 688 9 of 20

2.4. SOLWEIG Simulation Model The spatial variations of Tmrt are calculated using the SOLWEIG model. SOLWEIG is the open-source model of SOlar Long Wave Environmental Irradiance Geometry, a climate design program assembled on Urban Multi-scale Environmental Predictor (UMEP). UMEP is an extendable plugin for the QGIS, open-source desktop geographic information systems (GIS) that enables spatial data visualization and analysis capabilities [36]. SOLWEIG is suitable for analyzing the complex interaction of urban geometry and thermal environment at a district scale when the extended meteorological datasets are available [37]. It is a validated model to measure Tmrt resulted from the complex geometry of the built environment using 3D short and longwave radiation flux [11,14,15,37–40]. Comparing SOLWEIG with other parallel models (ENVI-met and Ray Man), Gála and Kántor (2020) demonstrated that the SOLWEIG model is highly reliable to predict the daytime Tmrt during the hottest period of the day, around noon and afternoon, when the local air temperature reaches the peak. However, for points located near the physical obstruction (e.g., close to buildings), using the domain-wide mean surface temperature leads to an overestimation of Tmrt where the site is in the shade and an underestimation where the site is sunlit. Moreover, when the sun elevation decreases, the rate of expectation for Tmrt errors increases as well [40]. SOLWEIG derives Tmrt from the Stefan–Boltzmann law using the mean radiant flux density, which is the result from the summation of all long and shortwave radiation fields in three dimensions on a standing person at the level of 1.1 m, where the angular factor is set to 0.22 for radiation fluxes from four lateral cardinal points and 0.06 for radiation fluxes from above and below sides [11,37]. It is noted to highlight the reflected fluxes are equally shared between the cardinal points [37,38], so, for the nonuniformed sky-obstructions (e.g., next to buildings), the model tends to errors in reflected fluxes [39]. The model assumes 0.7 and 0.97 for the absorption coefficients of short wave and longwave radiation, respectively. The default value of albedo is 0.20, and the emissivity of buildings and vegetation is considered 0.95. There is an empirical relationship between environment surfaces and air temperature where the sky is clear. The proportion of received sun stream for each surface depends on the anisotropic sky view factor and sun altitude [37]. Wind velocity is not included in the model. However, a future aim is to couple with a surface energy model and a convective boundary model to add the possibility of physiological equivalent temperature (PET) calculation [36]. In terms of satisfying the input datasets for SOLWEIG modeling, two datasets are required in various formats, meteorological data, and spatial data. The meteorological data sets for an alternative temporal period consist of 1 1 air temperature (◦C), specific humidity (kg.kg− ), barometric pressure (Kpa), wind speed (m.s− ), 2 2 1 relative humidity (%), hourly incoming shortwave radiation (W.m− ) and rainfall rate (kg.m− . s− ) as obligatory parameters which are explained in the next section. The spatial dataset is provided as the urban scenario and converted to DEM and DSM format at the district level. Using spatial data in the SOLWEIG model needs to pass the preprocessing level. In this case, building DSM and ground DEM are processed to generate wall aspect, wall height, and sky view factor within the UMEP application [36]. In the next step, SOLWEIG merges both datasets, meteorological and preprocessed spatial data output to produce Tmrt and the hourly pattern of shadow in the study area. Then, the SOLWEIG analyzer is the final stage for enabling results to be statistically calculated and plotted based on the SOLWEIG outputs such as the spatial distribution of Tmrt, sunlit, shadow, and more. Vectorizing the results and applying the PET index results in a map displaying the regions in the priority of heat mitigation called hot-spots map, a guide map appropriated for efficient intervention, and vegetation planning, see Figure7. ISPRS Int. J. Geo-Inf. 2020, 9, 688 10 of 20 ISPRS Int. J. Geo-Inf. 2020, 9, x 10 of 21

FigureFigure 7. 7.The The detailed detailed flowchart flowchart of data of data preparation preparation and processing and processing for hot spotfor hot map spot and vegetationmap and vegetation planning. planning. 2.5. Meteorological Data and Climate Scenario 2.5. Meteorological Data and Climate Scenario Increasing the population of urban areas leads to a massive need for new development and infrastructure,Increasing and the cities population are preparing of urban their areas new lead actions to plansa massive to face need the extensivefor new development demand for theand future.infrastructure, As discussed, and cities the proposedare preparing scenario their is new a response action plans to a future to face condition. the extensive Hence demand the research for the leveragedfuture. As a discussed, predicted climatethe proposed scenario scenario developed is a response by Hosseini to a etfuture al. 2020 condition. [41]. The Hence climate the scenario research forleveraged this research a predicted is derived climate from scenario the GFDL_ESM2 developed modelby Hosseini [42] concerning et al. 2020 climate[41]. The change climate resulting scenario fromfor this greenhouse research gasis derived emissions. from The the model GFDL_ESM2 data are model downscaled [42] concerning to Montré al-Pierreclimate change Elliott Trudeauresulting internationalfrom greenhouse airport gas with emissions. an hourly The resolution model data provided are downscaled for the next to 30 Montréal-Pierre years, year-by-year, Elliott taking Trudeau into accountinternational the pressure airport of with population an hourly increase, resolution slow pr incomeovided growth for the with next the 30 modestyears, year-by-year, rates of technology taking (RCPinto 8.5account W/m 2thescenario pressure [43 ])of from population 2020 onwards. increase The, slow model income used growth the statistically-based with the modest correction rates of totechnology eliminate (RCP the discovered 8.5 W/m2 scenario errors comparing [43]) from the 2020 previous onwards. forecast The model with used the observed the statistically-based data. Then, throughcorrection applying to eliminate a classification the discovered method errors used compar daily monitoreding the previous data from forecast 1953 with to 2014 the and observed regression data. techniques,Then, through a model applying was traineda classification to generate method hourly used weather daily monitored parameters data focusing from 1953 on temperature, to 2014 and radiation,regression wind techniques, speed, anda model atmospheric was trained pressure to ge [41nerate]. The hourly temperature weather variation parameters resulting focusing from on climatetemperature, change radiation, is significantly wind higherspeed, than and otheratmosp factors;heric seepressure Figure 8[41].b,d,f. Th Basede temperature on the analysis variation of projectedresulting years,from climate the intensity change and is frequencysignificantly of extremehigher than heat other have afactors; strong see tendency Figure to8b,d,f. increase Based if no on greenhousethe analysis gas of mitigationprojected years, planis the premeditated. intensity and frequency of extreme heat have a strong tendency to increaseReviewing if no thegreenhouse weather gas data mitigation analysis plan led byis premeditated. the RCP8.5 scenario represents the significant temperatureReviewing change the in weather Montreal. data In 2047,analysis this cityled experiencesby the RCP8.5 the warmestscenario periodrepresents and desperatethe significant need fortemperature cooling load change [41]. Thatin Montreal. was the keyIn 2047, point this to pick city up experiences 2047 weather the datawarmest as the period input ofand research desperate for theneed Tmrt for modelingcooling load and [41]. evaluation. That was the In this key year,point the to pi averageck up 2047 temperature weather data is expected as the input to increase of research by 2.76for the K,as Tmrt if 5.25 modeling K added and to evaluation. the average In of this wintertime year, the and average 1.54 Ktemperature added to the is averageexpected of to summer increase days,by 2.76 see K, Figure as if 85.25b. Comparing K added to importantthe average weather of wintertime profile factors and 1.54 for K 2020 added and to 2047, the average the period of ofsummer heat, thedays, average see Figure daily temperature8b. Comparing more important than 30 ◦weatherC, for 2047 profile is expected factors tofor be 2020 extended and 2047, from the one period month, of Julyheat, in the 2020, average to three daily months temperatur startinge in more May than and ending30 °C, for in August.2047 is expected The scenario to be predicts extended the from count one of hoursmonth, more July than in 2020, 20 ◦ toC willthree increase months 6.5%,starting from in Ma 1580y and h to ending 2138 h in in August. 2047, and Th Montreale scenario needs predicts to bethe preparedcount of hours for 153 more more than hours 20 °C of airwill temperature increase 6.5% higher, from than 1580 30 h to◦C. 2138 The h selected in 2047, temporaland Montreal period needs of Tmrtto beassessment prepared for for 153 the more project hours is two of air consecutive temperatur dayse higher from than August 30 °C. 10 The at 1:00 selected AM totemporal August period 11 at 24:00of Tmrt in 2047, assessment during for which the project the temperature is two consecutive variation days is between from August a minimum 10 at of1:00 20 AM◦C during to August night 11 timeat 24:00 and in a 2047, maximum during 35 which◦C at the 16:00 temperature of the first variat day.ion The is average between temperature a minimum ofof these20 °C during two days night is 26.19time ◦andC, almost a maximum 6 K more 35 than°C at the 16:00 same of timethe first in 2020, day. see The Figure average8a. Astemperature the input meteorologicalof these two days data is for26.19 the °C, alternative almost 6 scenario, K more than the model the same relies time on threein 202 components0, see Figure of 8a. solar As the radiation input meteorological flux during the daydata time,for the the alternative global horizontal scenario, as the the model primary relies factor, on directthree components normal and diofff solaruse horizontal radiation radiationflux during with the day time, the global horizontal as the primary factor, direct normal and diffuse horizontal radiation with the hourly average 348, 811 and 60 Wh.m-2, respectively. By comparing the radiation flux for the same days in 2020 and 2027, the direct and diffuse radiations do not show notable changes and

ISPRS Int. J. Geo-Inf. 2020, 9, 688 11 of 20

ISPRS Int. J. Geo-Inf. 2020, 9, x 11 of 21 2 the hourly average 348, 811 and 60 Wh.m− , respectively. By comparing the radiation flux for the same daysare entirely in 2020 overlaid and 2027, in the Figure direct 8e and. However, diffuse radiationsthe average do of not global show horizontal notable changes radiation and has are increased entirely overlaidsignificantly in Figure by 1088e. Wh.m However,−2 relatives the average to reference of global days horizontal in 2020, radiationsee in Figure has increased 8c. The average significantly wind 2 −1 byspeed 108 for Wh.m the− targetrelatives two to days reference is 2.73 days m.s in, 2020,and during see in Figurewhich8 thec. The cloudi averageness windof the speed sky is for at thethe 1 targetminimum. two days is 2.73 m.s− , and during which the cloudiness of the sky is at the minimum.

Figure 8. (a) The temperature difference between two alternative days of project (10 and 11 of August) Figure 8. (a) The temperature difference between two alternative days of project (10 and 11 of August) in 2047 and 2020; (b) pairwise comparison of the average monthly temperature for 2020 and 2047; in 2047 and 2020; (b) pairwise comparison of the average monthly temperature for 2020 and 2047; (c) (c) pairwise comparison of hourly global radiation flux for two alternative days of project (10 and 11 of pairwise comparison of hourly global radiation flux for two alternative days of project (10 and 11 of August 2020 and 2047); (d) pairwise comparison of monthly global horizontal radiation flux for 2020 August 2020 and 2047); (d) pairwise comparison of monthly global horizontal radiation flux for 2020 and 2047; (e) pairwise comparison of hourly direct and diffuse radiation flux for two alternative days and 2047; (e) pairwise comparison of hourly direct and diffuse radiation flux for two alternative days of project (10 and 11 of August 2020 and 2047); (f) pairwise comparison of monthly direct and diffuse of project (10 and 11 of August 2020 and 2047); (f) pairwise comparison of monthly direct and diffuse radiation flux for 2020 and 2047. radiation flux for 2020 and 2047. 3. Results and Discussion 3. Results and Discussion 3.1. Effect of Wall Surface Albedo Variation on the Mean Daytime Tmrt 3.1. Effect of Wall Surface Albedo Variation on the Mean Daytime Tmrt In the absence of vegetation, the distribution of Tmrt is dependent on the configuration layout of the buildingIn the absence and properties of vegetation, of surface the distribution material, precisely of Tmrt albedois dependent value. on Evaluating the configuration the effect layout of the variousof the building albedo valuesand properties on the wall of surface surfaces material, from 0.75 pr toecisely 0.15 showsalbedo a value. meaningful Evaluating decrease the ineffect the meanof the daytimevarious albedo Tmrt by values 0.5 K on per the twenty wall surfaces percent albedofrom 0.75 reduction to 0.15 asshows if the a maximummeaningful of decrease mean daytime in the mean Tmrt woulddaytime decrease Tmrt by by 0.5 almost K per 1.70twenty K. Thepercent open albedo spaces reduction showed the as if minimum the maximum reflection of mean in response daytime to Tmrt the wallwould surface decrease albedo by almost variation. 1.70 K. In The contrast, open narrowspaces showed spaces havethe minimum a maximum reflection reaction in response by raising to thethe averagewall surface of Tmrt albedo by 1 variation. K per 20 percentIn contrast, increase narrow in albedo spaces value, have seea maximum Table2. Using reaction albedo by inraising the low the valueaverage provides of Tmrt more by 1 comfortable K per 20 percent Tmrt duringincrease summer in albedo days. value, However, see Table it may 2. Using result albedo in uncomfortable in the low value provides more comfortable Tmrt during summer days. However, it may result in uncomfortable and cold stress in the narrow spaces between buildings, see Figure 9. When changing the wall surface albedo, the ground albedo was assumed the default set (0.20) on the model. The

ISPRS Int. J. Geo-Inf. 2020, 9, x 12 of 21

variation of ground surface albedo could be managed either by changing the default value or using the land cover layer produced in raster format. The probability of overestimating the Tmrt due to using the land cover layer discussed by Gála and Kántor (2020) [40] convinced us to use the default value in the rest of the modeling process. By changing the albedo values from 0.35 to 0.75, the results of SOLWEIG for the hourly Tmrt on the spots distributed in the site showed a considerable temperature variation at the peak hour, almost 5 K for the points near the buildings and less than 3 K for the central points in the surrounded urban spaces. Based on the literature review, the results are prone to overestimation for the points near the ISPRSbuildings Int. J. Geo-Inf. [24]. Moreover,2020, 9, 688 it was expected to see the maximum variation of Tmrt after or before12 of the 20 peak hour while the model starts dropping right after the peak to disappear by sundown [25]. Since the model is not comparable with the observed data, it suggested more investigation of the model’s and cold stress in the narrow spaces between buildings, see Figure9. When changing the wall surface accuracy concerning the effect of the material’s albedo for the next research. However, the default albedo, the ground albedo was assumed the default set (0.20) on the model. The variation of ground value of wall surface albedo, 0.20, has been tested and verified in many relevant studies using the surface albedo could be managed either by changing the default value or using the land cover layer SOLWEIG model [11,14,37–40]. Thus, we used it in the research since it is near the albedo of selected produced in raster format. The probability of overestimating the Tmrt due to using the land cover materials for buildings in the study area. layer discussed by Gála and Kántor (2020) [40] convinced us to use the default value in the rest of the modelingTable process. 2. The average daytime Tmrt (°C) in different urban spaces when the wall surface albedo changes. Table 2. The average daytime Tmrt (◦C) in diﬀerent urban spaces when the wall surface albedo changes. USp_ USp_ USp_ USp_ USp_ USp_ Avg. of Min–Max Albedo Value Avg. of Min–Max Albedo Value USp_11 USp_22 USp_33 USp_4 USp_54 USp_Medium5 Medium the Site of the Site the Site of the Site Avg. SVF 0.7 0.56 0.56 0.72 0.97 0.45 > Avg. SVF 0.7 0.56 0.56 0.72 0.97 0.45 > 0.75 35.81 32.38 32.96 37.47 40.00 31.72 37.43 27.24–45.38 0.75 35.81 32.38 32.96 37.47 40.00 31.72 37.43 27.24–45.38 0.550.55 35.0935.09 31.4831.48 32.05 32.05 36.80 36.80 39.88 39.8830.82 30.82 36.94 36.94 26.50–43.5326.50–43.53 0.350.35 34.3734.37 30.5830.58 31.15 31.15 36.13 36.13 39.76 39.7629.73 29.73 36.45 36.45 25.73–41.6625.73–41.66 0.150.15 33.6433.64 29.6829.68 30.24 30.24 35.43 35.43 39.64 39.6428.75 28.75 35.95 35.95 24.80–40.3024.80–40.30 Avg. of Tmrt Avg. of Tmrt 0.72 0.90 0.90 0.68 0.12 1.00 0.50 0.82–1.69 Diﬀerences 0.72 0.90 0.90 0.68 0.12 1.00 0.50 0.82–1.69 Differences

FigureFigure 9. 9.Spatial Spatial distribution distribution of of mean mean daytime daytime Tmrt Tmrt using usin digﬀ differenterent wall wall surface surface albedo albedo from from high tohigh low to value:low value: (a) albedo (a) =albedo0.75; (b=) albedo0.75; (b=) 0.55;albedo (c) albedo= 0.55;= (0.35;c) albedo (d) albedo = 0.35;= 0.15.3.2. (d) albedo The Spatiotemporal= 0.15.3.2. The DistributionSpatiotemporal of Tmrt Distribution and Identiﬁcation of Tmrt and of Hot-Spots. Identification of Hot-Spots.

ByThe changing measured the correlation, albedo values R2 (the from coefficient 0.35 to 0.75, of determination), the results of SOLWEIG between the for thealbedo hourly value Tmrt and onthe the average spots distributeddaytime Tmrt in theis 0.23. site It showed is not signi a considerableficant compared temperature to the sky variation view factor at the effect, peak which hour, almostprovides 5 K a for strong the points positive near correlation the buildings with and the lessvalue than of 3R2 K equal for the to central 0.92. It points states in the the high surrounded sky view urbanfactor spaces. goes with Based high on Tmrt the literature and determines review, the where results the are tendency prone to of overestimation the area to heat for stress the points increases, near theand buildings on the other [24]. Moreover,side, in which it was level expected of sky to exposure, see the maximum spaces have variation the primary of Tmrt aftercondition or before for heat the peakrelief hour Figure while 10. the The model spatial starts distribution dropping right of mean after thedaytime peak toTmrt disappear classified by sundownbased on [ 25physiological]. Since the modelequivalent is not temperature comparable (PET) with demonstrates the observed the data, thermal it suggested comfortability more investigationof urban spaces of resulted the model’s from accuracyvariation concerning of sky view the factor eﬀect and of thesolar material’s access, see albedo Figure for 10b. the nextPET index research. relies However, on dry temperature, the default valuerelative of wallhumidity, surface wind albedo, speed, 0.20, and has mean been testedradiant and temperature veriﬁed in to many represent relevant thestudies levels of using thermal the SOLWEIGperception model and physiological [11,14,37–40]. stress; Thus, see we in used Table it in 3 the[44]. research since it is near the albedo of selected materials for buildings in the study area.

The measured correlation, R2 (the coefficient of determination), between the albedo value and the average daytime Tmrt is 0.23. It is not significant compared to the sky view factor effect, which provides a strong positive correlation with the value of R2 equal to 0.92. It states the high sky view factor goes with high Tmrt and determines where the tendency of the area to heat stress increases, and on the other side, in which level of sky exposure, spaces have the primary condition for heat relief Figure 10. The spatial distribution of mean daytime Tmrt classified based on physiological equivalent temperature (PET) demonstrates the thermal comfortability of urban spaces resulted from variation of sky view factor and solar access, see Figure 10b. PET index relies on dry temperature, relative humidity, wind speed, ISPRS Int. J. Geo-Inf. 2020, 9, x 13 of 21

ISPRS Int. J. Geo-Inf. 2020, 9, x 13 of 21

Increasing 25% of the average sky view factor give rise almost 3.5 K to the mean daytime Tmrt of the urban spaces, although the direction of urban spaces has a considerable effect in the project’s case. The spatial pattern of mean daytime Tmrt in Figure 10b demonstrates the general increase in Tmrt at a distance from west-facing walls to the east-facing walls and north-facing walls to the south- facing walls in the N–S and the E–W canyons, respectively. The pattern and duration of shadow and sunlit area change the mean daytime Tmrt over a short distance in the surrounded urban spaces up to 25 K. When air temperature and global and direct radiation are at a high level, 12:00 to 19:00, the FigureFigure 10. 10.(a ()a The) The distribution distribution of of the the sky sky view view factor factor resulted resulted from from the the building building design design and and the the spread spread average of Tmrt exceeds the extreme heat stress, 41 °C, see Figure 8c. During this period, 75% of the ofof focal focal spots spots all all over over the the site; site; (b) ( theb) the distribution distribution of mean of mean daytime daytime Tmrt T classiﬁedmrt classified by PET by index;PET index; (c) the (c) site is sunlit, see Figure 11 and demands the dire need for heat mitigation. However, some areas are distributionthe distribution of mean of mean Tmrt classiﬁedTmrt classified by PET by index PET throughindex through the hottest the hottest period ofperiod the days of the (12:00–19:00). days (12:00– in buildings’19:00). shade and provide a moderate comfort level. The spatiotemporal behavior of Tmrt is differentTable 3.in Thermaleach urban sensitivity space, and depending grads of physiological on the characters stress, according and todirection Mayer and of Matzaraki the surrounding [44]. buildings.PET ( C) ◦ Above +41 +35 to +41 +29 to +35 +29 to +23 +23 to +18 +18 to +13 +13 to +8 +8 to +4 Below 4 Range Heat stress Extreme Strong Moderate Slight Slight Moderate Strong extreme Comfort category heat stress heat stress heat stress discomfort comfort cold stress cold stress cold stress

Increasing 25% of the average sky view factor give rise almost 3.5 K to the mean daytime Tmrt of the urban spaces, although the direction of urban spaces has a considerable eﬀect in the project’s case. The spatial pattern of mean daytime Tmrt in Figure 10b demonstrates the general increase in Tmrt at a distance from west-facing walls to the east-facing walls and north-facing walls to the south-facing walls in the N–S and the E–W canyons, respectively. The pattern and duration of shadow and sunlit area change the mean daytime Tmrt over a short distance in the surrounded urban spaces up to 25 K.

When air temperature and global and direct radiation are at a high level, 12:00 to 19:00, the average of Figure 11. (a) The percent of the time that the mean daytime Tmrt is above 41 °C during the hottest Tmrt exceeds the extreme heat stress, 41 ◦C, see Figure8c. During this period, 75% of the site is sunlit, Figure 10. (a) The distribution of the sky view factor resulted from the building design and the spread see Figureperiod 11 ofand days demands (12:00–19:00); the dire (b) needthe percent for heat of mitigation.the time which However, the area somecovered areas sunlit are during in buildings’ the of focal spots all over the site; (b) the distribution of mean daytime Tmrt classified by PET index; (c) shade andhottest provide period aof moderate days (12:00–19:00). comfort level. The spatiotemporal behavior of Tmrt is diﬀerent in each the distribution of mean Tmrt classified by PET index through the hottest period of the days (12:00– urban space, depending on the characters and direction of the surrounding buildings. 19:00).Table 3. Thermal sensitivity and grads of physiological stress, according to Mayer and Matzaraki [44].

PET Above +35 to +29 to +23 to +18 to +8 to Below (°C) +29 to +23 +13 to +8 +41 +41 +35 +18 +13 +4 4 Range Heat Extrem Stron Moderat Slight Slight Moderat Stron extrem stress Comfor e heat g heat e heat discomfor comfor e cold g cold e cold categor t stress stress stress t t stress stress stress y

FigureFigure 11.11.( (aa)) The The percent percent of of the the time time that that the the mean mean daytime daytime Tmrt Tmrt is is above above 41 41◦ °CCduring duringthe thehottest hottest periodperiod of of days days (12:00–19:00); (12:00–19:00); (b )( theb) the percent percent of the of timethe time which which the area the covered area covered sunlit duringsunlit theduring hottest the periodhottest of period days (12:00–19:00).of days (12:00–19:00).

Table 3. Thermal sensitivity and grads of physiological stress, according to Mayer and Matzaraki [44].

ISPRS Int. J. Geo-Inf. 2020, 9, 688 14 of 20 ISPRS Int. J. Geo-Inf. 2020, 9, x 14 of 21

In the publicpublic squaresquare ofof USp_1,USp_1, the the N_S N_S wide wide canyon, canyon, the the di differencesfferences between between East East and and West West of the of thesite site start start around around noon. noon. It takes It takes one hourone hour for the for west the andwest center and center of the siteof the to reachsite to the reach peak the of peak Tmrt of at Tmrt63 ◦C, at while 63 °C, it while takes 5it htakes for the 5 h eastfor the side east with side almost with 10almost K reduction 10 K reduction in the peak in the level. peak During level. theDuring five thehours, five the hours, discrepancy the discrepancy of temperature of temperature overtakes overtakes 25 K over 25 the K over public the square’s public square’s eastern andeastern western and westernsides. The sides. most The extended most extended duration durati of heaton of stress heat is stress for the is for western the western two-thirds two-thirds of the site,of the with site, 7 with h of 7extreme h of extreme heat stress. heat stress. In contrast, In cont therast, eastern the eastern part of part the of site the undergoes site undergoes less than less two than hours two hours above above 41 ◦C, 41see °C, Figure see Figure 12a. It 12a. is worthy It is worthy of indicating of indicating in the in N_S the squares N_S squares that the that eastern the eastern building’s building’s effect effect is more is morethan westernthan western in providing in providing efficient efficient shadow shadow during during the hottest the periodhottest ofperiod the day. of the The day. next The canyons, next canyons,USp_2 and USp_2 USp_3, and directedUSp_3, directed in N_S, in are N_S, influenced are influenced by a super by a super building building located located on the on eastern the eastern side sideof urban of urban spaces. spaces. The The buildings buildings have have a complex a complex form form above above 10 10 m m from from the the street streetline. line. At this level, level, the buildings have have a a 5 m setback and are converted to a detached form with almost 23 m distance between them, see Figure 22.. ForFor thethe non-uniformnon-uniform geometrygeometry modeling,modeling, thethe evaluationevaluation ofof aspectaspect ratioratio could notnot bebe beneficial, beneficial, and and measuring measuring the the sky sky view view factor factor in di ffinerent different sections sections of space of isspace more is reliable. more reliable.The design The eff ectdesign resulted effect in theresulted wavy skyin the view wavy factor, sky and view consequently, factor, and the curvyconsequently, distribution the of curvy mean distributiondaytime Tmrt of acrossmean daytime the canyon Tmrt changes across between the canyon 12 Kchanges to 28 K between for less than 12 K 2 to h. 28 There K for is less a significant than 2 h. Theredelay foris a spaces significant to reach delay their for peak spaces because to reach of the their surrounding peak because buildings’ of the elevation, surrounding see Figurebuildings’ 12b. elevation,It happens see after Figure 15:00 12b. for It the ha westernppens after side 15:00 and for at 18:00the western for the si easternde and sideat 18:00 of the for spaces.the eastern Overall, side ofthe the east spaces. side ofOverall, the urban the east spaces side is of shaded the urban 70% spaces of the is daytime shaded and70% representsof the daytime an area and withrepresents slight andiscomfort. area with Onslight the discomfort. other side, On the the west other side side, is located the west in moderateside is located heat in stress moderate for 85% heat of stress the time, for 85%see Figure of the 11time,. The see duration Figure 11. of heat The stressduration above of 41heat◦C stress is less above than 441 h, °C only is less for 30%than of 4 theh, only site nearfor 30% the ofeast-facing the site near walls. the east-facing walls.

Figure 12. The temporal and and spatial distribution distribution of of hourly hourly Tmrt Tmrt for for focal focal spots spots in the urban spaces demonstrated in Figure 10a;10a; (a) points (1,2,3); (b) points (4,5); (c) points (6,7,8);(6,7,8); ((d)) pointspoints (9,10).(9,10).

The courtyardcourtyard andand public public park, park, USp_4, USp_4, and and USp_5, USp_ have5, have almost almost the same the same behavior behavior in their in central their centralarea due area to thedue wide to the aspect wide ratio, aspect see ratio, Figure see 12 Figud. However,re 12d. However, the eﬀect the of effect surrounded of surrounded wall surfaces wall surfacesincreases increases the amount the ofamount longwave of longwave radiation radiation received forreceived the courtyard. for the courtyard. This results This in 2results K growth in 2 ofK growthTmrt at of the Tmrt peak at hour the peak compared hour compared to the public to the park public and park has alsoand reducedhas also reduced the time the in which time in space which is spaceilluminated is illuminated by 28%. Inby the28%. absence In the ofabsence surrounding of surrounding buildings, buildings, the heating the up heating of the public up of parkthe public starts parkearlier starts than earlier the other than regions, the other and regions, the duration and ofthe heat duration stress goesof heat above stress 7 h. goes The above last group 7 h. ofThe spaces last groupconnecting of spaces the principal connecting urban the principal spaces and urban buildings spaces with and the buildings average with sky the view average factor lowersky view than factor 0.45 loweris notably than a 0.45ﬀected is notably by surrounding affected by walls. surrounding In these spaces, walls. In the these building spaces, walls the decrease building their walls shortwave decrease their shortwave radiation access during the day time. It provides nearly 76% of the time under shade

ISPRS Int. J. Geo-Inf. 2020, 9, 688 15 of 20 ISPRS Int. J. Geo-Inf. 2020, 9, x 15 of 21 radiationduring the access heat duringstress period. the day For time. this It providesgroup of nearlyurban 76%spaces, of the the time average under daytime shade during Tmrt is the 36 heat °C, stressalmost period. 11 K lower For this than group the average of urban of spaces,the site, the see average Figure 12c. daytime Tmrt is 36 ◦C, almost 11 K lower than theTo averageachieve ofthermal the site, comfort see Figure at efficient 12c. and context-sensitive tree placement as the principal strategy,To achieve we overlaid thermal the comfort spatial atmap effi ofcient mean and Tmrt context-sensitive during heat stress tree placement with the designed as the principal human strategy,traffic pathway. we overlaid Applying the spatial the mapextreme of mean heat Tmrt thresh duringold above heat stress 41 °C with based the on designed the PET human index, tra itffi isc pathway.possible to Applying cluster the the regions extreme distributed heat threshold under above extreme 41 ◦C stress, based see on Figure the PET 13a. index, The it clustered is possible map to clusterdemonstrates the regions the preference distributed area under for extremeheat mitigation stress, see practice Figure in 13 thea. Thestudy clustered area’s dense map demonstratesstructure. The thefootpath, preference located area in forthe heatpublic mitigation park and practice far from in surrounded the study area’s walls, dense receives structure. a high solar The footpath,radiation locatedlevel with in themean public Tmrt park over and 57 far°C. fromThere surrounded is almost the walls, same receives heat stress a high for solarpathways radiation located level around with meanthe east-facing Tmrt over walls 57 ◦C. in There the urban is almost spaces the same USp_1, heat USp_2, stress forUSp_3, pathways and USp_4, located providing around the the east-facing average wallsTmrt inof the52, urban42, 41, spaces 57 during USp_1, the USp_2,hottest USp_3,period andof the USp_4, day, respectively. providing the In average the second Tmrt level, of 52, heat 42, 41,mitigation 57 during action the hottest goes for period the ofseries the day,of pathways respectively. distributed In the second in the level, rest heatof the mitigation site connecting action goesbuildings for the and series principle of pathways urban spaces, distributed medium in the spaces rest of with the sitethe connectingaverage sky buildings view factor and more principle than urban0.45. The spaces, variation medium of the spaces sky view with factor the average and shading sky view pattern factor results more in than noncontinuous 0.45. The variation regions ofabove the sky41 °C. view These factor areas and solicit shading for pattern deliberate results action in noncontinuous and pointwise regionsheat mitigation above 41 strategy.◦C. These The areas pathway solicit forarea deliberate located in action the andproximity pointwise of the heat west-facing mitigation strategy.walls and The spaces pathway with area the locatedlower sky in the view proximity factors of(0.45>) the west-facing is in the shade walls of and the spacessurrounded with thewalls lower for sky74% viewof the factors time and (0.45 displays>) is in thea nonsignificant shade of the surroundeddemand for wallsthermal for comfort 74% of the by timeproviding and displays the mean a nonsignificant Tmrt of 36 °C demandduring the for heat thermal stress comfort time, see by providingFigure 11. the mean Tmrt of 36 ◦C during the heat stress time, see Figure 11.

FigureFigure 13.13. TheThe overlaidoverlaid mapmap ofof thethe alternativealternative pathwaypathway andand thethe meanmean TmrtTmrt forfor thethe hottesthottest periodperiod ofof daysdays (12:00–19:00)(12:00–19:00) withwith thethe classiﬁcationclassification ofof PETPET inin threethree conditions;conditions; ((aa)) thethe baselinebaseline mapmap withoutwithout treetree plantingplanting scenario; scenario; ( b()b the) the ﬁrst first level level of treeof tree placement placement with with the prioritythe priority of pathways; of pathways; (c) the (c second) the second level oflevel tree of placement tree placement with the with priority the priority of some of urbansome urban spaces. spaces.

3.2.3.3. HeatHeat MitigationMitigation throughthrough TreeTree PlantationPlantation

ToTo systematicallysystematically placeplace thethe treestrees inin thethe firstfirst action,action, thethe clusteredclustered pathwaypathway aboveabove 4141◦ °CC inin thethe TIFTIF formatformat gotgot vectorizedvectorized throughthrough QGISQGIS beforebefore transferringtransferring toto CityEngine.CityEngine. InsideInside CityEngine,CityEngine, usingusing thethe completecomplete streetstreet rulerule CGACGA codecode fromfrom D.Wasserman,D.Wasserman, NorwayNorway maplemaple treestrees areare placedplaced onon thethe pathwaypathway sidewalks,sidewalks, whichwhich fallsfalls onon thethe centralcentral axisaxis ofof thethe clusteredclustered pathways.pathways. TheThe trees’trees’ placingplacing waswas consciousconscious withwith 1010 m m distancing distancing regarding regarding the averagethe average canopy canopy cover cover for the for target the Norwaytarget Norway maple (10 maple m-diameter) (10 m- withdiameter) 15 m height.with 15 Them height. tree generation The tree generation process’s proce outcomess’s leadsoutcome to264 leads trees to 264 with trees a su withfficient a sufficient canopy coveragecanopy coverage density of density 19% of of the 19% site (30%of the of site the (30% landscape) of the aslandscape) the mitigation as the action mitigation strategy. action The outcomestrategy. ofThe tree outcome planting of results tree planting in almost results 8.5 K improvementin almost 8.5 ofK meanimprovement Tmrt within of mean pathways Tmrt andwithin 7.5 Kpathways all over theand site 7.5 comparedK all over tothe without site compared tree condition to without during tree the condition stress time, during see the Figure stress 13 b.time, Figure see Figure14a shows 13b. theFigure improvement 14a shows ofthe average improvement Tmrt during of average the hottest Tmrt periodduring ofthe the hottest day, andperiod Figure of the 14 bday, represents and Figure the di14bfferences represents between the thedifferences conditions, between with andthe withoutconditions, tree with placement. and without The tree tree placement placement. in pathways The tree alsoplacement upgrades in pathways the average also of upgrades Tmrt in urbanthe average spaces of by Tmrt 5 K, in demonstrated urban spaces inby Table 5 K, 4demonstrated. This process in purposelyTable 4. This was process focused purposely on the pathways was focused as the on first th levele pathways of thermal as the comfort first level betterment of thermal and appliedcomfort betterment and applied the isolated-linear strategy of tree planting. However, the clustered strategy of tree placement is more recommended for a more substantial effect [19].

ISPRS Int. J. Geo-Inf. 2020, 9, x 16 of 21

In terms of improving thermal comfort in urban spaces, depending on their timing use and level of activity, it is required to repeat the process of overlaying of the mean Tmrt during the heat stress with the target urban space boundaries and applying the threshold of above 41 °C for the area under extreme stress. Subtracting the region out of stress provides a clustered map for the second level of

ISPRSheat Int. mitigation J. Geo-Inf. 2020 practice, 9, 688 through tree placement. Like the first level, the second level’s clustered16 of 20map output is converted to shapefile using QGIS and imported to CityEngine. Within CityEngine, for all the polygon patches of open spaces falling on the heat stress region, the CityEngine inbuilt “plant theloader.cga” isolated-linear rule strategy file is tweaked of tree planting.and applied However, to the open the clusteredspace polygon strategy patches of tree to placement match the isboundary more recommendedcondition of for1 tree a more per 120 substantial square meters. eﬀect [19 ].

FigureFigure 14. 14.Average Average Tmrt Tmrt for thefor hottestthe hottest period period of days of days (12:00–19:00) (12:00–19: following00) following the strategic the strategic addition addition of treesof intrees level in 1 level (a) and 1 ( levela) and 2 (levelb); the 2 change(b); the ( ∆change) in Tmrt (Δ) with in T themrt baseline with the map baseline (the condition map (the without condition treewithout placement) tree inplacement) level 1 (b )in and level level 1 (b 2) (andd). level 2 (d).

In termsThe second of improving level of thermaltree generation comfort results in urban in spaces,increasing depending the average on their density timing of usetrees and in levelthe site of activity,by 4% of it the is required site (6% toof repeatthe landscape), the process 58 oftrees. overlaying It leads ofto the meangeneral Tmrt improvement during the of heat mean stress Tmrt withby the 1.3 targetK in the urban site spacemeantime boundaries 1 K in andthe pathways applying thecompared threshold to ofthe above first mitigating 41 ◦C for the plan area during under the extremehottest stress. period Subtracting of the day, the see region Figure out 13c. of Moreover stress provides, the mean a clustered Tmrt changes map for 3.5 the K secondto better level for three of heattarget mitigation spaces practiceof USp_2, through USp_4, tree and placement. USp_5, which Like were the first subject level, to thenew second improvement level’s clustered because mapof their outputlocal isactivity converted purposes to shapefile and public using park QGIS functionalit and importedy, see toTable CityEngine. 4. Figure Within14c depicts CityEngine, the improvement for all theof polygon mean Tmrt patches in the of opensite, and spaces Figure falling 14d on repres the heatents stressthe difference region, the of CityEngineTmrt between inbuilt the condition “plant loader.cga”without tree rule and file isthe tweaked summation and appliedof the firs tot the and open second space level polygon of tree patches placement. to match the boundary condition of 1 tree per 120 square meters. The second level of tree generation results in increasing the average density of trees in the site by 4% of the site (6% of the landscape), 58 trees. It leads to the general improvement of mean Tmrt by 1.3 K in the site meantime 1 K in the pathways compared to the first mitigating plan during the hottest period of the day, see Figure 13c. Moreover, the mean Tmrt changes 3.5 K to better for three target spaces of USp_2, USp_4, and USp_5, which were subject to new improvement because of their local activity purposes and public park functionality, see Table4. Figure 14c depicts the improvement of mean Tmrt in the site, and Figure 14d represents the difference of Tmrt between the condition without tree and the summation of the first and second level of tree placement. ISPRS Int. J. Geo-Inf. 2020, 9, 688 17 of 20

Table 4. The average of mean daytime Tmrt in diﬀerent urban spaces where vegetation strategy is applied during the hottest period of day (12:00–19:00).

Avg. of Min–Max of Main Parameters USp_1 USp_2 USp_3 USp_4 USp_5 USp_Medium the Site the Site Avg. SVF 0.7 0.56 0.56 0.72 0.97 0.45 > Average of Tmrt L1 40.51 46.17 39.60 49.73 53.97 36.90 47.02 30.34–56.98 Average of Tmrt L2 36.18 42.76 39.23 39.34 42.83 36.42 39.80 30.34–56.98 Average of Tmrt L3 36.18 42.76 37.76 36.81 36.88 36.42 38.52 30.34–54.04 Percent of time above 45 67 42 78 95 24 67 0–100 41 ◦C (baseline) Percent of time above 19 50 42 29 52 24 35 0–100 41 ◦C in level 1 Percent of time above 19 50 33 17 17 24 29 0–100 41 ◦C in level 2

4. Conclusions This research identifies a systematic workflow to evaluate and upgrade the outdoor thermal comfort relying on the contribution of 3D city modeling and climate assessment application for various time scales. A case study district under development in Montreal was used to analyze 3D geometry’s impact on thermal comfort. Geometries were generated using the ArcGIS CityEngine. The generated spatial design was rasterized to serve the SOLWEIG program to model the spatiotemporal distribution of mean radiant temperature (Tmrt) as an outdoor thermal comfort indicator. Analyzing the outcome of SOLWEIG reveals the duration and pattern of the area under heat stress. The spatial variation of Tmrt demonstrates almost 25 K difference over the spaces investigated. One of the critical reasons was the variation of the average sky view factor resulting from surrounding buildings, with a non-uniform distribution of geometry in three dimensions. Moreover, the North-South dominant direction for some urban spaces resulted in a durable shadow for the area adjacent to the west-facing walls. Whereas the areas in the proximity of east-facing walls in the wide canyons suffered from being long-time exposed to solar irradiance during the hottest period of the day. Moreover, varying the albedo value on the wall surfaces showed an almost 0.5 K decrease per twenty percent reduction of albedo. Focusing on the albedo value, deep canyons with low sun accessibility showed more sensitivity to albedo effects, and their average mean daytime Tmrt changes more than wider canyons. Analyzing and classifying the results with the physiological equivalent temperature (PET) index indicated the potential regions under extreme heat stress. By overlaying the classified map and the human walkway’s principal structure, the priority of regions subject to heat mitigation action were clustered as a hot-spots map. The hot-spots map converted to polygon shapefile was used for automated tree generation in the CityEngine and converted to vegetation DSM for reassessment in the SOLWEIG program. The workflow was iterated two times. In the first round, the heat mitigation action’s target area was the improvement of human walkways. The response strategy results in proposing 264 number of trees with a 19% density of canopy coverage over the site (30% of the landscape). It led to improving the mean Tmrt during the hottest period of the day by 7.5 K in the entire study area, 8.5 K on the footpaths, and 5 K in the urban spaces. In the second try, the focus was on the urban space enhancement regarding their level of activity. In this case, the local community spaces and the park were in priority. The second intervention’s outcome added 58 more trees and increased tree canopy density by 4% (6% of the landscape). At this level, the mean Tmrt of the site was reduced by 1.3 K and determined urban spaces by 3.5 K during the hottest period of the day. The designed workflow is scalable and allows for the outdoor thermal evaluation in multiple urban levels. The simultaneous designing, visualizing, and assessment supports architects and urban designers for a collaborative practice toward climate-responsive decision-making to reduce the cost of human errors, as today’s design decision has implications for the next decades. However, using such a workflow, as demonstrated in the present work, reiterates the need for urban designers, geoinformatics, and sustainability experts to work hand-in-hand. That is because executing such a 3D city modeling ISPRS Int. J. Geo-Inf. 2020, 9, 688 18 of 20 driven workflows needs an acceptable level of experience and familiarity with the geographic information system (GIS), 3D data modeling methods, and basic computer programing ability in the case of using ArcGIS CityEngine. From the technical side in the SOLWEIG model, the hourly Tmrt on the spots near the buildings, in a nonuniformed surrounding, is prone to overestimating around the peak hour when the albedo value increases significantly from 0.15 to 0.75. Whereas in the central points with uniform surroundings and lower albedo value, the estimation of hourly Tmrt is close to the expected influence of albedo effect, namely lower than 2 K. The next issue is relevant to converting 3D building geometry with open space between the bottom surface of a bridge-type building and the ground surface (Figure6) provided complication and unreal data when converting to DSM. This problem resulted in missing correct data under the bridge building section and, consequently, no tree plantation suggestion could be made through the workflow. Such an issue can result in a wrong heat stress calculation from models such as SOLWEIG for which primary inputs are raster datasets. Since such complexity in the arrangement of building blocks is also available in the real-world, development of SOLWEIG type models which can directly accept 3D city models as inputs are suggested. It reduces the information loss, which happens while converting 3D city model geometries to raster datasets. Finally, this study clearly demonstrates the workflow’s capacity to provide an effective collaboration between 3D city modeling and climate assessment application to mitigate outdoor thermal stress through objective and systematic intervention (tree placement), which is currently a challenge for urban planners due to the lack of easy-to-use tools.

Author Contributions: Conceptualization, SeyedehRabeeh HosseiniHaghighi, and Fatemeh Izadi; methodology, SeyedehRabeeh HosseiniHaghighi, and Fatemeh Izadi and Rushikesh Padsala; software, Fatemeh Izadi and Rushikesh Padsala; formal analysis, SeyedehRabeeh HosseiniHaghighi; writing-original draft preparation, SeyedehRabeeh HosseiniHaghighi, and Rushikesh Padsala; writing—review and editing, Ursula Eicker and SeyedehRabeeh HosseiniHaghighi; visualization, Fatemeh Izadi, and Rushikesh Padsala. All authors have read and agreed to the published version of the manuscript. Funding: This research received no external funding. Conﬂicts of Interest: The authors declare no conﬂict of interest.

References

1. Canada’s Changing Climate Report. Available online: https://changingclimate.ca/CCCR2019/ (accessed on 23 September 2020). 2. Bastin, J.-F.; Clark, E.; Elliott, T.; Hart, S.; Hoogen, J.V.D.; Hordijk, I.; Ma, H.; Majumder, S.; Manoli, G.; Maschler, J.; et al. Correction: Understanding climate change from a global analysis of city analogues. PLoS ONE 2019, 14, e0224120. [CrossRef][PubMed] 3. Communicating the Health Risks of Extreme Heat Events—Canada.ca. Available online: https://www.canada.ca/en/health-canada/services/environmental-workplace-health/reports-publications/ climate-change-health/communicating-health-risks-extreme-heat-events-toolkit-public-health- emergency-management-oﬃcials-health-canada-2011.html (accessed on 24 September 2020). 4. Quebec Says Up to 70 People May Have Died in Connection with Recent Heat Wave | CTV News. (n.d.). published on 9 July 2018. Available online: https://www.ctvnews.ca/canada/quebec-says-up-to-70-people- may-have-died-in-connection-with-recent-heat-wave-1.4006431 (accessed on 17 November 2020). 5. Canada Health Act Annual Report 2011–2012—Canada.ca. Available online: https://www.canada.ca/en/ health-canada/services/health-care-system/reports-publications/canada-health-act-annual-reports/report- 2011-12.html (accessed on 24 September 2020). 6. Environment and Climate Change Canada—Weather and Meteorology—Weather and Meteorology—Glossary. Available online: https://www.ec.gc.ca/meteo-weather/default.asp?lang= En&n=B8CD636F-1&def=allShow (accessed on 24 September 2020). 7. Mayer, H.; Hoppe, P. Thermal comfort of man in diﬀerent urban environments. Theor. Appl. Climatol. 1987, 38, 43–49. [CrossRef] 8. Thorsson, S.; Rocklöv, J.; Konarska, J.; Lindberg, F.; Holmer, B.; Dousset, B.; Rayner, D. Mean radiant temperature—A predictor of heat related mortality. Urban Clim. 2014, 10, 332–345. [CrossRef] ISPRS Int. J. Geo-Inf. 2020, 9, 688 19 of 20

9. ASHRAE Handbook Online. Available online: https://www.ashrae.org/technical-resources/ashrae-handbook/ ashrae-handbook-online (accessed on 23 September 2020). 10. Nazarian, N.; Fan, J.; Sin, T.; Norford, L.; Kleissl, J. Predicting outdoor thermal comfort in urban environments: A 3D numerical model for standard effective temperature. Urban Clim. 2017, 20, 251–267. [CrossRef] 11. Lindberg, F.; Grimmond, C.S.B. The influence of vegetation and building morphology on shadow patterns and mean radiant temperatures in urban areas: Model development and evaluation. Theor. Appl. Climatol. 2011, 105, 311–323. [CrossRef] 12. Thorsson, S.; Lindberg, F.; Björklund, J.; Holmer, B.; Rayner, D. Potential changes in outdoor thermal comfort conditions in Gothenburg, Sweden due to climate change: The influence of urban geometry. Int. J. Climatol. 2011, 31, 324–335. [CrossRef] 13. Ali-Toudert, F.; Mayer, H. Numerical study on the effects of aspect ratio and orientation of an urban street canyon on outdoor thermal comfort in hot and dry climate. Build. Environ. 2006, 41, 94–108. [CrossRef] 14. Lindberg, F.; Holmer, B.; Thorsson, S.; Rayner, D. Characteristics of the mean radiant temperature in high latitude cities—Implications for sensitive climate planning applications. Int. J. Biometeorol. 2014, 58, 613–627. [CrossRef] 15. Thom, J.K.; Coutts, A.; Broadbent, A.M.; Tapper, N.J. The influence of increasing tree cover on mean radiant temperature across a mixed development suburb in Adelaide, Australia. Urban For. Urban Green. 2016, 20, 233–242. [CrossRef] 16. Lindberg, F.; Thorsson, S.; Rayner, D.; Lau, K. The impact of urban planning strategies on heat stress in a climate-change perspective. Sustain. Cities Soc. 2016, 25, 1–12. [CrossRef] 17. Johansson, E. Influence of urban geometry on outdoor thermal comfort in a hot dry climate: A study in Fez, Morocco. Build. Environ. 2006, 41, 1326–1338. [CrossRef] 18. Oke, T.R.; Johnson, G.T.; Steyn, D.G.; Watson, I.D. Simulation of surface urban heat islands under “ideal” conditions at night part 2: Diagnosis of causation. Boundary-Layer Meteorol. 1991, 56, 339–358. [CrossRef] 19. Wang, Y.; Akbari, H. The effects of street tree planting on Urban Heat Island mitigation in Montreal. Sustain. Cities Soc. 2016, 27, 122–128. [CrossRef] 20. Konopacki, S.; Akbari, H.; Gartland, L. Cooling Energy Savings Potential of Light-Colored Roofs for Residential and Commercial Buildings in 11 US Metropolitan Areas; Berkeley National Laboratory: Berkeley,CA, USA, 1997. [CrossRef] 21. Akbari, H.; Pomerantz, M.; Taha, H. Cool surfaces and shade trees to reduce energy use and improve air quality in urban areas. Sol. Energy 2001, 70, 295–310. [CrossRef] 22. Çengel, Y.A. Heat & Mass Transfer: Fundamentals and Applications; Tata McGraw-Hill: NewDelhi, India, 2011. 23. Erell, E.; Pearlmutter, D.; Boneh, D.; Bar Kutiel, P. Effect of high-albedo materials on pedestrian heat stress in urban street canyons. Urban Clim. 2014, 10, 367–386. [CrossRef] 24. Alchapar, N.L.; Correa, E. Comparison of the performance of different facade materials for reducing building cooling needs. In Eco-Efficient Materials for Mitigating Building Cooling Needs; Woodhead Publishing: Cambridge, UK, 2015; pp. 155–194. [CrossRef] 25. Li, H. Pavement Materials for Heat Island Mitigation: Design and Management Strategies; Elsevier: Amsterdam, The Netherlands, 2015. 26. Montreal climate: Average Temperature, Weather by Month, Montreal Weather Averages—Climate-Data.org. Available online: https://en.climate-data.org/north-america/canada/quebec/montreal-3704/ (accessed on 23 September 2020). 27. Padsala, R.; Coors, V. Conceptualizing, managing and developing: A web-based 3D city information model for urban energy demand simulation. In Proceedings of the UDMV15: Eurographics Workshop on Urban Data Modelling and Visualisation, Delft, The Netherlands, 23 November 2015; pp. 37–42. [CrossRef] 28. Watson, B.; Müller, P.; Veryovka, O.; Fuller, A.; Wonka, P.; Sexton, C. Procedural urban modeling in practice. IEEE Comput. Graph. Appl. 2008, 28, 18–26. [CrossRef] 29. Padsala, R.; Fink, T.; Peters-Anders, J.; Gebetsroither-Geringer, E.; Coors, V. From Urban Design to Energy Simulation—A Data Conversion Process Bridging the Gap Between Two Domains. In Proceedings of the 25th International Conference on Urban Planning, Regional Development and Information Society, Aachen, Germany, 15–18 September 2020; Volume 8, pp. 365–375. 30. Grêt-Regamey,A.; Celio, E.; Klein, T.M.; Hayek, U.W.Understanding ecosystem services trade-offs with interactive procedural modeling for sustainable urban planning. Landsc. Urban Plan. 2013, 109, 107–116. [CrossRef] ISPRS Int. J. Geo-Inf. 2020, 9, 688 20 of 20

31. Biljecki, F.; LeDoux, H.; Stoter, J. Generation of multi-LOD 3D city models in CityGML with the procedural modelling engine Random3Dcity. In Proceedings of the ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Delft, The Netherlands, 5 September 2016. [CrossRef] 32. Fink, T.; Koenig, R. Integrated parametric urban design in grasshopper/rhinoceros 3D demonstrated on a master plan in Vienna. In Proceedings of the 37 eCAADe and XXIII SIGraDi Joint Conference, “Architecture in the Age of the 4th Industrial Revolution”, Porto, Portugal, 11–13 September 2019; pp. 313–322. [CrossRef] 33. Reitz, T.; Schubiger-Banz, S. The ESRI 3D city information model. In IOP Conference Series: Earth and Environmental Science; IOP Publishing: Bristol, UK, 2014; Volume 18, p. 12172. [CrossRef] 34. Modèle Numérique de Terrain—Modèle Numérique de Terrain (MNT) 2015—Arrondissement Lachine—CityGML—PortailDonnéesOuvertes. Available online: http://donnees.ville.montreal.qc.ca/ dataset/modele-numerique-de-terrain-mnt/resource/54cb894b-551d-47ba-9cd5-0caf2a200ab1 (accessed on 23 September 2020). 35. GitHub—D-Wasserman/Complete_Street_Rule: The Complete Street Rule for ArcGIS CityEngine is a Scenario-Oriented Design Tool Intended to Enable Users to Quickly Create Procedural Generated Multimodal Streets. Available online: https://github.com/d-wasserman/Complete_Street_Rule (accessed on 23 September 2020). 36. UMEP Manual—UMEP Manual Documentation. Available online: https://umep-docs.readthedocs.io/en/ latest/index.html (accessed on 17 November 2020). 37. Lindberg, F.; Holmer, B.; Thorsson, S. SOLWEIG 1.0—Modelling spatial variations of 3D radiant fluxes and mean radiant temperature in complex urban settings. Int. J. Biometeorol. 2008, 52, 697–713. [CrossRef] 38. Lindberg, F.; Onomura, S.; Grimmond, C.S.B. Influence of ground surface characteristics on the mean radiant temperature in urban areas. Int. J. Biometeorol. 2016, 60, 1439–1452. [CrossRef] 39. Konarska, J.; Lindberg, F.; Larsson, A.; Thorsson, S.; Holmer, B. Transmissivity of solar radiation through crowns of single urban trees—Application for outdoor thermal comfort modelling. Theor. Appl. Climatol. 2014, 117, 363–376. [CrossRef] 40. Gál, C.V.; Kántor, N. Modeling mean radiant temperature in outdoor spaces, A comparative numerical simulation and validation study. Urban Clim. 2020, 32, 100571. [CrossRef] 41. Hosseini, M.; Bigtashi, A.; Lee, B. Generating future weather files under climate change scenarios to support building energy simulation—A machine learning approach. Energy Build. 2020, 184, 110543. [CrossRef] 42. Dunne, J.P.; John, J.G.; Shevliakova, E.; Stouffer, R.J.; Krasting, J.P.; Malyshev, S.L.; Milly, P.C.D.; Sentman, L.T.; Adcroft, A.J.; Cooke, W.; et al. GFDL’s ESM2 global coupled climate–Carbon Earth system models. Part II: Carbon System Formulation and Baseline Simulation Characteristics. J. Clim. 2013, 26, 2247–2267. 43. Riahi, K.; Rao, S.; Krey, V.; Cho, C.; Chirkov, V.; Fischer, G.; Kindermann, G.; Nakicenovic, N.; Rafaj, P. RCP 8.5—A scenario of comparatively high greenhouse gas emissions. Clim. Chang. 2011, 109, 33–57. [CrossRef] 44. Matzarakis, A.; Mayer, H.; Iziomon, M.G. Applications of a universal thermal index: Physiological equivalent temperature. Int. J. Biometeorol. 1999, 43, 76–84. [CrossRef]

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional aﬃliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).