Atmospheric Research 248 (2021) 105220 Contents lists available at ScienceDirect Atmospheric Research journal homepage: www.elsevier.com/locate/atmosres Highly resolved WRF-BEP/BEM simulations over Barcelona urban area with T LCZ I. Ribeiroa,b, , A. Martillic, M. Fallsa,d, A. Zonatoc,e, G. Villalbaa,f ⁎ a Sostenipra, Institute of Environment Sciences and Technology (ICTA), Autonomous University of Barcelona, Bellaterra, Spain b Now at the Swedish Meteorological and Hydrological Institute (SMHI), Norrköping, Sweden c CIEMAT (Research Centre for Energy, Environment and Technology), Av. Complutense, 40, 28040, Madrid, Spain d Now at the Department of Earth Sciences, Barcelona Supercomputing Center (BSC-CNS), Barcelona, Spain e University of Trento, Department of Civil, Environmental and Mechanical Engineering, Trento, Italy f Department of Chemical Engineering, Autonomous University of Barcelona, Bellaterra, Spain ARTICLE INFO ABSTRACT Keywords: This study evaluates the performance of urban schemes integrated in the Weather Research and Forecasting Local climate zones model (WRF) using Local Climate Zones (LCZ) as land use classification. We applied two multi-layer urban Multi-layer urban schemes schemes: 1) Building Effect Parameterization (BEP) and 2) Building Energy Model coupled with BEP WRF (BEP + BEM), over the Metropolitan Area of Barcelona (MAB) at 1km2 horizontal resolution for July 2016. Barcelona These two simulations were compared with observations and a standard WRF simulation (BULK approach). Corine Land Cover 2012 provides background information for the entire simulation domain, while the LCZ covers MAB classifying the land cover into 10 classes according to urban morphology and thermal properties. BULK and multi-layer urban scheme experiments present a similar general error trend: overestimation of relative humidity and planetary boundary layer height and underestimation of temperature. Although BEP has the best correlation with observations, this is the scheme with the highest value of bias and RMSE for tem- perature and relative humidity, in particular during the night/morning. On the other hand, BEP + BEM per- formed with the minimum RMSE associated for temperature and relative humidity in the entire domain. BEP + BEM has shown to be more sensitive than the other schemes over locations where the land use in the model grid differs to the real one, which is a common consequent limitation of horizontal model resolution. This study also suggests that depending on the synoptic condition the scheme accuracy on determining PBLH might change considerably. 1. Introduction social, environmental and health perspectives (de la Paz et al., 2016; Li et al., 2018; Rafael et al., 2017; Salamanca et al., 2012). In the last few decades, humanity has seen a substantial shift in the The urban atmosphere is strongly modified by anthropogenic ac- distribution of people. Not only does the global population continue to tivities (industry, traffic, domestic heating/cooling, etc.), as well as by grow at a significant pace, as this pace is faster in urban areas. surface materials that retain more energy than in rural areas. Currently, 54% of the world population is living in urban areas and it is Meteorological features at different scales also play an important role in expected to increase to 66% by 2050 (WHO, 2016). Furthermore, cities determining the general air circulation in the city (mesoscale phe- consume about 75% of global primary energy and emit between 50 and nomena) while local meteorological features take place near the surface 60% of the world's total greenhouse gases. This figure rises to ap- where the general air circulation might be influenced by urban mor- proximately 80% when the indirect emissions generated by urban in- phology (Middel et al., 2014; Santiago and Martilli, 2010; Xu et al., habitants are included (URL1, 2020). As a consequence, urban areas are 2017). Atmospheric models have been used to better understand the responsible for a disproportionately large contribution to a wide range dynamics of the urban atmosphere and to test the best urban planning of environmental problems, from global (climate change) to local (air strategies in order to improve thermal human comfort, wind flow at the pollution and thermal stress). This has resulted in the importance of street level and the influence on atmospheric pollutants dispersion, as urban planning to mitigate the impact of cities on its inhabitants from the UrbanSIS project (Gidhagen et al., 2020) and others have shown. ⁎ Corresponding author. E-mail address: isabel.ribeiro@smhi.se (I. Ribeiro). https://doi.org/10.1016/j.atmosres.2020.105220 Received 28 January 2020; Received in revised form 29 July 2020; Accepted 4 August 2020 Available online 28 August 2020 0169-8095/ © 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). I. Ribeiro, et al. Atmospheric Research 248 (2021) 105220 Usually, these strategies include studying the effect of building height, ▪ Single Layer Urban Canopy Model (SLUCM - Kusaka et al., 2001); street directions, implementation of trees in the streets, parks or green ▪ Building Effect Parameterization (BEP, Martilli et al., 2002): a multi- roof tops and strategies for the abatement of anthropogenic air pollu- layer layer scheme; tant concentrations (Amorim et al., 2013; Morini et al., 2017; Rafael ▪ BEP coupled to the Building Energy Model (BEP + BEM, Martilli et al., 2018; Vos et al., 2013). et al., 2002; Salamanca and Martilli, 2010): second generation of The urban schemes developed and integrated in the Weather BEP that considers energy consumption in buildings (heating/ Research and Forecasting model (WRF, Skamarock et al., 2008) are: cooling) for a more accurate effect on urban heat budget. Single Layer Urban Canopy Model (SLUCM - Kusaka et al., 2001), Building Effect Parameterization (BEP, Martilli et al., 2002) and BEP SLUCM and BEP compute the thermal and momentum interactions coupled to the Building Energy Model (BEP + BEM, Salamanca and between the atmosphere and the city in a single- and multi-layer model Martilli, 2010). They have been proved to be appropriate tools for grid, respectively. SLUCM has a simple 2D geometry, symmetrical street studying the urban heat island in different European cities such as: canyons with infinite length, canyon orientation and diurnal variation Madrid (de la Paz et al., 2016; Salamanca et al., 2012), Vienna of solar azimuth angle. Thus, the radiation treatment is three-dimen- (Hammerberg et al., 2018), Berlin (Kuik et al., 2016), Athens sional (3D) (Kusaka et al., 2001). On the other hand, BEP parameterizes (Giannaros et al., 2018), Rome (Morini et al., 2017) and Budapest a 3D urban morphology in a multi-layer model grid, and it is capable to (Göndöcs et al., 2017). estimate the heat fluxes from roofs, ground and walls, individually Additional input data related to the urban morphology and surface (Martilli et al., 2002). This way, BEP scheme computes the impact of thermal characteristics are needed so that the effects of buildings are buildings on the airflow and turbulence (term included in the con- simulated as well as the urban heat fluxes. Thus, Local Climate Zones servation equation for the Turbulent Kinetic Energy – TKE), as well as (LCZ, Stewart and Oke, 2012) database developed by Gilabert et al. the source/sinks of heat by solving the energy budget for each surface (2018) for the MAB region is used to characterize urban areas into 10 (de la Paz et al., 2016; Martilli et al., 2002). Unlike BEP that keeps the classes. Non-urban land covers are then characterized by Corine Land indoor temperature constant, BEP + BEM calculates the anthropogenic Cover 2012 (CLC12), used as background information. The most im- heat generated by air conditioning systems, as well as the heat ex- portant difference between CLC12 and LCZ is that the first considers changes between the interior of building and outer atmosphere only the land cover and human activity, while the last results from a (Brousse et al., 2016; Göndöcs et al., 2017; Salamanca and Martilli, combination of those with urban morphology and building thermal 2010). In this study, BULK, BEP and BEP + BEM options are tested to properties, making it an important base for urban climate studies. conclude about the performance of such schemes in Barcelona urban The goal of this study is to evaluate the performance of WRF and its region. multi-layer urban schemes (BEP and BEP + BEM) having the As indicated in Table 1, the vertical turbulent transport is para- Metropolitan Area of Barcelona as case study. One of the novelties of metrized using the Bougeault and Lacarrere (1989) turbulent scheme. this study resides in the use of the LCZ map to define urban land covers. This scheme has a prognostic equation for the TKE and includes a non- This approach, fostered by the WUDAPT project, is becoming more and local term during convective conditions. It has been chosen because it is more popular and it has already been used for Madrid (Brousse et al., among the best for the coastal area (Banks et al., 2016), and it is the 2016), Vienna (Hammerberg et al., 2018) and Bologna (Zonato, 2016) PBL scheme that has been tested most extensively coupled with BEP and in Europe. However, this study is the first that uses this approach for a BEP-BEM. coastal Mediterranean city, increasing the body of knowledge about the importance and applicability of this methodology for different city typologies. 2.2. Experiment set up A detailed description of the modelling set up, simulation domains and input data used to perform this study are described in section 2, This study focuses on the Metropolitan Area of Barcelona (MAB) and followed by the analysis and discussion of results (section 3) and lastly its surrounding, located at 41°26′24”N, 2°6′18″E. To simulate the target the conclusions (section 4). region, two two-way nested domains were defined. The parent domain (CAT), covering the whole Catalonia region (WE: 278 km, NS: 270 km), has a horizontal resolution of 3 km × 3 km. CAT domain is followed by 2. Data and methods a finer domain comprehending the Metropolitan Area of Barcelona (MAB) at 1 km × 1 km horizontal resolution (WE: 78 km; NS: 78 km) 2.1. Modelling system (Fig. 1). The domain geometry is based on previous studies conducted by the meteorological services of Catalonia over the MAB (MeteoCat, The Weather Research Forecasting (WRFV3.9.1) model (Skamarock personal communication in March 2018). Vertically, both domains are et al., 2008) is a numerical weather prediction and atmospheric simu- described by 45 layers (model top pressure: 100 hPa), being the urban lation system designed for both research and operational applications. canopy described by four layers about 10 m thick each. Finally, 0.25° Its dynamics solver integrates compressible and non-hydrostatic Euler global tropospheric analysis on a 6 h-basis (NCEP/NOAA, 2015) pro- equations with several physical and dynamic options designed to phe- vide initial and boundary meteorological conditions to the mother do- nomena at regional scale. However, motivated by the increasing in- main. terest on simulating the urban atmosphere and also due to its complex Table 1 dynamics, developments have been integrated into the main WRF Parameterizations used for each experiment. parameterizations (Salamanca et al., 2018). The bulk scheme (hereafter denoted as BULK), included in the Noah Surface Model is one example BULK BEP BEP + BEM of this (de la Paz et al., 2016; Liu et al., 2006), through the modification Urban scheme Included in the Noah BEP BEP + BEM of some parameters to better represent urban areas in what wind-speed Land Surface Model and heat storage capacity concerns. More recently, urban canopy Land surface model Noah Land Surface Model (Liu et al., 2006) schemes were developed and included into WRF, as physics options. PBL scheme Bougeault-Lacarrère PBL (BouLac), designed to use with These more advanced urban schemes were specifically designed to re- urban schemes (Bougeault and Lacarrere, 1989) Microphysics WRF Single Moment 6-class scheme present city morphology (e.g. building and street canyon geometry) and Long- and short-wave Rapid Radiative Transfer Model for General circulation surface characteristics (e.g. albedo, heat capacity, emissivity, urban/ radiation models (RRTMG) scheme vegetation fraction). The currently available urban canopy schemes are: 2 I. Ribeiro, et al. Atmospheric Research 248 (2021) 105220 satisfactory definition of LCZ comparing with the real morphology of the city. Once the LCZ classification is produced, it is incorporated within the WRF model choosing the most present value for each cell in the in- nermost domain. In order to implement the modified urban land mask in WRF, original 33 land use classes have been extended to 41 to allow the code to read the LCZ classification. MAB is the second largest Spanish metropolitan area and it is the sixth in Europe. Within the boundaries of Barcelona city, the most prominent LCZ class is Compact midrise, covering 31.7% of the city and hosting 75.8% of its dwellers (Table 2). Thus, this is the most densely populated area in the city (18,678 people·km2). The Barcelona city centre, mainly classified as Compact midrise (LCZ 2), is known by its two distinct urban geometries: 1) L'Eixample district characterized by regular blocks crossed by wide and straight avenues (75% is LCZ 2). In spite of its regular geometry, 20% of L'Eixample is classified as LCZ 1. Other land use classes represent ~5% of the area of this district and 2) three neighborhoods of the old city (Gràcia, Raval and Gótico) have with narrow streets and an unregularly pattern. Here, ~4% is LCZ 1, ~50% is LCZ 2, ~10% is LCZ 3 and ~ 20% is non-urban. The second largest LCZ is Dense trees (14.6% of Barcelona area), mainly occupied by the Collserola Park (~512 m above sea level) Fig. 1. Simulation domains: Catalonia domain (CAT) and the Metropolitan Area which is the most important green area of the city. The urbanized of Barcelona (MAB) with 3 km × 3 km and 1 km × 1 km horizontal resolution, suburbs, where 17.2% of the MAB population lives, are mainly classi- respectively. fied as Compact midrise. Each urban LCZ is characterized by typical geometrical parameters Three experiments are performed to figure out what is the best and thermal properties of building materials. The parameters used in urban scheme to simulate meteorological fields in MAB during typical this study are shown in the supporting material. Geometrical para- summer conditions. For this reason the whole month of July 2016 was meters are set to typical values for each LCZ, as suggested in Stewart simulated. The spin up period comprehends five days (1–5 July 2016). and Oke (2012) to evaluate the capability of LCZ classification to be Table 1 presents the main model parameterizations used within this adopted for most of the cities in the world. On the other hand, building work for both domains. thermal properties are the same for LCZ 2,3,5 and 9, whose walls are supposed to be built with solid brick, while roofs with tiles of the same 2.2.1. Case study and land use materials. LCZ 8, that describes industrial areas, is supposed to be built In both domains, the background land use is represented by Corine with concrete, and its thermal properties are assumed both for walls Land Cover 2012 (CLC12) while urban areas in MAB are represented by and roofs. Finally, LCZ 1 and 4 are composed of modern buildings, so the Local Climate Zone (LCZ) (Brousse et al., 2016) (Fig. 2). their respective thermal properties are used. For all classes, the physical CLC12, originally at 100 m resolution, consists of 44 land use properties of asphalt have been set for the ground. Concerning the classes. In order to be incorporated into the WRF model, CLC12 classes parameters for the rooms in buildings, air conditioning systems are have been remapped to the 20 MODIS land use classes following Pineda supposed to be turned on all day, with a target temperature of 25 °C, for et al. (2004), to allow the compatibility with the Noah LSM. Further- all LCZ, with a comfort range of 0.5 °C. Moreover, equipment inside more, CLC12 have been interpolated at each model domain resolution, each room produce a peak heat of 20 W/m2 for all LCZ, apart for LCZ 1, choosing the most present value for each cell. 4 and 8, that are supposed to produce a peak heat of 36 W/m2, due to The LCZ is a product of the level 0 methodology proposed by the higher density of people in LCZ 1 and 4 and higher density of equip- World Urban Database Access Portal Tool (WUDAPT) consisting on ment for industrial purposes in LCZ 8. classifying urban and rural lands with similar climatic characteristics, form and function (Brousse et al., 2016; Stewart and Oke, 2012; Zonato 2.3. Observations et al., 2020). In total, the LCZ comprises 17 land use classes, of which 10 are dedicated to urban areas. The WRFV3.9.1 used in this study The observed data used to evaluate the WRF output are provided by includes appropriate adaptations to the code (Brousse et al., 2016) the Meteorological Service of Catalunya (MeteoCat). As the LCZ in- based on the instructions by Martilli et al. (2016). in this way, the 10 formation is not available for the whole simulation domain, only those urban classes (from LCZ) plus the “Bare rock or paved” (considered as stations placed within the area with LCZ information were selected asphalt, class E in Stewart and Oke, 2012) are taken by the model with (Fig. 3) for the evaluation performance of WRF modelling system. specific thermal and geometric parameters. Thus, BEP and BEP + BEM Table 3 compiles information on meteorological stations, namely co- consider 11 urban classes (Figure 2, LCZ: 1–10 and E) instead of the ordinates, variables measured, sensor height above the ground as well traditional three assumed within WRF (1. Low density residential, 2. as the land use class in the LCZ database at 100 m × 100 m horizontal High density residential and 3. Commercial). resolution and in the model cell (1000 m × 1000 m) corresponding to The LCZ map of Barcelona is created using the WUDAPT method at each station. Due to the re-gridding from LCZ database to model grid, Level 0 (see www.wudapt.org), with a resolution of 100 m. It consists of some information is diluted which leads to a potential mismatch of the a supervised classification of Landsat images using a Random Forest land use class. An example of this is the group of sites classified as “Low Classification algorithm (Breiman, 2001).The method requires first to plants”. Their majority are placed on lawns and surrounded by building draw polygons of uniform land cover in Google Earth (training areas, at a distance larger enough to not be present in the 100 m × 100 m cell TA). Then the algorithm combines spectral bands with the TA defined of the original land use database but at less than 1000 m distance. before to obtain the LCZ classification to be used as input for WRF. In Another example is the Zoo station, located in a city-park (the Parc de la this case, Landsat images from different days (18/03/2013, 16/07/ Ciutadella) that is classified as “Low plants” in the 100mx100m LCZ 2013 and 10/06/2014) are used. TA are then modified to obtain a database. This park is surrounded by “Compact midrise” areas. After the 3 I. Ribeiro, et al. Atmospheric Research 248 (2021) 105220 Fig. 2. Land use in MAB domain: Corine Land Cover 2012 (CLC12, left panel) and Local Climate Zones (LCZ, right panel). Urban classes are inside the right rectangles (LCZ: 1–10 and E). MAB and Barcelona limits are marked as solid black lines in LCZ map. CLC12 is aggregated into its level 2 of classes (x.x – i.e. Urban fabric is 1.1 in CLC12 and it includes all subclasses 1.1.x). Table 2 used to determine boundary layer height (PBLH) for further comparison Area and population of MAB and Barcelona city. with modelled data. The use of radiosounding data for this purpose is LCZ class MAB BCN an accepted reference approach in the scientific community (Seibert et al., 2000). According to Stull (2000) and Seibert et al. (2000), the top Area Area Population Area Area Population of the boundary layer is defined as the minimum high where the virtual (km2) (%) (%) (km2) (%) (%) potential temperature (θv) equals the surface value. The same approach is used to determine PBLH by the BouLac parameterization within the 1.Compact high-rise 4.82 0.76 4.77 4.08 4.03 8.02 2.Compact midrise 57.94 9.09 64.14 32.17 31.74 75.80 WRF model. Thus, vertical measurements of temperature (°C), baro- 3.Compact low-rise 18.84 2.96 9.57 2.50 2.47 3.52 metric pressure (hPa), relative humidity (%), and dew point tempera- 4.Open high-rise 0.13 0.02 0.03 0.11 0.11 0.05 ture (°C) were taken into consideration to calculate θv for different al- 5.Open midrise 3.93 0.62 1.60 1.72 1.70 1.46 titudes under non-cloudy conditions, so that PBLH is inferred. 6.Open low-rise 42.82 6.72 5.34 4.03 3.98 1.42 8.Large low-rise 72.97 11.45 6.10 17.35 17.12 4.14 9.Sparsely built 3.28 0.51 0.17 0.19 0.19 0.02 10.Heavy industry 6.22 0.98 0.08 5.56 5.48 0.16 3. Results E.Bare rock or paved 25.46 3.99 2.82 3.77 3.72 2.59 No urban 401.04 62.91 5.39 29.86 29.47 2.83 3.1. Statistical analysis Total 637.45 101.35 The statistical analysis of the simulation performances is presented in Fig. 4, through the correlation factor (R), the systematic error (bias) re-gridding, the park area is split into four model cells what makes and the root mean square error (RMSE). Here, the observation towers “Compact midrise” class taking over park-related classes. As a con- are aggregated in accordance to the land use classes assumed by WRF sequence, the Zoo site is classified as “Compact midrise” instead of (see Table 3). All the experiments present the same error trend: over- “Low Plants”. estimation of relative humidity (bias > 0) and underestimation of The radiosonding measurements launched by MeteoCat and placed temperature (bias < 0). In average, the highest correlations with ob- in the Physics Faculty of University of Barcelona (blue dot in Fig. 3) are servations for relative humidity and temperature, respectively, are 4 I. Ribeiro, et al. Atmospheric Research 248 (2021) 105220 anthropogenic heat fluxes. Thus, it reaches Sabadell station and the entire Valley region with slightly higher relative humidity and lower temperature than the other urban schemes (see also Fig. 7). The Zoo station is presented here (Fig. 6) as an example of how an unreal input information can affect the analysis of results when com- paring with observations. Due to the mismatch of the land use class (see section 2.3), the Zoo station, presents a different behaviour in com- parison to other stations under “Compact midrise”, namely Raval (Fig. 4). Here, the absence of vegetation in this model cell leads towards underestimation of relative humidity and overestimation of tempera- ture, especially by the BEP and BEP + BEM schemes. The flawed be- haviour of this station exacerbates the statistical analysis of “Compact midrise”. In what wind speed concerns, the statistical analysis (Fig. 4) sug- gests that the correlation between model results and observations is higher for BEP (R = 0.74) and BEP + BEM (R = 0.73) than for BULK (R = 0.62). On the other hand, BEP + BEM presents the highest RMSE and bias values for all land use classes analysed excepting to Compact midrise (C. mid) and Open midrise (O. mid), to which bias values are close to 0 m/s and RMSE is reduced to 48% (C. mid) and 26% (O. mid) in comparison to BULK. On the remaining locations, BEP is the scheme with lower RMSE for wind speed. All schemes underestimate wind speed in C. mid, O. mid and CLC12 (no urban classes) while they overestimate in the other locations. Fig. 3. Detail of the MAB domain: location and altitude above the sea (m) of the Unlike for temperature and relative humidity, it is during the sun- meteorological stations used for the evaluation performance of WRF modelling light period (from 10 h to 18 h) that wind speed in BEP + BEM presents system; MAB and LCZ boundaries. a different magnitude of the bias in comparison to the other schemes. At the Raval station, located on building rooftop, in a compact midrise given by BEP (0.71, 0.92), followed by BEP + BEM (0.68, 0.91) and urban structure, about 2 km away from the sea, BEP + BEM keeps wind BULK (0.66, 0.89) (Fig. 4). Nevertheless, BEP is the scheme with the speed bias closer to 0 during all day while BEP and BULK overestimate highest RMSE and bias, mainly during the night (from 19 h to 06 h UTC, wind speed up to 2 m/s in sunlight period. An opposite effect is ob- Fig. 5). BEP predicts, thus, colder and moister atmosphere during night served in ZUni, classified in the model grid as cropland in spite of its time than BULK and BEP + BEM in the entire domain. location is an urbanized area within the city of Barcelona. In ZUni, Comparing BEP + BEM and BULK, the first presents, in general: 1) BEP + BEM underestimates wind speed mainly during sunlight while similar, yet slightly higher correlation for relative humidity and tem- the other two schemes present a bias curve closer to 0 yet on the po- perature, 2) lower RMSE for temperature and 3) similar RMSE for re- sitive side. Sabadell station is located in a model grid cell classified as lative humidity (Fig. 4). The daily profiles of the systematic error, de- Open low-rise. Nevertheless it is originally classified as Low plants, picted in Fig. 5, reveal that the main differences between these two located next to the Sabadell urban fabric. Here, BEP + BEM predicts schemes, still on temperature and relative humidity, occur during weaker winds than the other schemes during the whole period, espe- sunlight period, when sea breeze takes place. During this time period cially from 11 h to 19 h. This might be related to the influence of the sea BEP + BEM bias is closer to 0 than BULK. Out of that time range, both breeze together to the air warming up by the city, more intense within schemes present similar bias trends and magnitudes in all the stations the BEP + BEM scheme due to its accounting of anthropogenic heat analysed. As BULK accounts with less anthropogenic heat fluxes, the fluxes released by building cooling. Inconsistencies between land use Mediterranean air mass that crosses the city is exposed to lower classification in the model grid and the real one (Table 3) might also Table 3 Meteorological stations: measured parameters (Temperature (T), relative humidity (RH), and wind speed and direction (W)), altitude above the ground (m), land use class from LCZ database (at 100 m × 100 m) and in the model grid (at 1000 m × 1000 m) and coordinates. Station acronym Measurement Altitude above the ground (m) Land use (*non-urban class) Latitude Longitude LCZ Model Badalona T, RH 50 Compact midrise Compact midrise 41.452 2.248 Raval T, RH, W 40 Compact midrise Compact midrise 41.384 2.168 Zoo T, RH 2 Low plants* Compact midrise 41.389 2.188 Vallirana T, RH, W 10 Open low-rise Open low-rise 41.382 1.936 Castellbisbal T, RH, W 10 Bush, scrub* Open low-rise 41.479 1.975 Sabadell T, RH, W 10 Low plants* Open low-rise 41.566 2.070 ParetsV T, RH, W 40 Open midrise Open midrise 41.567 2.226 ElPlat T, RH, W 10 Low plants* Large low-rise 41.340 2.080 SCugat T, RH, W 10 Low plants* Large low-rise 41.483 2.080 ZALPrat T, RH, W 10 Low plants* Large low-rise 41.317 2.131 VilanovaV T, RH 10 Bush, scrub* Sparsely built 41.544 2.300 Zuni T, RH, W 10 Low plants* Crop./natural veg. Mosaic* 41.379 2.105 Viladecans T, RH 2 Bush, scrub* Croplands* 41.299 2.038 ObsFabra T, RH, W 10 Dense trees* Evergreen*Needleleaf Forest* 41.418 2.124 PNGarraf T, RH 2 Bush, scrub* Open Shrublands* 41.288 1.908 BocanaSud W 10 Water* Water * 41.317 2.165 5 I. Ribeiro, et al. Atmospheric Research 248 (2021) 105220 Fig. 4. Statistical analysis (R, bias and RMSE – columns) of BULK (black), BEP (grey) and BEP + BEP (green) performance concerning relative humidity, temperature and wind speed (rows). Fig. 5. Daily profile of the bias between observations and BULK (black), BEP (grey) and BEP + BEP (green) for relative humidity, temperature and wind U and V coordinates at Raval, Sabadell and ZUni. 6 I. Ribeiro, et al. Atmospheric Research 248 (2021) 105220 warmed up while passing over Barcelona. This is evident in the three schemes, although in BEP and BEP + BEM the wind at 10 m slows down in the city due to urban drag effect. It is interesting to see that, in spite of BEP and BEP + BEM are schemes physically closer to each other, BEP + BEM presents similar T2m patterns to BULK than to BEP in the morning and evening. As expected, both urban canopy schemes provide a more evident urban heat island feature during day than BULK does. However, BEP simulation results suggest that heat is lost more quickly as the solar intensity decreases. In fact, T2m by BEP are the lowest in the morning and evening in comparison to the other two schemes under study, but higher than BULK at noon. The same behaviour is also visible in PBLH (Fig. 8) and sensible heat fluxes (Fig. 9) (discussed below) that may support the underestimation of temperature and overestimation of relative humidity in the morning and evening (Fig. 5). 3.2.2. PBLH and sensible heat flux The three experiments present significant differences on PBLH as shown in Fig. 8. In general, BULK predicts lower PBLH than the other urban schemes throughout the study area by mid-day, while at night it is typically higher than BEP and lower than BEP + BEM (Fig. 8 and Fig. 9). During the sunlight period (08 h–17 h UTC), both urban canopy schemes predicted similar behaviour on PBLH development. On the other hand, at night, the PBLH by BEP decay significantly in compar- ison to BEP + BEM. This might be caused by less heat released by the urban surfaces to the atmosphere during evening/night, predicted by BEP that, unlike BEP + BEM, does not take into account the extra en- Fig. 6. Daily profile of the bias between observations and BULK (black), BEP ergy for indoor cooling. (grey) and BEP + BEM (green) for relative humidity and temperature at Zoo. Locations not classified as urban, such as Viladecans, PNGarraf, ObsFabra, ZUni and BocadaSud (only ZUni is shown in Fig. 8), are not justify such differences. In fact, BEP + BEM simulates the wind speed treated through urban schemes. All these stations, excepting ZUni, more accurately in Raval, ParetsV and Vallirana (not shown in this present very similar PBLH daily profiles for all the three experiments paper) than in other locations where, due to the conversion from 100 m and lower PBLH heights that those found in urban areas. Unlike the to 1000 m, a different land class than the real one is assumed by the other stations that are surrounded by forest or agricultural areas, ZUni model. On the other hand, “Large low-rise” (L. low) group of stations is itself within the urbanized area of Barcelona. It may suggest that, in presents, in general, low error values associated despites the mismatch Zuni, most of the parameters analysed are affected by the city fluxes, of land use class due to the re-gridding process. This happen because occurring in the surroundings, via advection. In this case, the PBLH “Large low-rise” are areas generally characterized by low vegetation, predicted by BEP and BEP + BEM is somewhat higher than the BULK with sparse trees and sparsely build, which is not very different to “Low one. plants”. The impact of mismatch is most notorious when the station gets The sensible heat flux is negative in the night for BEP in Raval and a non-representative classification of the reality as in Zoo, classified as Sabadell and for all schemes in Zuni, indicating the occurrence of ra- compact midrise. diative cooling which is potentiated by high levels of humanity. In fact relative humidity is typically overestimated by BEP at night (Fig. 5). The time series of modelled PBLH and the one observed from 3.2. Special and in situ analysis radiosonde measurements at 12 UTC in Barcelona (Physics Faculty) is presented in Fig. 10. From this plot, it is notable that BULK prediction is 3.2.1. Temperature and wind speed the closest one to the observed values. This might be due to too high Fig. 7 presents the averaged wind at 10 m (wind speed and direc- heat fluxes predicted by the urban scheme, however, having more tion) and surface temperature at 2 m (T2m) fields over July 2016 at measurements to estimate PBLH on other periods of the day, namely by 05 h, 12 h and 19 h UTC, obtained by BULK, BEP and BEP + BEM ex- the sunrise and sunset would be crucial to better analyse the real/model periments. PBLH development. Nevertheless, BULK also overestimates PBLH in During the day, BULK predicts wind and surface temperature pat- about half of the days. This made us to suspect that the performance of terns in a more homogeneous way across the domain than the other the three schemes under study might be influenced by different sy- schemes here used, to which the compact urbanized areas are hot spots noptic conditions, as the vertical profiles of T, θ and θv (Fig. 11) sug- with higher T2m (Fig. 7). This is mainly because BULK does not allow gest. the discretization of urban areas in different land use and physical The analysis performed highlights that WRF model, namely BULK property classes, while BEP and BEP + BEM consider 11 classes with scheme, provide a good prediction of PBLH under high pressure system specific physical characteristics (e.g. heat capacity, thermal con- condition, as observed on July 30 (Fig. 10 and Fig. 11a) as well as in the ductivity, albedo, among many others). Moreover, BEP and BEP + BEM day before and the following day. The synoptic condition on these days consider building structures and street canyons as well aiming a more is considered most common in the Iberian (IP) during summertime realistic simulation of air flows and heat fluxes within the urban area, (Valverde et al., 2015), characterized by the presence of a high pressure which are more intense during the day. system over Azores islands and a low pressure system over Scandinavia. According to the results, the Valley region, north of Barcelona This situation is well captured by BULK. However, BEP and BEP + BEM (comprehending Sabadell and Sant Cugat del Vallés cities) presents overestimate the PBLH, which might be related to a higher surface higher temperatures than in Barcelona due to the direct influence of the sensible heat flux then BULK. sea that cools down the air temperature in Barcelona (Figs. 7, 12 h From 10 to 12 July (10 July is depicted as an example in Fig. 11b), UTC). The Mediterranean air mass transported towards the Valley is the three schemes significantly overestimate PBLH (bias > 0). During 7 I. Ribeiro, et al. Atmospheric Research 248 (2021) 105220 Fig. 7. Averaged wind fields (10 m) and T2m at 05 h, 12 h and 19 h (UTC) regarding the month of July 2016: BULK (top panel), BEP (central panel) and BEP + BEM (bottom panel). The black marks place Barcelona city (circle), Sabadell (square) and S. Cugat del Vallès (star). these days, the high pressure system is over the Azores islands and, at were applied over the Metropolitan Area of Barcelona at high resolution 500 hPa, there is a well-defined anticyclone ridge over the Iberian Pe- using detailed urban land use from the Local Climate Zone (LCZ) of the ninsula (IP), which makes the atmosphere highly stable above the MAB. The results were then analysed and compared with a default WRF boundary layer. Moreover, the solar radiation in the IP boosts the for- setup (BULK) and observations from local meteorological towers and mation of a low thermal system (Millán et al., 1991). However, this low radiosonde. The analysis performed was based on temperature, relative pressure is not strong enough to face the high pressure system above humidity and wind (U and V components) for the entire month of July (Fig. 11b). 2016, as well as on boundary layer height and heat fluxes. A low pressure system placed over the Atlantic, in between Great As expected, significant differences were found among BULK and Britain and Greenland, invades southwards to IP latitudes provoking both multi-layer urban canopy scheme simulations. The use of such North Atlantic advection to Western of IP, bringing saturated air (from schemes and LCZ made the study area warmer during the daytime while 18th to 20th). On 21st, the low pressure system dissipates and the in the evening the heat over urbanized areas is more quickly dissipated moisture air warms up leading to atmospheric unstable conditions than in BULK simulations. This is associated to the overall more de- (Fig. 11c). In fact, according to the meteorological stations in Barce- tailed representation of the urban canopy, in particular the thermal lona, the relative humidity is roughly 20% higher than the monthly properties of the urban surfaces. Additionally, the air conditioning (A/ average. In this case, WRF underpredicted PBLH (bias < 0). Never- C) systems for house cooling considered in the BEP + BEM scheme theless, both BEP and BEP + BEM present a better performance as a induce more heat released to the atmosphere than in both BEP and consequence of their overprediction trend. BULK, making the urbanized areas warmer in BEP + BEM experiment The synoptic analysis was based in the Global Forecast System (GFS) than in BULK and BEP, not only during sunlight but also in the night. analysed fields in the Wetterzentrale archive (www.wetterzentrale.de). The boundary layer heights predicted by both multi-layer urban canopy schemes are typically higher than those by BULK and the ones 4. Summary and conclusions derived by measurements. Nevertheless, BULK, BEP and BEP + BEM presented the same bias signal. During July 2016, three synoptic con- High quality input data and parameterizations are crucial to obtain ditions with different behaviours on PBLH calculation were identified: accurate modelling results. Within this work, the two multi-layer urban 1) Good prediction by BULK and overestimation by the BEP and canopy schemes available in the WRF model, BEP and BEP + BEM, BEP + BEM when the high pressure system over the Azores islands and 8 I. Ribeiro, et al. Atmospheric Research 248 (2021) 105220 Fig. 8. Averaged differences in PBLH height: BEP – BULK (top panel), BEP + BEM – BULK (central panel) and BEP + BEM – BEP (bottom panel) over 05 h, 12 h and 19 h of July 2016. The black marks locate Raval, Sabadell and ZUni meteorological stations. Fig. 9. Daily profile of PBL height (m) and the sensible heat flux (W/m2) in: Raval, Sabadell and ZUni meteorological stations. a low pressure system over Scandinavia; 2) Strong overestimation of predicted PBLH more accurately than BULK for the situation 3), mainly PBLH for all the simulations under a well-defined anticyclone ridge due to their overestimation tendency. over the IP and low thermal at surface; 3) Underestimation under un- Regarding the statistical analysis of the three simulations perfor- stable atmospheric conditions promoted by a North Atlantic advection mance over Barcelona urban area during July 2016, BEP has the best that brought saturated air towards IP. The multi-layer urban schemes correlation with observations. On the other hand, this is the scheme 9 I. Ribeiro, et al. Atmospheric Research 248 (2021) 105220 Fig. 10. PBLH measured at 12 h UTC (black dots) and hourly-based time series predicted by BULK (blue line), BEP (magenta line) and BEP + BEP (green line) at Physics Faculty. with the highest value of bias and RMSE for most of the variables and 100 m × 100 m. It is, thus, a limitation imposed in part by the model urban classes, in particular during the night: overestimation of relative resolution. Hereupon, further studies involving high quality input data: humidity and underestimation of air temperature. This phenomenon is detailed land cover classification (e.g. LCZ), more accurate information also verified on daily profiles of PBLH and sensible heat fluxes, which is on urban canopy structure and on urban surfaces physical/thermal likely related to the sea-land-mountain breezes developed within the characteristics; as well as higher horizontal model resolution are study region. BEP bias and RMSE are more prominent than other strongly recommended to perform urban simulations. schemes. In fact, BEP has lower sensible heat fluxes mainly because it neglects anthropogenic heat and probably because the indoor tem- Author statement perature assumed is too cold, which may induce most of the difference in temperature. In addition, due to the higher overestimation of relative Term Definition humidity by BEP at night, the radiative cooling effect might be stronger than in the other schemes, contributing to the reduction of the air Conceptualization This work was part of the 1st year of the Post-doc research temperature and then the underestimation of this physical property. grant of Isabel Ribeiro, entitled “Quantifying the impact of BEP + BEM performed with the lower RMSE associated for relative green infrastructures on (peri-) urban atmospheres (GUrbs)” that took place at the Autonomous University of Barcelona and humidity and temperature all over the domain, as well as for wind was founded by P-SPHERE project under Marie Skłodowska- speed at compact midrise (main LCZ class in Barcelona city) and open Curie program. Supervised by Gara Villalba midrise urban structures – locations where the LCZ in the grid model at The main goal of this work is to evaluate the performance of 1 km × 1 km is in accordance to the original LCZ database at the urban schemes in WRF over Barcelona. Fig. 11. Vertical profiles of temperature (T), potential temperature (θ) and virtual potential temperature (θv) from radiosonding measurements, as well as the PBLH predicted by BULK (dotdashed line), BEP (dashed line) and BEP + BEP (solid line) at Physics Faculty, on a) 30th, b) 10th and c) 21st of July 2016 at 12 h UTC. Systematic error (bias, model-observation) concerning the days and hour of each vertical profile. 10 I. Ribeiro, et al. Atmospheric Research 248 (2021) 105220 Methodology The urban land use of the Metropolitan area of Barcelona was Appendix A. Supplementary data improved through the use of LCZ and CLC databases. The performance of the BULK and two multi-layer urban schemes in WRF (BEP and BEP + BEM) were assessed using observed Supplementary data to this article can be found online at https:// data. doi.org/10.1016/j.atmosres.2020.105220. The methodology was designed by Isabel Ribeiro, Alberto Martilli, Andrea Zonato and Gara Villalba References Software WRF V3.9.1 (code adaptation by Andrea Zonato; simulations by Isabel Ribeiro Phython scripts developed specifically for this work (Isabel Amorim, J.H., Rodrigues, V., Tavares, R., Valente, J., Borrego, C., 2013. CFD modelling of Ribeiro and Marcus Falls) the aerodynamic effect of trees on urban air pollution dispersion. Sci. Total Environ. 461–462, 541–551. https://doi.org/10.1016/j.scitotenv.2013.05.031. QGIS 3.3 (Andrea Zonato) Banks, R.F., Tiana-Alsina, J., Baldasano, J.M., Rocadenbosch, F., Papayannis, A., Solomos, Validation Some simulations and the analysis of the results were repli- S., Tzanis, C.G., 2016. Sensitivity of boundary-layer variables to PBL schemes in the cated within the producing phase of the work to make sure the WRF model based on surface meteorological observations, lidar, and radiosondes results presented are correct. during the HygrA-CD campaign. Atmos. Res. 176–177, 185–201. https://doi.org/10. Isabel Ribeiro and Alberto Martilli made sure all the results 1016/j.atmosres.2016.02.024. present are correct. Bougeault, P., Lacarrere, P., 1989. Parameterization of Orography-Induced Turbulence in Formal analysis The analysis and discussion of the results were mainly per- a Mesobeta–Scale Model. Mon. Weather Rev. https://doi.org/10.1175/1520- formed by Isabel Ribeiro with important contributions by 0493(1989)117<1872:POOITI>2.0.CO;2. Alberto Martilli. Breiman, L., 2001. Random Forests. Mach. Learn. https://doi.org/10.1023/ Investigation The research involved was conducted by Isabel Ribeiro with A:1010933404324. contributions by Alberto Martilli and Andrea Zonato. Isabel Brousse, O., Martilli, A., Foley, M., Mills, G., Bechtel, B., 2016. WUDAPT, an efficient land Ribeiro also performed all the experiments, prepared the use producing data tool for mesoscale models? Integration of urban LCZ in WRF over observed/modelled for analysis and the analysis/discussion Madrid. Urban Clim. https://doi.org/10.1016/j.uclim.2016.04.001. itself. de la Paz, D., Borge, R., Martilli, A., 2016. Assessment of a high resolution annual WRF- BEP/CMAQ simulation for the urban area of Madrid (Spain). Atmos. Environ. 144, Resources The simulations and post-processing was done in the compu- 282–296. https://doi.org/10.1016/j.atmosenv.2016.08.082. tational cluster “Port d'Informació Científica” (PIC) https:// Giannaros, C., Nenes, A., Giannaros, T.M., Kourtidis, K., Melas, D., 2018. A comprehen- www.pic.es/ sive approach for the simulation of the Urban Heat Island effect with the WRF/ Computational resources were paid by the URBANCO2FLUX SLUCM modeling system: the case of Athens (Greece). Atmos. Res. 201, 86–101. grant (Marie Skłodowska-Curie) (Gara Villalba) https://doi.org/10.1016/j.atmosres.2017.10.015. Data Curation Isabel Ribeiro made sure that the model setup and source code, Gidhagen, L., Olsson, J., Amorim, J.H., Asker, C., Belusic, D., Carvalho, A.C., Engardt, M., as well as pre- and post-processing are available to be used by Hundecha, Y., Körnich, H., Lind, P., Lindstedt, D., Olsson, E., Rosberg, J., Segersson, the students and researchers currently involved in the on-going D., Strömbäck, L., 2020. Towards climate services for European cities: Lessons learnt research activities at ICTA-UAB under the purview of Gara from the Copernicus project Urban SIS. Urban Clim. https://doi.org/10.1016/j.uclim. Villalba 2019.100549. Writing - Original The paper was mainly written by Isabel Ribeiro that counted to Gilabert, J., Llasat, M.C., Corbera, J., 2018. Using local climate zones methodologies to Draft the scientific and grammar revision by Alberto Martilli and assess vulnerability and potential risk in urban ecosystems at a framework of climate Gara Villalba. change. In: 10th International Conference on Urban Climate/14th Symposium on the Urban Environment. AMS, New York. Writing - Review & The paper was written by Isabel Ribeiro that counted to the Göndöcs, J., Breuer, H., Pongrácz, R., Bartholy, J., 2017. Urban heat island mesoscale Editing scientific and grammar revision by Alberto Martilli and Gara modelling study for the Budapest agglomeration area using the WRF model. Urban Villalba. Clim. https://doi.org/10.1016/j.uclim.2017.05.005. The reply to reviewers was also mainly performed by Isabel Hammerberg, K., Brousse, O., Martilli, A., Mahdavi, A., 2018. Implications of employing Ribeiro with contributions by Alberto Martilli. detailed urban canopy parameters for mesoscale climate modelling: a comparison Visualization All work involved in post-processing of data, data visualization between WUDAPT and GIS databases over Vienna, Austria. Int. J. Climatol. https:// and presentations were done by Isabel Ribeiro. doi.org/10.1002/joc.5447. Supervision The scientific leadership is by Isabel Ribeiro with a strong Kuik, F., Lauer, A., Churkina, G., Denier van der Gon, H.A.C., Fenner, D., Mar, K.A., contribution by Alberto Martilli Butler, T.M., 2016. Air quality modelling in the Berlin–Brandenburg region using Oversight by Gara Villalba since the work was part of her main WRF-Chem v3.7.1: sensitivity to resolution of model grid and input data. Geosci. projects. Model Dev. 9, 4339–4363. https://doi.org/10.5194/gmd-9-4339-2016. Project administra- The management and coordination of responsibility for the Kusaka, H., Kondo, H., Kikegawa, Y., Kimura, F., 2001. A simple single-layer urban ca- tion research activity planning and execution was done by Isabel nopy model for atmospheric models: comparison with multi-layer and slab models. Boundary-Layer Meteorol. 101, 329–358. https://doi.org/10.1023/ Ribeiro and Gara Villalba A:1019207923078. Funding acquisition Acquisition of the financial support for the project leading to Li, H., Meier, F., Lee, X., Chakraborty, T., Liu, J., Schaap, M., Sodoudi, S., 2018. this publication by Gara Villalba Interaction between urban heat island and urban pollution island during summer in Berlin. Sci. Total Environ. 636, 818–828. https://doi.org/10.1016/j.scitotenv.2018. 04.254. Liu, Y., Chen, F., Warner, T., Basara, J., 2006. Verification of a mesoscale data-assim- Declaration of Competing Interest ilation and forecasting system for the Oklahoma City area during the joint urban 2003 field project. J. Appl. Meteorol. Climatol. https://doi.org/10.1175/JAM2383.1. The authors declare that they have no known competing financial Martilli, A., Clappier, A., Rotach, M.W., 2002. An urban surface exchange para- meterisation for mesoscale models. Boundary-Layer Meteorol. 104, 261–304. https:// interests or personal relationships that could have appeared to influ- doi.org/10.1023/A:1016099921195. ence the work reported in this paper. Martilli, A., Brousse, O., Ching, J., 2016. Urbanized WRF Modeling Using WUDAPT (WEB Version). Middel, A., Häb, K., Brazel, A.J., Martin, C.A., Guhathakurta, S., 2014. Impact of urban form and design on mid-afternoon microclimate in Phoenix local climate zones. Declaration of Competing Interest Landsc. Urban Plan. https://doi.org/10.1016/j.landurbplan.2013.11.004. Millán, M.M., Artíñano, B., Alonso, L., Navazo, M., Castro, M., 1991. The effect of meso- scale flows on regional and long-range atmospheric transport in the western None. Mediterranean area. Atmos. Environ. Part A, Gen. Top. https://doi.org/10.1016/ 0960-1686(91)90137-V. Morini, E., Touchaei, A.G., Rossi, F., Cotana, F., Akbari, H., 2017. Evaluation of albedo Acknowledgements enhancement to mitigate impacts of urban heat island in Rome (Italy) using WRF meteorological model. Urban Clim. https://doi.org/10.1016/j.uclim.2017.08.001. NCEP/NOAA, 2015. Updated daily. NCEP GDAS/FNL 0.25 D egree global tropospheric The authors would like to thank MeteoCat for providing the ob- analyses and forecast grids. Res. Data Arch. Nation al Cent. Atmos. Res. Comput. Inf. served data for the Barcelona region, and the ICGC for the Local Climate Syst. Lab. https://doi.org/10.5065/D65Q4T4Z. Zones database. This project has received funding from the European Pineda, N., Jorba, O., Jorge, J., Baldasano, J.M., 2004. Using NOAA AVHRR and SPOT VGT data to estimate surface parameters: application to a mesoscale meteorological Union's Horizon 2020 research and innovation programme under Marie model. Int. J. Remote Sens. 25, 129–143. https://doi.org/10.1080/ Skłodowska-Curie grant agreements No 665919 (P-SPHERE project) 0143116031000115201. and No 653950 (URBANCO2FLUX). Rafael, S., Martins, H., Marta-Almeida, M., Sá, E., Coelho, S., Rocha, A., Borrego, C., 11 I. Ribeiro, et al. Atmospheric Research 248 (2021) 105220 Lopes, M., 2017. Quantification and mapping of urban fluxes under climate change: (Technical Report NCAR/TN–475+STR). Application of WRF-SUEWS model to Greater Porto area (Portugal). Environ. Res. Stewart, I.D., Oke, T.R., 2012. Local climate zones for urban temperature studies. Bull. https://doi.org/10.1016/j.envres.2017.02.033. Am. Meteorol. Soc. https://doi.org/10.1175/BAMS-D-11-00019.1. Rafael, S., Vicente, B., Rodrigues, V., Miranda, A.I., Borrego, C., Lopes, M., 2018. Impacts Stull, R., 2000. Meteorology for Scientists and Engineers, 2nd ed. Brooks/Cole, Pacific of green infrastructures on aerodynamic flow and air quality in Porto's urban area. Grove, CA c2000. Atmos. Environ. 190, 317–330. https://doi.org/10.1016/j.atmosenv.2018.07.044. Valverde, V., Pay, M.T., Baldasano, J.M., 2015. Circulation-type classification derived on Salamanca, F., Martilli, A., 2010. A new Building Energy Model coupled with an Urban a climatic basis to study air quality dynamics over the Iberian Peninsula. Int. J. Canopy Parameterization for urban climate simulations-part II. Validation with one Climatol. https://doi.org/10.1002/joc.4179. dimension off-line simulations. Theor. Appl. Climatol. https://doi.org/10.1007/ Vos, P.E.J., Maiheu, B., Vankerkom, J., Janssen, S., 2013. Improving local air quality in s00704-009-0143-8. cities: to tree or not to tree? Environ. Pollut. https://doi.org/10.1016/j.envpol.2012. Salamanca, F., Martilli, A., Yagüe, C., 2012. A numerical study of the Urban Heat Island 10.021. over Madrid during the DESIREX (2008) campaign with WRF and an evaluation of WHO, 2016. Global Report on Urban Health: Equitable, Healthier Cities for Sustainable simple mitigation strategies. Int. J. Climatol. 32, 2372–2386. https://doi.org/10. Development. World Health Organization - UN Habitat https://doi.org/https:// 1002/joc.3398. apps.who.int/iris/handle/10665/204715. Salamanca, F., Yizhou, Z., Michael, B., Fei, C., Alex, M., Shiguang, M., 2018. Evaluation of Xu, Y., Ren, C., Ma, P., Ho, J., Wang, W., Lau, K.K.L., Lin, H., Ng, E., 2017. Urban the WRF-Urban Modeling System coupled to Noah and Noah-MP Land Surface morphology detection and computation for urban climate research. Landsc. Urban Models over a Semiarid Urban Environment. J. Geophys. Res. Atmos. 123, Plan. https://doi.org/10.1016/j.landurbplan.2017.06.018. 2387–2408. https://doi.org/10.1002/2018JD028377. Zonato, A., 2016. Evaluating the Urban Heat Island for Cities of Emilia-Romagna Region Santiago, J.L., Martilli, A., 2010. A Dynamic Urban Canopy Parameterization for through Numerical Simulations. University of Bologna. Mesoscale Models based on Computational Fluid Dynamics Reynolds-Averaged Zonato, A., Martilli, A., Di Sabatino, S., Zardi, D., Giovannini, L., 2020. Evaluating the Navier-Stokes Microscale Simulations. Boundary-Layer Meteorol. https://doi.org/10. performance of a novel WUDAPT averaging technique to define urban morphology 1007/s10546-010-9538-4. with mesoscale models. Urban Clim. https://doi.org/10.1016/j.uclim.2020.100584. Seibert, P., Beyrich, F., Gryning, S.E., Joffre, S., Rasmussen, A., Tercier, P., 2000. Review and intercomparison of operational methods for the determination of the mixing height. Atmos. Environ. https://doi.org/10.1016/S1352-2310(99)00349-0. Web references Skamarock, W.C., Klemp, J.B., Dudhia, J., Gill, D.O., Barker, D.M., Huang, X.Y., Wang, W., Powers, J.G., 2008. A Description of the Advanced Research WRF Version 3 URL1, 2020. https://unhabitat.org/urban-themes/energy/ (accessed on 27/02/2019). 12