How to understand the heat island effects in high-rise compact urban canopy?

Alleviating the urban heat island effect is of great significance to improve thermal comfort, energy saving and carbon reduction, and realize sustainable urban development. At present, several methods are developed to investigate urban heat island effect, including meteorological observation data analysis, mesoscale WRF numerical simulation and remote sensing image analysis, etc. Among them, remote sensing image is widely used in the comparative study of heat island effect in different urban development periods. The local climate zone theory (LCZs), proposed by Stewart and Oke (Bull Am Meteorol Soc 93:1879–1900, 2012) provides a new tool for the downscaling study of urban heat island effect and forms a systematic classification scheme for different urban forms and surface landscapes. The results currently using LCZs to study the heat island effect, usually illustrate the horizontal differentiation at pedestrian level. However, the high-rise compact urban canopy of megacities in China is characteristic of three-dimensional space pattern, leading to the three dimensional differentiation of urban thermal and wind environment. Together with the local climate zones, the two-layer analysis scheme of the surface building-vegetation mixing layer and high building effect layer is thus proposed in this short review to understand the three-dimensional differentiation of urban canopy. This two-layer analysis scheme will provide a new insight for the study of urban heat islands and heat mitigation, deepening the existing local climate zone theory.


Introduction
Currently, more than half the world's population lives in cities (Grimm et al. 2008). Concentrated economic and human activities have greatly changed the underlying surfaces and forms a unique urban climate environment (Han et al. 2022). Urban heat island effects (UHI) has become one of the main urban environmental problems (Memon et al. 2008), which is the phenomenon that the air temperature in the urban area is significantly higher than that in the suburbs. The greater the temperature difference, the stronger the urban heat island effect (Oke et al. 2017).
UHI is formed by the combined effects of human factors and near surface meteorological factors. Among the human factors, the change of the underlying surface (land use shortcomings for three-dimensional evaluation in high-rise urban canopies. And a vertical stratification scheme is established for the study of UHI to fill the gap. In this scheme, the urban canopy layer is divided into two layers, one is defined as the Surface Building-Vegetation Mixed Layer (hereinafter referred to as Mixed Layer), which is composed of underlying surface, buildings and all vegetation beneath the top of tree dense canopy, and the other is High Building Effect Layer (hereinafter referred to as Building Effect Layer composed of the buildings above the tree height). This stratification scheme can help to improve the current parameterized model for urban canopy simulation, and obtain a more accurate three-dimensional wind field, temperature field and humidity field inside the urban canopy, which will help to further understand the characteristics of urban heat island effect and explore the influence of current heat mitigation measures on thermal environment through scenario simulations. So as to provide scientific guidance for urban heat island heat mitigation and urban planning. The flow chart of this paper is shown in Fig. 2.

Remote sensing image analysis of urban heat island effect
The research methods of urban heat island effect generally include meteorological data analysis Song et al. 2003), field measurements and analysis (Dong et al. 2011), numerical simulation (He et al. 2020Giannaros et al. 2013) and remote sensing (Yang et al. 2018;Zhang et al. 2015;Nichol and Hang 2012). Among them, remote sensing is widely used in the comparative study of surface heat island effect in different urban development periods (Yang et al. 2018), which belongs to macro scale research (Strategic Consulting Center of Chinese Academy of Engineering 2020). According to the meteorological data, the intensity of the urban heat island in Xi'an has increased annually since the 1970s. From 1981 to 2017, the intensity of urban heat island in Xi'an increased significantly, from 0.3 °C to 1.4 °C . Yang et al. (2018) used the Landsat TM /ETM+ /TIRS remote sensing data in summer and analyzed the temporal and spatial variation characteristics of surface urban heat island effect in Beijing since 1985. It was found that heat island effect in Beijing has the characteristics of overall sub-high temperature and local high temperature. During a heat event in the summer of 2010, the average surface heat island intensity within the Fourth Ring of Beijing was 2.93 °C and the maximum value was 5.5 °C (Zhang et al. 2015). In addition, modern cities face the dual effects of urban heat island and heat wave (HW), and the heat island effect strengthens the impact of heat wave (Rizvi et al. 2019). He et al. (2020) used WRF (weather research and forecasting model) and observation data to simulate a heat wave event in Beijing. The results showed that the temperature rise of heat wave caused by heat island effect was 0.78 °C, which was greater during the nighttime than that during the daytime.
However, remote sensing models should be treated with caution since they rely only on sensible heat and the conversion of received shortwave solar radiation into latent heat through the mechanism of evapotranspiration is neglected (Völker et al. 2013). As a result, it tends to overestimate daytime temperatures and underestimate nocturnal ones ( (Mohan et al. 2013). In addition, the LST inversed from remote sensing images cannot fully reflect the atmospheric motion or atmospheric environmental parameters. Therefore, it's still necessary to obtain the accurate atmospheric temperature difference between urban canopy and suburbs simultaneously by meteorological data or numerical simulation, to evaluate the intensity of urban heat island effect.

Local climate zone theory and its application
At present, the local climate zone theory is mainly used to study the urban heat island effect Stewart et al. 2014;Middel et al. 2014;Huang et al. 2019). For example, Cardoso and Amorim (2018) analyzed the effects of urban morphology and surface cover on UHI intensity based on LCZ system, it was found that compact built zones have higher temperatures, followed by open midsize, lightweight low-rise, and low plants zones. Kaloustian and Bechtel (2016) applied local climatic zoning mapping for Beirut, a dense city situated along the Mediterranean Sea, and provides details on the dominating urban form and materials of the city thus contributing to the identification of the main parameters that can promote the formation of an urban heat island (UHI) in the city. They also evaluated the thermal effect of LCZs according to the results of urban energy balance (TEB) model. Wang and Ouyang (2017) proposed a modified version of LCZ framework which is referred as Local Thermal Zone (LTZ). By figuring out the impacts of land surface indicators on the land surface temperature in each LTZ, heat mitigation strategies with detailed guidelines can be explored.
In addition to the basic classification of LCZ mentioned in Introduction, many researchers further subdivide LCZ to give a more accurate description of urban thermal environment. For example, Anjos et al. (2020) identified six additional LCZ sub-classes in Londrina to give a better description of the city's land cover. According to the LCZ map and air temperature spatial distribution, the LCZ1 3 and LCZ 3 were considered to be zones for large UHI intensity, and LCZG A and LCZ9 Gw were considered to be zones that need to be preserved and promoted. Zhang et al. (2016) extended the LCZs classifications when simulating the urban wind environment in Xi'an City. The "medium high" buildings were subdivided into two types: medium high I (building height 4-6 floors) and medium high II (building height 7-15 floors). Buildings with more than 16 floors are classified as high-rise buildings to adapt to the characteristics of urban buildings in China. This is because a large number of buildings in cities of China have a height of more than 30 floors, and Stewart & Oke's LCZs classifications were not fully applicable (Zhang et al. 2021).
To further characterize the temporal and spatial differences of the thermal and humid environment within the urban canopy, urban climate research experts from the fields of architecture, urban geographic science and environmental science put forward the World Urban Database and Access Port Tools (WUDAPT) which is suitable for urban climate research (www. wudapt. org). It comprises information on urban form and function that has been gathered in a consistent manner and captures variation across the urbanized landscape (Ching et al. 2014(Ching et al. , 2015Mills et al. 2015). The database will be structured into levels, which indicate the level of detail available for a city. The lowest level (level 0) data employs the Local Climate Zone (LCZ) typology to classify a city and its surrounding area. For example, Brousse et al. (2016) used WUPAPT database to improve the Building Effect Parameterization and Building Energy Model (BEP-BEM) schemes in WRF, and discussed the application potential of LCZs in mesoscale WRF simulation. In their work, images from the Landsat 8 satellite are used in SAGA GIS software to discriminate between the spectral attributes of different LCZ types based on training areas. And Corine Land Cover data were extracted from a previous work (Salamanca et al. 2012) which used the database of the year 2000 from the European Environment Agency. Through the numerical simulation of winter and summer in Madrid, Spain, the simulation results using LCZs data and CORINE land cover data were compared to the measurements from a network of 24 meteorological stations respectively and root mean square error (RMSE) was used to estimate the error of the simulation. The results showed that LCZs scheme improved the performance of WRF model. A slight improvement could be seen in both cases, about 5% averaged in winter and 10% in summer. It was also pointed out that it is necessary to further enrich the basic urban data in order to expand the value of LCZs model in urban planning. Shi et al. (2021) studied the heatwave spatial patterns by utilizing the WUDAPT Level 0 data. The LCZ maps were generated for both of the years 2009 and 2016 in Guangdong, China. By comparing the change of LCZ maps with heatwave conditions (i.e. the number of hot days in each heat events and frequency of heatwave events in each year), it was found that the effect of urbanization on heatwave conditions becomes stronger during the past several years. Zonato et al. (2020) proposed a modified WUDAPT method to define urban morphology whose urban morphology parameters were obtained from LIDAR data. This technique is tested by means of simulations with the Weather Research and Forecasting (WRF) model at 500 m resolution for the city of Bologna (Italy), located in the Po Plain. Simulation output is compared with measurements from weather stations. Results showed that simulations using the modified WUDAPT method reproduce better atmospheric dynamics with respect to those implementing the standard WUDAPT method.

Multi-scale numerical simulations of urban wind environment
According to the characteristics of the scales of the research object, the numerical simulation models of urban atmospheric environment can be divided into Urban Scale Model, Urban Sub-domain Scale Model (USSM) and Urban Street Canyon Model (USCM) (Wang et al. 2005), and the urban scale atmospheric motion is controlled in larger scale (horizontal scale of hundreds of kilometers). Mesoscale models, such as WRF, the simulation area usually cover the whole city and the horizontal spatial resolution of the model is greater than 1 km. This model focus on simulating the changes of atmospheric environment and meteorological conditions at the urban scale, which can basically reflect the climate and environmental information at the height of urban residents. The Regional Boundary Layer Model (RBLM) is a three-dimensional non hydrostatic regional meteorological prediction model which developed on the basis of WRF. It is suitable for the simulation of complex underlying surface conditions, and the horizontal grid distance is as small as 0.5 ~ 1.0 km. In this model, the influence of urban buildings is introduced into the horizontal motion equation and turbulent kinetic energy equation to reflect the drag effect of buildings on urban wind field. And an artificial heat source term is added to the ground temperature prediction equation to take the special thermal effect of cities into account (Wang et al. 2012;Miao 2017). In addition, urban mesoscale numerical simulation results have shown that only in simulations with 1 km × 1 km highresolution grids, the introduction of anthropogenic heat flux can effectively improve the spatial distribution of meteorological elements, especially in the high-value anthropogenic heat areas in big cities (Feng et al. 2014).
For local scale wind environment, the detailed information of thermal and humid environment perceived by human bodies is of great significance to urban planning and design. In the urban canopy, buildings of different heights are randomly distributed and roads are crisscross. There are great differences in the physical properties of the building facades and the greening of the street surface, resulting in the very complex characteristics of the underlying surface; the atmospheric stability in the street valley changes with the diurnal variation of solar radiation; even if the atmospheric flow is stable, it usually presents the characteristics of coherent structure turbulence after passing through the upstream buildings, which has a complex impact on the downstream flow (Gu et al. 2011). For local scale atmospheric flow simulation or wind environment assessment of urban canopy buildings, large eddy simulation method has become an effective and high-precision numerical method, and remarkable achievement has been made (Cui et al. 2013;Gu and Zhang 2014). The large eddy simulation method of time-varying incoming wind environment at the building and block scale has been further developed in many researches (Li et al. 2019;Fan et al. 2022).

Two-layer analysis scheme for the urban thermal and wind environment of high-rise compact canopy
The local climate zone theory (LCZs) proposed by Stewart and Oke (2012) describes the urban underlying surface characteristics of different urban forms and surface landscapes, forming a systematic classification scheme with multiple uses. However, when LCZs is used to study the heat island effect at present, both the analysis of meteorological data and the inversion of surface and atmospheric temperature from remote sensing images only reflect the temperature distribution on the two-dimensional horizontal spaces. For example, Leconte et al (2015) conducted mobile survey to gather the air temperature in different LCZs at 2 m height. It was found that the average nocturnal temperature difference between pairs of LCZ types varied from less than 1℃ for close LCZ types to more than 4 ℃ for dissimilar LCZ types. Badaro-Saliba et al. (2021) studied SUHI intensity between different LCZs by the use of satellite Land Surface Temperature data. It was found that the difference in average temperatures between high-rise densely built LCZs and mostly pervious zones exceeded two degrees at nighttime.
However, for high-density urban canopy with high proportion of high-rise buildings, the thermal and humid environment has differentiation in three-dimensional spaces. Therefore, for fully understanding the heat island effect and the effectiveness of thermal mitigation measures in high-density urban canopy with high proportion of high-rise buildings, it's necessary to study the effects of three-dimensional spatial differentiation characteristics on the thermal and humid environment, which is usually achieved by local scale numerical simulation. As the thermal dynamic process involved in the urban canopy is extremely complicated, the current canopy model needs to be further improved. To date, many researchers have proposed different parameterization schemes for urban canopy. For example, Santiago et al. (2010) and Krayenhoff et al. (2020) parameterized the influence of buildings and vegetation on wind and turbulent kinetic energy by adding source terms to momentum equation and turbulent kinetic energy equation, so as to obtain the wind field. Yan et al. (2020) and Redon (2020) parameterized the influence of buildings, vegetation and underlying surfaces on air temperature and humidity by adding source terms to air temperature equation and water vapor equation, so as to obtain the air temperature and humidity field. The source terms given by Redon is shown as follows: (1) Where S T and S q are the source terms for air temperature and humidity, Q H and Q E are the sensitive and latent heat exchange between the surfaces (roof (R), road (r), wall (w), natural soil & low vegetation (nat) and trees (t)) and the canopy air. S is the surface area and V air is the total air volume.
In Redon's parameterized scheme, the heat fluxes of walls don't vary with height as trees are as tall as buildings. In this paper, we focus on the thermal environment of compact high-rise urban canopy (LCZ1 and LCZ 2(i)) in cities of China, where the building height is much higher than that of trees. The LCZ classification standard used in this paper was proposed by He et al. (2019) where LCZ 1 is defined as a land parcel with compact high-rise buildings with a height of 16 storeys or more, and LCZ 2(i) is defined as a land parcel with compact mid-high-rise buildings with a height of 7-15 storeys (LCZ 2(i) occupies a low proportion in Xi'an and it is mapped together with LCZ 1 (Fig. 1)). Therefore, the heat fluxes of walls vary with height, and the parameterized scheme is supposed to be modified.
It is suggested in this paper to divide the vertical urban canopy into two layers with the tree height as the boundary, i.e. the Mixed Layer and the Building Effect Layer, as shown in Fig. 3.
In the mixed layer, the interactions among buildings, vegetation and the underlying surfaces need to be fully considered. For example, the multiple reflection and absorption of radiation in the mixed layer; mutual shading between buildings and vegetation; and the shading of buildings and vegetation on the underlying surfaces. In the building effect layer, the effects of vegetation and underlying surfaces are ignored. Through such vertical stratification, we can express the impact of various elements in the canopy on the thermal environment more accurately.
Based on such vertical stratification scheme, the Eqs. (1) and (2) are modified as: (3) Fig. 3 The vertical stratification analysis scheme of urban canopy heat island effect Where s represents the influence of shading and ns represents the situation without shading, m represents the wall in Mixed layer and b represents the wall in Building effect layer.
The latent heat flux and sensible heat flux in the equation are suggested to be obtained through field measurements in this paper. For example, the net radiation exchange between each surface and canopy air can be first measured, and then measure the Bowenratio (Bo) of the underlying surfaces through the profile method (Foken 2017). According to the energy balance, the sum of latent and sensible heat flux can be expressed as: Where Q * is the net radiation, Q H is the sensitive heat flux, Q E is the latent heat flux and Q G is the heat storage.
Q G can be measured by soil heat flux-plates or be estimated by empirical formula (Foken 2017): Therefore, the sensible heat flux and latent heat flux of each surface are obtained as: Finally, substituting the sensible heat and latent heat fluxes measured into the equation, an accurate urban canopy temperature and humidity field can be obtained through local scale numerical simulation.
As the area of each surface in Eqs. (3) and (4) is influenced by the configuration of vegetation, buildings and underlying surfaces in the canopy. Therefore, this parameterization scheme should be combined with LCZs method. The differences of heat island effect in different LCZs can be explored, and an accurate comparison of heat island effect within cities or between suburbs can be achieved.
In addition, with this canopy model, the mechanism of different thermal mitigation measures (such as cool facade Technology (Xiang and Zhang 2018)) to improve the thermal environment in urban canopy can be identified through scenario simulations by changing the proportion or types of vegetation, underlying surface and buildings, which can provide a scientific basis for UHI mitigation and urban planning.

Conclusion
In the process of urbanization, big cities in China have a large number of buildings with the height of more than 100 m, showing a three-dimensional development trend of intensive and tall buildings. As a result, the height of urban canopy has increased (4) significantly, and the residential and living patterns formed by the three-dimensional development of cities lead to concentration of population, consumption and emission (Zhang and Gu 2013). Therefore, it is necessary to further understand the heat island effect in the high-density urban canopy with high proportion of high-rise buildings. This relies on the accurate numerical simulation of the thermal environment of urban canopy with different structures.
In this paper, the urban heat island effect and urban heat mitigation studies were briefly reviewed, and main research methods and their shortcomings at present were analyzed. A vertical stratification scheme was proposed for the study of UHI in highrise and high-density canopy spaces in this paper. The contribution of this scheme is as follows: Firstly, through this vertical stratification scheme, the differences of the effects of the elements in the canopy (buildings, vegetation, underlying surface) on the thermal environment at different heights in the three-dimensional spaces can be expressed more accurately.
Secondly, through the combination of high-resolution numerical simulation and field measurements, an accurate temperature and humidity field in the urban canopy with different configurations can be obtained. The differences of heat island effect in different climate zones can be explored, and therefore, an accurate comparison of heat island effect within cities or between suburbs can be achieved.
Finally, by using the high-resolution numerical simulation method, the effectiveness of urban heat mitigation strategies and technologies can be discussed through scenario simulation by changing the proportion or types of vegetation, underlying surface and buildings in urban canopy, which contributes to scientific evaluation for urban planning (Ren and Wu 2012).
Recommendations for future studies are as follows: 1. Further expand the LCZs theory, so as to adapt to the rapidly changed urban underlying surfaces. 2. Focus on the study of the thermal environment in the urban canopy in the threedimensional space, rather than only on two-dimensional horizontal space, so as to better understand the effects of UHI. 3. Further study the effectiveness of existed heat mitigation measures (such as cool roofs, green roofs, high-reflectance facades) in different LCZs, so as to provide scientific guidance for urban heat island heat mitigation and urban planning.