Effective Factors for Reducing Land Surface Temperature in Each Local Climate Zone Built Type in Tokyo and Shanghai

Abstract

Identifying the main factors influencing the land surface temperature (LST) of each local climate zone (LCZ) built type is of great significance for controlling LST. This study investigated the main factors influencing the LST of each LCZ built type in two Asian megacities: Tokyo and Shanghai. Each area in both megacities was classified according to the LCZ scheme. The diurnal LST, pervious surface fraction (PSF), surface albedo (SA), average building height ($⟨BH⟩$), and gross building coverage ratio (λp) of each LCZ were also calculated. Finally, the influence of the properties of each LCZ built type on LST was investigated. The results demonstrated that the main factors influencing LST of different LCZ built types differed in Tokyo and Shanghai. $⟨BH⟩$ was the main factor influencing LST for compact mid-rise and open high-rise types in Tokyo, and the compact mid-rise type in Shanghai; PSF was the main factor influencing LST for other LCZ built types. Moreover, both $⟨BH⟩$ and PSF negatively correlated with LST. Based on the above results and characteristics of each LCZ built type, specific LST mitigation strategies for each LCZ built type were proposed for Tokyo and Shanghai. The approach in this study can contribute to perspectives for urban planners and policymakers to develop highly feasible and reasonable LST mitigation strategies.

1. Introduction

1.1. Understanding and Mitigating Urban Heat Islands through Land Surface Temperature Analysis

Urbanization and intensive human activities have significantly changed the urban climate and caused a series of environmental problems [1,2]. Under the influence of various factors such as increased urban density, use of building materials, reduction in natural land cover, decrease in urban wind speed, and increase in anthropogenic heat release, the urban thermal balance has been altered, thereby triggering the urban heat islands (UHIs) phenomenon [3,4,5,6,7].

UHIs worsen the conditions in the urban thermal environment and negatively affect energy consumption [8,9,10] and the health of urban residents [5,11,12,13,14]. As UHIs and global warming continue, the conditions in the urban thermal environment will become more severe and cause even greater negative impacts on energy consumption and the health of urban residents [15,16]. Therefore, mitigating UHIs to improve the conditions in the urban thermal environment is an important issue that is receiving increased attention from researchers, urban planners, and policymakers [17,18].

To mitigate UHIs, it is first necessary to correctly describe their conditions. The UHIs were traditionally described by measuring the air temperature at 1–2 m above the ground [5]; air temperature at this height was usually obtained through fixed stations or mobile measurements [19,20,21,22]. However, the number of fixed stations was generally limited, resulting in poor spatial resolution of the observed air temperature. Mobile measurements also had limitations in the measurement area and simultaneity. These make it difficult to correctly describe the condition of UHIs [23,24,25,26]. With the development of remote sensing technology, obtaining high spatial resolution and real-time land surface temperature (LST) over large areas has become possible. LST is related to the air temperature of the lowest layers of the urban atmosphere, and many researchers have used it to describe and study UHIs [23,24,25,26]. In addition, LST also affects the comfort of city dwellers [27,28,29]. Therefore, controlling LST is very important, and understanding the main factors influencing LST is a prerequisite for its effective control [30].

1.2. Literature Review

1.2.1. LST and LST Mitigation

In complex urban areas, LST is influenced by multiple factors [31,32]. Previous studies have examined the influence of land cover, urban morphology, human activities, meteorological conditions, and other factors on LST [33,34,35,36,37,38,39]. Many studies have shown that LST is closely related to land cover type [3,38,40,41]. Vegetation and water bodies usually have low LST, while buildings and roads have high LST [3,38]. LST is also negatively correlated with vegetation coverage and water body area [42,43], and positively correlated with an impervious surface area [41]. The relationships between urban morphology and LST were also been investigated by many studies [33,44,45]. Due to the cooling effect of the shadow of high-rise buildings, the LST in the high-rise building area of the urban center is the lowest and increases from the high-rise to low-rise building areas [45,46]. Meanwhile, LST is also influenced by building density, sky view factor (SVF), and so forth [39]. Land use functions can also influence LST, which is caused by the fact that various land use functions provide specific urban facilities for human activities, meaning that they have different heat absorption and dissipation capabilities [32]. For example, commercial and industrial areas usually have the highest LST, while recreational areas have the lower LST [47,48]. In addition, some studies have also shown that LST is related to population density, the use of fossil fuels, elevation, and precipitation [39,44,49].

To improve the conditions in the urban thermal environment, a series of mitigation strategies have been proposed, primarily focusing on urban greening, surface materials, urban water, and urban form optimization [50]. Among them, urban greening and surface materials are the two strategies most concerned by researchers [4]. Urban greening can reduce the LST through the transpiration of vegetation and the creation of green shade [51,52]. It is also the most well-known strategy for urban residents to improve the conditions in the urban thermal environment [50]. Diverse urban greening strategies span from incorporating parks, green urban spaces, and street trees to incorporation below green such as shrubs, grasses, and yards [22,53,54]. Surface materials contribute to the reduction of LST by using innovative materials such as retroreflective materials, and super cool materials that alter the albedo, heat capacity, or emissivity of urban surfaces [3,55]. Water bodies can reduce LST through water evaporation [42]. This effect can be used by adding water features like ponds, lakes, and fountains as effective measures to reduce LST [15,56]. Meanwhile, the effect of the water body to reduce the LST can be influenced by wind speed [42]. The urban form optimization can reduce LST by optimizing the building height and density, and rationally arranging buildings, landscapes, roads, and so forth [50]. He et al. emphasized the importance of the co-benefits of various strategies in improving the condition in the urban thermal environments and mitigating climate change, thereby achieving sustainable urban development [4]. In addition, the influence of these mitigation strategies on enhancing urban resilience can also be given due consideration, thereby improving the ability of cities to respond to various natural and manmade hazards that may threaten their competitiveness, livability, and function [57].

In addition, a previous study pointed out that due to the different characteristics of land cover, urban morphology, and other factors, the degree of influence of each factor on LST varied across different urban areas [5]. Therefore, the main factors influencing LST differ from one urban area to the next [41]. Understanding these complex interactions and identifying the main factors influencing LST in different areas is essential for effective urban planning, LST mitigation strategies, and the development of heat adaptation measures in urban areas [30].

1.2.2. The Local Climate Zone Scheme and Its Relationship to LST

To better describe the influence of land cover, urban morphology, and human activities on the thermal environment, Stewart and Oke proposed the local climate zone (LCZ) scheme [5]. LCZs consist of a variety of surface covering, structures, and materials, with horizontal extents ranging from hundreds to thousands of meters. The LCZ scheme has 17 standard types: 10 LCZ built types and 7 LCZ land cover types (Figure A1) [5]. These types were established to provide a consistent and structured framework for the study and modeling of urban climates and to facilitate the development of sustainable urban environments [58]. In recent years, the LCZ scheme has been proven effective in identifying areas within cities that display different thermal behaviors, thereby capturing the spatial and temporal variations in LST caused by urbanization and city development [59]. Moreover, as a highly versatile classification method for land cover and urban morphology, LCZ can also be used in different environments to analyze urban characteristics and facilitate comparisons between cities under a unified standard [59,60,61]. Therefore, it is now widely used in urban climate studies and applied to the exploration and formulation of mitigation strategies to reduce the UHI effect [62,63]. In addition, the LCZ scheme has also been used in studies of energy, air pollution distribution, and thermal comfort [22].

The relationship between LCZ and LST has been investigated in multiple cities worldwide and demonstrated that LCZ types were generally consistent with LST, and that each LCZ type had a characteristic LST [22,30,64,65,66,67,68,69,70,71,72]. Overall, LCZ built types usually have higher LST than LCZ land cover types [22,30]. Among LCZ built types, open high-rise and sparsely built types usually have low LST, and compact low-rise, large low-rise, and heavy industry types usually have high LST [71,72]. Among LCZ land cover, dense trees and water types usually have low LST [22,71,72]. In addition, several studies have explored the main factors influencing the LST of different LCZ built types in several large cities, showing that there are differences in the main factors influencing the LST of different LCZ built types, and LCZ can be used as an effective tool to investigate the main factors influencing the LST of different urban areas [30,73,74].

1.3. Research Purpose

Although several previous studies have investigated the main factors influencing the LST of different LCZ built types, these studies only investigate the influence of a single type of property (only land cover properties or only urban morphological properties), which is not comprehensive. In addition, their study areas are all large cities [30,73,74]. Compared with large cities, megacities usually have more complex urban structures, resulting in the conditions of urban thermal environment being more severe, especially for Asian megacities [75,76]. However, there are currently no relevant studies for Asian megacities. Therefore, this study aimed to identify the main factors influencing the LST of each LCZ built type in two Asian megacities: Tokyo and Shanghai.

Section 2 is the introduction of Tokyo and Shanghai. In Section 3, we generate LCZ maps of Tokyo and Shanghai and calculate a series of data using satellite images or geographic information system (GIS) data, including LST, land cover properties, and urban morphological properties of each LCZ built type for each city. In Section 4, we investigate the main factor influencing LST for each LCZ built type. Section 5 discusses the effective LST mitigation strategies for each LCZ built type for each city, and Section 6 summarized the conclusions of this study.

2. Study Areas

This study selected Tokyo in Japan and Shanghai in China as the study areas. Tokyo and Shanghai are highly urbanized global cities [77,78] and were both classified as megacities according to the 2018 World Urbanization Prospects Report of the United Nations Department of Economic and Social Affairs [79].

Tokyo is in the middle of the Main Island facing the Pacific Ocean and is the most populous city in Japan. In 2021, Tokyo had a population of 14.03 million and a total area of 2194 km² [80]. Shanghai is located on the banks of the Yangtze River Delta in the East China Sea of eastern China, and it is China’s economic and financial center. In 2021, Shanghai had a population of 24.87 million and a total area of 6341 km² [81].

Tokyo and Shanghai have a similar average annual temperature, average annual relative humidity, and average annual precipitation. From 2015 to 2020, the average annual temperature in Tokyo and Shanghai was 16.4 °C and 17.5 °C, respectively, the average annual relative humidity was 69.3% and 76.8%, respectively, and the average annual precipitation was 1650.9 mm and 1518.8 mm, respectively. [82,83]. According to the Köppen climate classification, both cities have a Cfa climate (humid subtropical climates), with hot and humid summers and cool to mild winters [84].

As megacities with high urbanization, Tokyo and Shanghai suffer from severe UHIs [78,85,86] and extreme heat events [87] during summer. Tokyo experienced a series of heatwaves in 2010, and the number of heat stroke patients soared to about six times that of previous years; since 2010, the number of heat stroke patients has continued to rise [16]. Heat waves in Shanghai in 2003 and 2013 caused 516 and 1347 deaths, respectively [88]. Shanghai was defined as one of the Chinese cities severely affected by extreme heat exposure [89]. The elderly are more sensitive and vulnerable to the deterioration of conditions in the urban thermal environment [90], and both Tokyo and Shanghai are facing population aging problems [91,92]. Shanghai is the first city in China to enter an aging society, and it is also the city with the highest aging rate [92], reaching 16.3% in 2019 [93,94]. In 2019, the aging rate of Tokyo reached 23.1%, and according to the forecast of the Japanese Cabinet Office, Tokyo’s aging rate will reach 30.7% in 2045 [95]. All of these highlight the necessity and urgency of improving the conditions in the urban thermal environment of these two cities.

3. Materials and Methods

Figure 1 depicts the flowchart of this study, which contains 3 main steps. The first step is to generate LCZ maps of Tokyo and Shanghai using the World Urban Database and Access Portal Tools (WUDAPT) Level 0 (L0) method, and the second step is to calculate LST, two land cover properties (previous surface fraction (PSF) and surface albedo (SA)), and two urban morphological properties (average building height ($\langle BH\rangle$) and gross building coverage ratio (λp)) using satellite images or GIS data. The last step is to use the data of the above steps and the linear regression model to identify the main factor influencing the LST of each LCZ built type in Tokyo and Shanghai. The details of each step are described in this section.

3.1. LCZ Classification Using the WUDAPT L0 Method

In this study, the LCZ maps of two cities were generated using satellite images and the machine learning-based WUDAPT L0 method [61].

The WUDAPT L0 method consisted of three main steps: (1) acquiring and processing satellite images of the study area from the U.S. Geological Survey Earth-Explorer; (2) defining LCZ types within the study area and selecting training samples from Google Earth Pro (GEP); (3) importing satellite images and training samples of the study area into a system for automated geoscientific analyses geographic information system (SAGA-GIS), and using the random forest algorithm (machine learning) in SAGA-GIS to generate LCZ maps. Steps (2) and (3) require continuous iterations to generate suitable LCZ maps [22,61].

In this study, we selected four sets of multispectral satellite images from Landsat 8 Level 1 data representing the four seasons in Tokyo and Shanghai according to the WUDAPT L0 process [96]. Because the GIS data of Tokyo and Shanghai were collected in 2015 and 2018, the satellite images with the lowest cloud cover in Tokyo and Shanghai from 2015 to 2018 were selected to ensure the consistency of the generated LCZ maps and GIS data. The cloud coverage of all the satellite images was less than 10%, as listed in

Table 1. Satellite images for LCZ classification and properties calculations.

Study Area	Entity ID	Date
Tokyo	LC08_L1TP_107035_20160317	17 March 2016
	LC08_L1TP_107035_20160707	7 July 2016
	LC08_L1TP_107035_20180102	2 January 2018
	LC08_L1TP_107035_20181102	2 November 2018
Shanghai	LC08_L1TP_118038_20170402	2 April 2017
	LC08_L1TP_118039_20170402
	LC08_L1TP_118038_20170824	24 August 2017
	LC08_L1TP_118039_20170824
	LC08_L1TP_118038_20180115	11 January 2018
	LC08_L1TP_118039_20180115
	LC08_L1TP_118038_20181217	17 December 2018
	LC08_L1TP_118039_20181217

In addition to being used for LCZ classification, satellite images in bold black were also used for PSF and SA calculations.

. The satellite images of each city were clipped according to their administrative boundaries.

The WUDAPT L0 method requires one to define the LCZ types for the study area and select training samples for these defined LCZ types, and the size of each training sample should be no smaller than 1 km² [96]. The $\langle BH\rangle$ and λp are the most crucial properties for defining the LCZ built types [22]. Therefore, the $\langle BH\rangle$ and λp of Tokyo and Shanghai at a resolution of 1 km were calculated using their GIS data, and the recommended ranges for $\langle BH\rangle$ and λp of each LCZ built type provided by Stewart and Oke [5] were also used to define the LCZ built types in Tokyo and Shanghai. Figure 2 depicts the distribution of $\langle BH\rangle$ and λp at 1 km resolution in Tokyo and Shanghai. The recommended ranges for $\langle BH\rangle$ and λp of each LCZ built type are shown in

Table 2. Recommended ranges for $\langle BH\rangle$ and λp of each LCZ built type, data from Ref. [5].

LCZ Built Type	$\mathbf{\langle }\mathit{B}\mathit{H}\mathbf{\rangle } \mathbf{\left[}\mathbf{m}\mathbf{\right]}$	λp
LCZ 1 (Compact high-rise)	>25	0.4–0.6
LCZ 2 (Compact Mid-rise)	10–25	0.4–0.7
LCZ 3 (Compact Low-rise)	3–10	0.4–0.7
LCZ 4 (Open High-rise)	>25	0.2–0.4
LCZ 5 (Open Mid-rise)	10–25	0.2–0.4
LCZ 6 (Open Low-rise)	3–10	0.2–0.4
LCZ 7 (Lightweight low-rise)	2–4	0.6–0.9
LCZ 8 (Large Low-rise)	3–10	0.3–0.5
LCZ 9 (Sparsely Built)	3–10	0.1–0.2
LCZ 10 (Heavy industry)	5–15	0.2–0.3

. The calculation methods of $\langle BH\rangle$ and λp are mentioned in Section 3.2.3 and Section 3.2.4.

As shown in Table 2, the recommended ranges of λp and $\langle BH\rangle$ for compact high-rise type is from 0.4 to 0.6 and more than 25 m, respectively. However, as shown in Figure 2, there is no area (1 km × 1 km) in Tokyo and Shanghai that matches these ranges, so the compact high-rise type was not defined in this study for both cities. Similarly, the recommended range of λp for the lightweight low-rise type is from 0.6 to 0.9. However, the maximum λp in Tokyo and Shanghai is only 0.5. Moreover, according to previous studies [22,97], the lightweight low-rise type only exists in developing cities. Therefore, the lightweight low-rise type was not defined for the developed Tokyo and Shanghai in this study. Heavy industry type had some overlapped ranges with open mid-rise type and open low-rise type. Through GEP, it was found that the heavy industry type exists only in the areas neighboring Tokyo, such as Kawasaki, Yokohama, and Chiba, but not in Tokyo itself, which is consistent with previous studies [86,98]; hence, heavy industry type was not defined in Tokyo in this study.

LCZ land cover types were defined through the description of each LCZ land cover type (Figure A1) and satellite images in GEP. Bush & scrub type (LCZ C) and bare soil or sand type (LCZ F) were not found in Tokyo, consistent with the previous study [86,98], so they were not defined in Tokyo. Dense trees type (LCZ A) and shrubs & bushes type (LCZ C) were located outside the administrative boundaries of Shanghai [99], so they were not defined in Shanghai.

Excluding these LCZ types mentioned above, 12 LCZ types were finally defined in Tokyo, including 7 LCZ built types and 5 LCZ land cover types. On the other hand, 13 LCZ types, including 8 LCZ built types and 5 LCZ land cover types, were defined in Shanghai. According to the sample selection criteria of WUDAPT [96], 5 to 20 training samples were selected for each defined LCZ type in each city through GEP. Examples of the training samples of each LCZ built type in Tokyo and Shanghai are illustrated in Figure 3a,b, respectively.

Subsequently, the satellite images and training samples of each city were imported into SAGA-GIS as input data, and default 100 m resolution LCZ maps of each city were generated using the random forest algorithm in SAGA-GIS. The accuracy of each LCZ map was verified using a confusion matrix [61].

3.2. Calculating Land Cover and Urban Morphological Properties of LCZ Built Types

The land cover properties of the LCZ built type include impervious surface fraction (ISF), PSF, surface admittance, and SA, and the urban morphological properties include $\langle BH\rangle$, λp, SVF, and aspect ratio (AR) [5].

PSF and ISF are two complementary properties [100], and both were used in LST-related studies [64,66]. Compared to ISF, in addition to being used in the formulation of LST mitigation strategies, PSF also plays an important role in enhancing urban resilience, such as flood control and stormwater management, ecosystem services, and improving air quality [57]; therefore, PSF was selected in this study. Surface admittance was not selected because it is difficult to estimate at the local scale [5]. AR was not selected because the street geometry is complex and there is no standardized method for calculating AS [22].

$\langle BH\rangle$ and λp are the most important properties for defining LCZ built types [22], and they were also widely used to formulate LST mitigation strategies [30]. SVF is usually calculated with GIS data and is thus highly relevant with the $\langle BH\rangle$ and λp [2]; compared with SVF, $\langle BH\rangle$ and λp can be used in the analysis of LST more directly [2], so $\langle BH\rangle$ and λp were selected. Furthermore, $\langle BH\rangle$ and λp were also important indicators of resilient urban forms [57,101]. For example, high $\langle BH\rangle$ and λp can not only provide a shading effect to reduce urban temperature, but also improve the overall efficiency of the urban energy system by benefiting from economies of scale, thereby helping to reduce carbon emissions. However, high $\langle BH\rangle$ and λp may disrupt natural ventilation patterns, increase humidity levels, and reduce passive cooling through evaporation and transpiration. Therefore, a previous study [57] suggested that when using $\langle BH\rangle$ and λp to analyze urban heat-related problems, SA and PSF must also be considered to propose reasonable mitigation strategies. Therefore, SA was also selected in this study. Overall, selecting PSF, SA, $\langle BH\rangle$ and λp for LST analysis is not only conductive to formulating LST mitigation strategies, but can also improve urban resilience.

The PSF and SA were calculated using multispectral satellite images from the Landsat 8 Level 1 data product, as listed in Table 1. $\langle BH\rangle$ and λp were calculated using GIS data. The GIS data for Tokyo were the 2015 GIS data collated by the Environmental Systems Research Institute Japan Corporation [102]; the GIS data for Shanghai were the 2018 GIS data collated by the Resource and Environmental Science Data Center [103]. Each property was calculated for each zone of the LCZ map at a resolution of 100 m.

3.2.1. Pervious Surface Fraction (PSF)

Since there are always few water bodies in the built type zones [73], the PSF can be estimated based on the assumption that the built type zones are only comprised of vegetation cover and impervious surfaces (buildings and roads) [22]. It represents the total proportion of all vegetation (street trees, vegetation in parking or vacant lots, plants in yards) in the 100 m × 100 m grid in this study. This value can be calculated using the following equations:

$$NDVI=\frac{{\alpha }_5-{\alpha }_4}{{\alpha }_5+{\alpha }_4}$$

(1)

$$PSF={\left(\frac{NDVI-NDV{I}_{min}}{NDV{I}_{max}-NDV{I}_{min}}\right)}^2$$

(2)

where $NDVI$ is the normalized difference vegetation index, ${\alpha }_4$ and ${\alpha }_5$ are the top of atmosphere reflectance (TOA) values of band 4 (red) and band 5 (near-infrared) of Landsat 8, respectively, $NDV{I}_{min}$ and $NDV{I}_{max}$ are the minimum and maximum $NDVI$ values of each city, respectively. In Tokyo, the $NDV{I}_{min}$ and $NDV{I}_{max}$ were 0.01 and 0.94, respectively; in Shanghai, the $NDV{I}_{min}$ and $NDV{I}_{max}$ were 0.01 and 0.98, respectively.

3.2.2. Surface Albedo (SA)

SA represents the ratio of the amount of reflected solar radiation to the amount of solar radiation incident on the surface. This value can be calculated using the following equation [104,105]:

$$A=0.356{\alpha }_2+0.130{\alpha }_4+0.373{\alpha }_5+0.085{\alpha }_6+0.072{\alpha }_7-0.0018$$

(3)

where ${\alpha }_i$ is the TOA value of band $i$ of Landsat 8.

3.2.3. Average Building Height ($\langle BH\rangle$)

The $\langle BH\rangle$ [m] represents the average building height in each zone (100 m × 100 m) (in Section 3.1, the size was 1 km × 1 km) of the LCZ map. This value can be calculated using the following equation [2]:

$$\langle BH\rangle =\frac{{{\displaystyle \sum }}_{i=1}^n(B{H}_i\times B{A}_i)}{{{\displaystyle \sum }}_{i=1}^{n}BAi}$$

(4)

where $B{H}_i$ is the height of each building, $B{A}_i$ is the footprint area of the building, and $n$ is the total number of buildings in the zone.

3.2.4. Gross Building Coverage Ratio (λp)

The gross building coverage ratio (λp) represents the footprint area of all buildings relative to the zone area (in Section 3.1, the size was 1 km × 1 km). This value can be calculated using the following equation:

$${\mathsf{\lambda }}_p=\frac{{{\displaystyle \sum }}_{i=1}^{n}B{A}_i}{GA}$$

(5)

where $GA$ is the area of one zone.

3.3. Retrieving Land Surface Temperature (LST) of Summer in Tokyo and Shanghai

The most widely used LST data were from the Moderate Resolution Imaging Spectroradiometer (MODIS), Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER), and Landsat [106]. The spatial resolution of MODIS LST was 1 km, which was too coarse for the 100 m resolution LCZ maps used in this study. Although ASTER LST had a spatial resolution of 90 m, which was close to the resolution of the LCZ maps in this study, the scene size of ASTER LST was only 60 km × 60 km, which made it difficult to fully cover Tokyo and Shanghai. The Landsat 8 thermal infrared sensor had a spatial resolution of 100 m and a scene size of 185 km × 180 km, which was suitable for this study. Therefore, Landsat 8 was selected to retrieve LST. Landsat 8 was in a sun-synchronous orbit; therefore, only diurnal LST data were available.

Summer (July–September) was selected as the target season because the conditions in the thermal environment of urban areas are most severe during this period. Considering that the shaded area on the land surface varies greatly at different sun elevations and azimuths, it was difficult to draw general conclusions about the influence of land cover and urban morphological properties on LST using the LST of a specific day. Therefore, 5 sets of satellite images from Tokyo and 16 sets from Shanghai were selected in this study. Landsat 8 was launched in 2013; hence, the satellite images selected to retrieve LST were all summer (July–September) satellite images of Tokyo and Shanghai with less than 10% cloud coverage from 2013 to 2022. The sun elevation of the satellite images in Tokyo ranged from 48.93° to 66.23°, and the sun azimuth ranged from 119.46° to 150.88°. The sun elevation of the satellite images in Shanghai ranged from 59.97° to 67.56°, and the sun azimuth ranged from 110.25° to 133.56°.

In this study, diurnal $LST$ (°C) was calculated using the single-channel algorithm (SCA) [107,108], which can be calculated using the following equation:

$$LST=\mathsf{\gamma }\left[\frac{{\psi }_1{L}_{\mathsf{\lambda }}+{\psi }_2}{\epsilon }+{\psi }_3\right]+\delta -273.15$$

(6)

where $\epsilon$ is the land surface emissivity [109]; $\mathsf{\gamma }$ and $\delta$ can be calculated using Equations (7)–(9):

$$\mathsf{\gamma }\approx \frac{T^2}{b_{\mathsf{\gamma }}{L}_{\mathsf{\lambda }}}$$

(7)

$$\delta \approx T-\frac{T^2}{b_{\mathsf{\gamma }}}$$

(8)

$$T=\frac{K_2}{\ln \left(\frac{K_1}{L_{\mathsf{\lambda }}}+1\right)}$$

(9)

where $T$ is the at-sensor brightness temperature; $b_{\mathsf{\gamma }}$, $K_1$, and $K_2$ are constants; for band 10 of Landsat 8, $b_{\mathsf{\gamma }}$ is 1320 [K], $K_1$ is 774.8853 [K], and $K_2$ is 1321.0789 [K]; $L_{\mathsf{\lambda }}$ is the at-sensor radiance of band 10 of Landsat 8, which can be calculated using Equation (10):

$$L_{\mathsf{\lambda }}=M_L\times Q_{CAL}+A_L$$

(10)

where $Q_{CAL}$ is the digital number (DN) of the satellite image, $M_L$ is the band-specific multiplicative rescaling factor, and $A_L$ is the band-specific additive rescaling factor. For band 10 of Landsat 8, $M_L$ was 0.0003342 and $A_L$ was 0.1.

${\psi }_1$, ${\psi }_2$, and ${\psi }_3$ are atmospheric functions that can be calculated using Equation (11):

$${\psi }_1=\frac{1}{\tau }; {\psi }_2=-L_{\mathsf{\lambda }}^{\downarrow }-\frac{L_{\mathsf{\lambda }}^{\uparrow }}{\tau }; {\psi }_3=L_{\mathsf{\lambda }}^{\downarrow }$$

(11)

where $\tau$ is the atmospheric transmittance, $L_{\mathsf{\lambda }}^{\uparrow }$ is the atmospheric path radiance, and $L_{\mathsf{\lambda }}^{\downarrow }$ is the sky radiance. All these values can be obtained using the Atmospheric Correction Parameter Calculator provided by the National Aeronautics and Space Administration [110].

For LST retrieval errors in zones with clouds [98], the quality assessment band of the satellite image set was used as a mask to remove clouds from each satellite image. Subsequently, the retrieved LST values were used to calculate the average LST for each zone.

3.4. Correlation Analysis of Properties and LST

Regression models can help to evaluate the relationship between two or more properties of a feature and have been widely used to investigate the influencing factor of LST [22,111,112,113]. Some previous studies have proved that there is a linear relationship between LST and $\langle BH\rangle$, PSF and other properties [22,114,115]. Therefore, this study used a linear regression model to investigate the relationship between land cover properties, urban morphological properties, and the LST of each LCZ built type for Tokyo and Shanghai. The coefficient of determination (R²) was used to measure the correlation between LST and these properties. Based on previous studies [19,36,40], a threshold R² value of 0.1 was set in this study. If R² was less than 0.1, there was no correlation; if R² was higher than 0.1, there was a correlation; the higher the R² value, the stronger the correlation. By comparing the value of R², the main factor influencing LST of each LCZ built type can be clarified.

In addition, a significance test was also performed, and p-values were used to test the reliability of the properties correlated with LST. Based on previous studies, a threshold p-value of 0.05 was set in this study. If the p-value was less than 0.05, the correlations were reliable.

4. Results

4.1. LCZ Distributions in Tokyo and Shanghai

The LCZ distributions for Tokyo and Shanghai are illustrated in Figure 4a,b, respectively. The confusion matrixes of Tokyo and Shanghai are illustrated in

Table A1. Confusion matrix for the LCZ classification of Tokyo, Japan.

CLASS		Reference												Sum Row	User Accuracy
		LCZ 2	LCZ 3	LCZ 4	LCZ 5	LCZ 6	LCZ 8	LCZ 9	LCZ A	LCZ B	LCZ D	LCZ E	LCZ G
Output	LCZ 2	154	19	2	0	2	1	0	0	0	0	0	0	195	0.79
	LCZ 3	9	340	0	1	9	12	6	0	1	0	11	0	389	0.87
	LCZ 4	0	0	95	0	0	0	0	0	0	0	0	0	101	0.94
	LCZ 5	1	0	9	99	6	13	3	0	3	0	0	0	135	0.73
	LCZ 6	0	11	1	1	254	6	7	0	4	0	0	0	284	0.89
	LCZ 8	5	2	2	3	2	260	5	0	2	0	6	0	290	0.9
	LCZ 9	0	0	0	9	21	1	334	0	32	6	2	0	405	0.82
	LCZ A	0	0	0	0	0	0	0	753	7	0	0	0	760	0.99
	LCZ B	0	0	1	2	0	1	10	32	509	0	1	0	556	0.92
	LCZ D	0	0	0	0	0	0	39	0	7	124	0	0	170	0.73
	LCZ E	0	0	1	0	4	2	0	0	0	0	122	0	130	0.94
	LCZ G	0	0	1	0	0	0	0	0	0	0	1	523	525	1
Sum Column		172	372	129	115	298	298	404	785	565	130	143	523
Producer Accuracy		0.90	0.91	0.74	0.86	0.85	0.87	0.83	0.96	0.90	0.95	0.85	1.00
Overall Accuracy		0.90
Kappa Coefficient		0.89

and

Table A2. Confusion matrix of the LCZ classification of Shanghai, China.

CLASS		Reference													Sum Row	User Accuracy
		LCZ 2	LCZ 3	LCZ 4	LCZ 5	LCZ 6	LCZ 8	LCZ 9	LCZ 10	LCZ B	LCZ D	LCZ E	LCZ F	LCZ G
Output	LCZ 2	30	0	2	0	0	0	2	0	0	0	0	0	0	34	0.88
	LCZ 3	8	40	0	0	0	0	0	0	0	0	0	0	0	48	0.83
	LCZ 4	0	1	221	2	0	0	0	2	3	0	0	31	0	260	0.85
	LCZ 5	0	17	85	274	9	1	0	0	1	0	0	2	0	389	0.7
	LCZ 6	0	3	55	10	184	1	1	1	3	4	0	12	0	274	0.67
	LCZ 8	0	0	1	12	1	439	0	3	0	1	1	0	0	458	0.96
	LCZ 9	0	0	5	0	15	0	269	1	17	216	0	41	1	565	0.48
	LCZ 10	0	0	2	0	0	48	0	36	0	0	2	0	0	88	0.41
	LCZ B	0	0	4	0	8	1	0	1	375	2	0	75	0	466	0.8
	LCZ D	0	0	0	0	0	0	52	1	12	3480	0	74	102	3721	0.94
	LCZ E	0	0	0	0	2	2	0	6	0	0	128	0	0	138	0.93
	LCZ F	0	0	0	0	0	0	0	0	0	32	0	361	0	393	0.92
	LCZ G	0	0	0	0	0	0	0	0	0	15	0	68	5339	5422	0.98
Sum Column		38	61	375	298	219	492	324	51	411	3750	131	664	5442
Producer Accuracy		0.79	0.66	0.59	0.92	0.84	0.89	0.83	0.71	0.91	0.93	0.98	0.54	0.98
Overall Accuracy		0.90
Kappa Coefficient		0.88

, respectively. The overall accuracy and kappa values of Tokyo were 90.3% and 89.2%, respectively, and those of Shanghai were 90% and 88.4%, respectively. The accuracy of these results was close to or higher than that of the LCZ maps of Japanese or Chinese cities in previous studies [22,98,116,117], confirming that the LCZ maps in this study were satisfactory.

As illustrated in Figure 4a, the main land cover in areas of Tokyo where there are no buildings is dense trees. There was a clear dividing line (solid white line in Figure 4a) between the dense tree type zones and built type zones in LCZs (hereafter, built type zones) near longitude 139°15′. The dense tree type zones were mainly located on the western side of the dividing line, whereas the built type zones were mainly located on the eastern side. On the eastern side of the dividing line, the built type zones were distributed as three approximately concentric circles (dotted white circles in Figure 4a). The open high-rise type zones were concentrated in the central ring (Tokyo Station), the compact mid-rise type zones were in the middle ring, and the compact low-rise type zones were in the outer ring, extending into the suburban area. The open mid-rise type and open low-rise type zones were mainly located on the southern side of Tokyo.

As illustrated in Figure 4b, low plants are the main land cover type in Shanghai, where there are no buildings. The urban area of Shanghai was dominated by open built type zones (open high-rise, open mid-rise, and open low-rise type zones). Compact built type zones (compact mid-rise and compact low-rise type zones) were concentrated in a small region on the west side of the Huangpu River near the Shanghai Railway Station, and there were more built type zones on the west side of the Huangpu River than on the east side. Large low-rise type zones were mainly distributed on the western side of Shanghai. Sparsely built type zones were mainly distributed on Chongming Island and the eastern side of Shanghai.

Table 3. The proportion of each LCZ to LCZ built types in Tokyo and Shanghai.

LCZ Built Type	Tokyo		Shanghai
	Number *	Proportion	Number *	Proportion
LCZ 2 (Compact mid-rise)	11,590	11%	1087	0.28%
LCZ 3 (Compact low-rise)	32,188	30.6%	2552	0.67%
LCZ 4 (Open high-rise)	5666	5.39%	23,836	6.21%
LCZ 5 (Open mid-rise)	10,125	9.63%	64,641	16.83%
LCZ 6 (Open low-rise)	17,503	16.64%	79,314	20.65%
LCZ 8 (Large low-rise)	8823	8.39%	88,089	22.94%
LCZ 9 (Sparsely built)	19,298	18.35%	106,125	27.63%
LCZ10 (Heavy industry)	-	-	18,395	4.79%
Total	105,193	100%	384,039	100%

* Number means the total number of zones of 100 m resolution.

illustrates the proportion of each LCZ built type among all those in Tokyo and Shanghai. The proportion of compact built type (compact low-rise and compact mid-rise type) zones differed significantly between Tokyo and Shanghai, especially for compact low-rise type zones. In Tokyo, the proportion of compact low-rise type zones was 30.6%, which was the largest proportion of LCZ built types. However, in Shanghai, the proportion was only 0.67%. The proportion of compact mid-rise type zones in Tokyo was 11%, whereas that in Shanghai was only 0.28%. The proportions of other built type zones in Tokyo were lower than those in Shanghai.

4.2. Maps of Land Cover and Urban Morphological Properties in Tokyo and Shanghai

Figure 5 and Figure 6 showed urban morphological and land cover property maps, respectively. The distribution characteristics of these properties were consistent with their LCZ distribution in both Tokyo and Shanghai.

In the urban area of Tokyo, the distribution of $\langle BH\rangle$ and λp also demonstrated three concentric circles (dotted white circles in Figure 6(a-1,b-1), similar to the distribution of built type zones (Figure 4a). The center (Tokyo Station), which was dominated by open high-rise type zones, had the highest $\langle BH\rangle$ and high λp; the middle ring, which was dominated by compact mid-rise type zones, had reduced $\langle BH\rangle$ and high λp; and the outer ring, which was dominated by compact low-rise type zones, had relatively high λp and the lowest $\langle BH\rangle$.

In Shanghai, the proportion of areas with $\langle BH\rangle$ higher than 15 m was much larger than that of Tokyo, and the proportion of areas with high λp was much lower than that of Tokyo. In Shanghai, the areas with low $\langle BH\rangle$ and high λp were concentrated on the west side of the Huangpu River near the Shanghai Railway Station. This was also the area where compact low-rise zones were concentrated.

The distribution of PSF and SA (Figure 6) clearly distinguished built type zones and land cover type zones in both Tokyo and Shanghai. The PSF and SA of areas dominated by land cover type zones were significantly higher than those dominated by built type zones. The PSF of the urban areas in Tokyo was relatively low, and the PSF of the urban areas in Shanghai was higher than that of Tokyo. In addition, whether in Tokyo or Shanghai, the differences in SA of different urban areas were small, ranging from 0.1 to 0.17.

In addition, based on the above data, the average values and ranges of PSF, SA, $\langle BH\rangle$, and λp of each LCZ built type in Tokyo and Shanghai were also determined, as shown in Figure A2. In Figure A2, the value at the left end of the range of Tokyo and Shanghai is referred to as (mean value − standard deviation), and the value at the right end of the range is referred to as (mean value + standard deviation).

4.3. LST Distributions in Tokyo and Shanghai

The LST distributions of Tokyo and Shanghai were illustrated in Figure 7a,b, respectively. The LST distributions of these two cities were correlated with their LCZ distributions (Figure 4a,b).

In Tokyo, there was also a clear dividing line (solid black line in Figure 7a) between the high and low LST areas near longitude 139°15′. The eastern part of the dividing line, dominated by built type zones, demonstrated a higher LST than the western side. The LST distribution of the urban area in Tokyo also demonstrated three concentric circles (dotted black circles in Figure 7a), similar to the distribution of built type zones (Figure 4a). The center (Tokyo Station), dominated by open high-rise type zones, had a low LST, the middle ring, which was dominated by compact mid-rise type zones, had an increased LST, and the outer ring, dominated by compact low-rise type zones, had the highest LST.

In Shanghai, areas dominated by built type zones also demonstrated a higher LST than areas dominated by land cover type zones. Areas dominated by compact low-rise type zones and large low-rise type zones demonstrated the highest LST. The west side of the Huangpu River demonstrated more high-LST areas than the east side because there were more built type zones on the west side. The areas dominated by open high-rise type zones demonstrated relatively low LST. The LST on the east side of Shanghai and Chongming Island, dominated by sparsely built type zones, was low.

4.4. Average LST of Each LCZ Built Type in Tokyo and Shanghai

In this study, we also calculated the average diurnal LST to investigate the differences in LST between the LCZ built types in each city (Figure 8). The average LST of each LCZ built type was the average value of all zones of the same LCZ built type in each city.

The compact low-rise type had the highest average LST in both cities, as it had the lowest average PSF and average $\langle BH\rangle$ (Figure A2a,c). The open high-rise type had a low average LST in both cities, reflecting the better solar radiation-blocking effect of high-rise buildings. The open mid-rise and sparsely built types in Tokyo and the open low-rise and sparsely built types in Shanghai, which have a high average PSF (Figure A2a), had relatively low average LST.

4.5. Correlations between Land Cover and Urban Morphological Properties and LST in Tokyo and Shanghai

Table 4. Correlations between properties and LST of each LCZ built type in Tokyo and Shanghai.

LCZ Built Type	R2 between Properties and LST of Each LCZ Built Type in Tokyo and Shanghai
	Tokyo				Shanghai
	PSF	SA	$\mathbf{\langle }\mathit{B}\mathit{H}\mathbf{\rangle }$	λp	PSF	SA	$\mathbf{\langle }\mathit{B}\mathit{H}\mathbf{\rangle }$	λp
LCZ 2 (Compact mid-rise)	0.014	0.001	0.154	0.003	0.012	0.001	0.112	0.020
LCZ 3 (Compact low-rise)	0.122	0.019	0.095	0.050	0.162	0.099	0.003	0.138
LCZ 4 (Open high-rise)	0.022	0.029	0.227	0.002	0.292	0.01	0.029	0.155
LCZ 5 (Open mid-rise)	0.336	0.014	0.003	0.147	0.336	0.016	0.050	0.123
LCZ 6 (Open low-rise)	0.364	0.033	0.014	0.153	0.342	0.000	0.001	0.078
LCZ 8 (Large low-rise)	0.102	0.005	0.077	0.039	0.282	0.009	0.001	0.203
LCZ 9 (Sparsely built)	0.502	0.905	0.000	0.151	0.257	0.010	0.015	0.056
LCZ 10 (Heavy industry)					0.115	0.038	0.000	0.090
No correlation			Negative correlation			Positive correlation

Properties correlated with LST have been tested by p-value, with their p-values all being less than 0.05.

illustrates the correlations between properties and LST in Tokyo and Shanghai. For compact mid-rise type, $\langle BH\rangle$ demonstrated a negative correlation with LST in both cities. However, for open high-rise type, there was a correlation between $\langle BH\rangle$ and LST in Tokyo, but no correlation in Shanghai. Through satellite images and street views, we found that open high-rise zones in Tokyo were mainly dominated by high-rise commercial buildings, while in Shanghai, in addition to such open high-rise type zones, many of these zones were dominated by high-rise residential buildings. The building spacing, building arrangement, PSF, and pavement of these two types of open high-rise type zones were completely different. This led to the conclusion that even if these two types of open high-rise type zones have the same $\langle BH\rangle$, their LST were completely different; thus, the correlation between $\langle BH\rangle$ and LST of open high-rise type in Shanghai was significantly weakened.

For the other LCZ built types in Tokyo and Shanghai, the PSF showed the highest R² with LST; it was the main factor influencing LST for these LCZ built types. Meanwhile, PSF was negatively correlated with LST, reflecting the cooling effect of plant transpiration on LST.

LCZ built type zones with higher λp had more building shadows and thus, lower LST; hence, λp was expected to demonstrate a negative correlation with LST. However, λp was positively correlated with LST for the open mid-rise, open low-rise, and sparsely built types in Tokyo and the compact low-rise, open high-rise, open mid-rise, and large low-rise types in Shanghai. Combined with the fact that the PSF of these LCZ built types demonstrated relatively strong negative correlations with LST, one of the possible reasons for the positive correlations between λp and LST was that, for these LCZ built types, zones with high λp may have low PSF, and the PSF of these LCZ built types had relatively strong negative correlations with LST; thus, λp exhibited positive correlations with LST.

Table 5. Correlations between PSF and λp of some LCZ built types in Tokyo and Shanghai.

Study Area	LCZ Built Types	R2 between PSF and λp
Tokyo	LCZ 5 (Open Mid-rise)	0.2772
	LCZ 6 (Open Low-rise)	0.1947
	LCZ 9 (Sparsely Built)	0.2419
Shanghai	LCZ 3 (Compact Low-rise)	0.1078
	LCZ 4 (Open High-rise)	0.1035
	LCZ 5 (Open Mid-rise)	0.1306
	LCZ 8 (Large Low-rise)	0.1411
No correlation	Negative correlation	Positive correlation

The correlations in this table have been tested by p-value, with their p-values all being less than 0.05.

illustrates the correlations between PSF and λp for these LCZ built types, and it can be observed that λp was negatively correlated with PSF for these LCZ built types, as expected.

There was no correlation between SA and LST for each LCZ built type. Because the SA of each LCZ built type was within a narrow range (Figure A2b), the differences in SA between zones of the same LCZ built type were too small to demonstrate the correlations between SA and LST. Properties correlated with LST in Table 4 and the correlations in Table 5 have been tested by p-value, with their p-values all being less than 0.05, which proved that the above correlations were reliable.

5. Discussion

5.1. Comparison of Main Factors Influencing LST of Different LCZ Built Types between Tokyo, Shanghai, and Large Cities

Few studies have investigated the main factors influencing LST of different LCZ built types, and they have only investigated the influence of a single type of properties (only urban morphological or land cover properties) on the LST of different LCZ built types. It should also be noted that their study areas were all large cities with smaller scales than megacities.

Yang [30] and Zhou [73] investigated the influence of several urban morphological properties on the LST of different LCZ built types in Shenyang and Xi’an, China, respectively. Their results showed that, unlike Tokyo and Shanghai, the main factor influencing the LST of compact mid-rise types in Shenyang and Xi’an was not $\langle BH\rangle$. This may be because in a large city like Shenyang and Xi’an, the $\langle BH\rangle$ of compact mid-rise type zones was mainly distributed in a narrow range (the maximum difference was only around 10 m); hence, the differences in $\langle BH\rangle$ between different compact mid-rise type zones were too small, so the $\langle BH\rangle$ did not show a strong correlation with LST. In contrast, in megacities like Tokyo and Shanghai, the $\langle BH\rangle$ of compact mid-rise zones was distributed in a wider range (the maximum difference exceeded 20 m).

In addition, similar to Shanghai, the main factor influencing the LST of open high-rise types in Shenyang and Xi’an was not $\langle BH\rangle$. This may be because, as mentioned in Section 4.5, the open high-rise type zones of megacities and large cities in China often contain different building types, which weakened the relationship between $\langle BH\rangle$ and LST. Therefore, when generating LCZ maps for Chinese cities, it may be necessary to subclassify open high-rise type zones according to the type of buildings, to analyze their characteristics in more detail.

Although the two previous studies mentioned above did not investigate the influence of land cover properties on LST, the main factors influencing the LST of most LCZ built types in Shenyang and Xi’an were impervious surface fraction (ISF) or λp. ISF was a property that strongly related to PSF, and as shown in this study, λp also tended to have a strong correlation with PSF, so it can be speculated that PSF may also have a strong influence on the LST of most LCZ built types in large cities, such as Shenyang and Xi’an.

Mushore [74] investigated the influence of several land cover properties on the LST of mid-rise and low-rise built types in Harare, a large city in Zimbabwe. His results showed that although there were differences in the main factors influencing LST for different LCZ built types, the main influencing factors were all PSF-related properties, such as the NDVI and Normalized Difference Bareness Index. His results were consistent with this study, emphasizing the important influence of PSF on the LST of most LCZ built types.

5.2. LST Mitigation Strategies for LCZ Built Types in Tokyo and Shanghai

5.2.1. Specific Strategies for Increasing Vegetation of Compact Low-Rise Type Zones and Open Low-Rise Type Zones in Tokyo

Figure 7a illustrates the LST distribution of Tokyo. It can be seen that the LST in the urban area of Tokyo was distributed in three concentric circles, and the outer ring that was dominated by compact low-rise type zones demonstrated the highest LST and extended to suburban areas. Figure 7a illustrates the average LST of each LCZ built type in Tokyo. In addition to confirming that the compact low-rise type had the highest average LST, it was also found that the open low-rise type had the second-highest average LST. Their zones accounted for 47.24% of all built type zones in Tokyo (Table 3), which means that nearly half of the built type zones in Tokyo belong to the LCZ built types with high LST. The results of Section 4.5 demonstrate that the PSF was the main factor influencing LST of compact low-rise and open low-rise types and had negative correlations with LST in Tokyo. This meant that increasing the PSF of these LCZ built types could be an effective method to reduce their LST. In the compact low-rise type zones of Tokyo, there were mostly dense low-rise residential buildings with few green plants in the streets between buildings. Therefore, adding street trees could be a simple, effective method to reduce the LST of compact low-rise type zones, and this method was also available for compact low-rise type zones in other Japanese cities [22]. The buildings of open low-rise type zones were almost all low-rise residential buildings, but there were already many street trees between these buildings; it may not be suitable to increase the PSF by adding street trees. Compared with buildings in compact low-rise type zones, buildings in open low-rise type zones tend to have large yards; encouraging residents to plant in their yards could be an effective method to reduce LST in open low-rise type zones.

5.2.2. Improving the Conditions in the Thermal Environment of Compact Low-Rise Type Zones through the Urban Renewal Plan in Shanghai

In Shanghai, the compact low-rise type had a significantly higher average LST (Figure 8b) than other built types. Although these zones accounted for only 0.67% of all built type zones in Shanghai (Table 3), they were concentrated in the Hongkou, Jing’an, and Huangpu Districts, the three most densely populated districts in Shanghai (Figure 7b). In 2020, the population densities of the Hongkou, Jing’an, and Huangpu Districts reached 32,317, 26,516, and 32,185 persons/km², respectively [118], which means that there were many residents in the compact low-rise type zones in Shanghai, who were likely to be suffering from high LST environments. The results in Section 4.5 demonstrate that the PSF was the main factor influencing LST of the compact low-rise type, which negatively correlated with LST. Increasing the PSF of compact low-rise type zones should be an effective method to reduce LST. However, according to the observation of satellite images, we found that the distance between the buildings in Shanghai’s compact low-rise type zones was too narrow to add plants.

In contrast, most of the buildings in Shanghai’s compact low-rise type zones were dilapidated old houses with poor living environments. To improve the living environment, the Shanghai government has been promoting the renewal, renovation, or reconstruction of the areas where these old dilapidated houses gathered. In 2021, Shanghai completed the renewal of 901,000 m² of houses [119]. Referring to the areas that had been renewed, those where compact low-rise zones were located were likely to be renewed to open high-rise type zones, open mid-rise type zones, or open low-rise type zones in the future. Section 4.4 demonstrate that the average LST of the open high-rise type, open mid-rise type, and open low-rise types were significantly lower than that of the compact low-rise type in Shanghai; therefore, the LST of these areas was likely to be reduced through urban renewal.

5.2.3. Increasing 〈BH〉 to Reduce LST in Future Urban Development

Although the $\langle BH\rangle$ was the main factor influencing LST of compact mid-rise type and open high-rise type in Tokyo and compact mid-rise type in Shanghai, it had negative correlations with LST, and adjusting the $\langle BH\rangle$ of the build-up areas was generally considered difficult. The significance of these results was to provide urban planners and policymakers with information that $\langle BH\rangle$ can be appropriately increased to create shadows and thus control LST when building these built type zones in the future. However, increasing the $\langle BH\rangle$ may cause negative effects, such as decreasing the sunlight hours of surrounding buildings [120]. Therefore, the multiple influences of increasing $\langle BH\rangle$ must be fully considered to create a comfortable living environment.

5.2.4. Specific Strategies for Increasing Vegetation of Other LCZ Built Types

For other LCZ built types with negative correlations between PSF and LST in Tokyo and Shanghai, their zones tend to have wide streets and bare parking or vacant lots; therefore, increasing street trees and vegetation in parking or vacant lots could be an effective method of reducing LST.

5.2.5. Research Highlights and Limitation

The suggested LST mitigation strategies of each LCZ built type in Tokyo and Shanghai are outlined in

Table 6. Specific LST mitigation strategies for each LCZ built type in Tokyo and Shanghai.

Study Area	LCZ Built Types	Specific LST Mitigation Strategies
Tokyo	LCZ 2 (Compact Mid-rise)	Increasing building height to create urban shadow.
	LCZ 3 (Compact Low-rise)	Increasing street trees.
	LCZ 4 (Open High-rise)	Increasing building height to create urban shadow.
	LCZ 5 (Open Mid-rise)	Increasing street trees; increasing vegetation in parking or vacant lots.
	LCZ 6 (Open Low-rise)	Encouraging residents to plant in their yards.
	LCZ 8 (Large Low-rise)	Increasing street trees; increasing vegetation in parking or vacant lots.
	LCZ 9 (Sparsely Built)	Increasing street trees; increasing vegetation in parking or vacant lots.
Shanghai	LCZ 2 (Compact Mid-rise)	Increasing building height to create urban shadow.
	LCZ 3 (Compact Low-rise)	Through urban renewal plan of Shanghai.
	LCZ 4 (Open High-rise)	Increasing street trees; increasing vegetation in parking or vacant lots.
	LCZ 5 (Open Mid-rise)	Increasing street trees; increasing vegetation in parking or vacant lots.
	LCZ 6 (Open Low-rise)	Increasing street trees; increasing vegetation in parking or vacant lots.
	LCZ 8 (Large Low-rise)	Increasing street trees; increasing vegetation in parking or vacant lots.
	LCZ 9 (Sparsely Built)	Increasing street trees; increasing vegetation in parking or vacant lots.
	LCZ 10 (Heavy Industry)	Increasing street trees; increasing vegetation in parking or vacant lots.

. It was apparent that the LST mitigation strategies differ among different LCZ built types. The findings suggest that effective LST mitigation strategies can be developed by integrating the characteristics of each LCZ built type and their main LST influencing factors. This approach could provide more perspectives for urban planners and policymakers to develop highly feasible and reasonable LST mitigation strategies.

However, in complex urban areas, LST is influenced by multiple factors [31,32]. This study only investigated the influence of $\langle BH\rangle$, λp, PSF, and SA on the LST of each LCZ built type, which might not be comprehensive. In future studies, it is necessary to explore the influence of more properties on the LST of each LCZ built type. This would help in developing more comprehensive and effective LST mitigation strategies.

6. Conclusions

Identifying the main factors influencing LST of different LCZ built types is of crucial importance for formulating specific and effective LST mitigation strategies for different LCZ built types. This study clarified the main factors influencing the LST of different LCZ built types in Tokyo and Shanghai by comprehensively investigating the influence of the land cover and urban morphological properties of each LCZ built type on LST in each city. The results confirmed that for the compact mid-rise and open high-rise types in Tokyo, and compact mid-rise type in Shanghai, $\langle BH\rangle$ was the main factor influencing LST and it had a negative correlation with LST; for other LCZ built types, PSF was the main factor influencing LST and also had a negative correlation with it. This study integrates the above results with the characteristics of LCZ built types in Tokyo and Shanghai to propose different LST mitigation strategies for different LCZ built types. Suggested LST mitigating strategies include increasing street trees for the compact low-rise type in Tokyo, planting more plants in the building yards for the open low-rise type in Tokyo, improving conditions in the thermal environment of the compact low-rise type in Shanghai through urban renewal plans, and considering the appropriate increase in building height for compact mid-rise types in future urban development in both Shanghai and Tokyo.

This study also has possible limitations. First, the WUDAPT L0 method is the most widely used LCZ mapping method [121], and it was used to generated the LCZ maps of Tokyo and Shanghai in this study. As Asian megacities, Tokyo and Shanghai have been proven to have a compact high-rise type [75,98,100]. However, as mentioned in Section 3.1 of this study, it is difficult to define the compact high-rise type in Tokyo and Shanghai using the WUDAPT L0 method. Since the compact high-rise type often has a high population density and severe conditions in the thermal environment [122], it is important to identify and subsequently analyze it in future studies. To better identify the compact high-rise type, future studies should consider using higher resolution data such as airborne LiDAR data and higher resolution satellite images [100,123]; employing other classification algorithms like fuzzy logic classification, convolutional neural networks, or deep neural networks [100,121]; or applying the GIS-based LCZ mapping method [75,124]. Second, this study only investigated megacities in the Cfa climate, and it is necessary to investigate megacities with diverse climates in future studies, to deepen our understanding of the main factors influencing LST in different LCZ built types under different climate conditions. Third, in the future, it is necessary to investigate the influence of more properties on the LST of different LCZ built types, to formulate more comprehensive LST mitigation strategies. Finally, this study focuses only on LST mitigation strategies, but adaptation strategies also play a pivotal role in reducing heat-related issues [4]. Therefore, future studies should also emphasize the development and formulation of adaptation strategies to construct more resilient urban spaces.