Evaluation of Urban Thermal Comfort and Its Relationship with Land Use/Land Cover Change: A Case Study of Three Urban Agglomerations, China

Abstract

With the acceleration of urbanization in China, the urban surface thermal environment has undergone significant changes. This work aims to calculate the urban thermal comfort index using a temperature and humidity model with the land surface temperature and relative humidity. It also aims to explain the association between the land use/land cover change (LUCC) and urban surface thermal environment of the Beijing–Tianjin–Hebei (BTH) Region, the Guangdong–Hong Kong-Macao Greater Bay Area (GBA) and the Yangtze River Delta (YZD) in 2020, 2015, 2010 and 2005 using geographically weighted regression. The results reveal that (1) the three urban agglomerations have substantial heat island intensity regions, which are clustered and zonally distributed, and the annual average rates of the heat island area growth in the three regions are 1.01%, 1.41% and 1.09%, respectively. (2) Many uncomfortable areas exist in the three urban agglomerations, with an exponential growth trend in summer, and the annual average proportion of the uncomfortable areas in the three regions are 60.8%, 56.8% and 49.4%, respectively. (3) From the spatial point of view, the high-thermal comfort index areas of Beijing, Tianjin, Hebei, Guangdong, Hong Kong and Macao expand to the coast, and the high-index areas of the Yangtze River Delta expand to the inland. In terms of time, the annual distribution of thermal comfort in Beijing–Tianjin–Hebei is discrete, but the annual distribution of thermal comfort in Guangdong, Hong Kong, Macao and the Yangtze River Delta is concentrated. (4) In LUCC, the change intensity in construction land has a remarkable effect on the change in thermal comfort. The areas where the thermal comfort index increases positively correlate with the areas where the construction land increases. This study enriches the research on the impact of LUCC on urban ecological performance, and thus provides the necessary scientific basis for urban environment construction.

1. Introduction

With 68% of the world’s population expected to live in cities by 2050, rapid urbanization has profound implications for urban climates [1,2]. In the rapid urbanization process, given the large-scale change and expansion of the pattern of urban land use/land cover change (LUCC), the urban thermal environment has changed significantly [3,4,5]. To meet the needs of urban dwellers, more and more vegetation areas have been converted into artificial surfaces, leading to the urban heat island (UHI) effect [6]. The UHI effect describes the influence of urban surfaces on temperature patterns in urban areas as opposed to the surrounding areas [7]. Combined with rising average temperatures caused by global warming, many cities face the dual pressures of global warming and UHI [8,9]. The deteriorating urban thermal comfort (UTC) has caused many ecological and environmental problems, such as excessive water and energy consumption [10,11,12], and has also affected the urban environment and human health [13,14]. Therefore, additional knowledge of the influences of LUCC on the UHI and UTC provides a preliminary but practical basis for urban development strategies and human health and welfare.

There are already many studies on urban thermal environments and their effects on a single city. Extensive human land use activities lead to changes in the thermal properties of the land surface (e.g., surface roughness, evapotranspiration and albedo), which leads to the increase in the surface UHI intensity (SUHII) in urban areas [15]. The surface UHI (SUHI) is often used to explicitly distinguish the UHIs measured using land surface temperatures (LST) from the air temperature patterns [16]. The LST have the advantage of spatially explicit datasets compared to single measurement points. Remote sensing data is related to above-ground air temperatures [17], and a variety of remote sensing products are also available in time series. Thus, temperature data measured from satellite sensors were used to monitor LST changes and showed the SUHI effect [18,19,20,21]. As a supplementary explanation for the SUHI and urban thermal environment, there exists a difference and an inner linkage between the two aspects. The commonality of these two factors lies in the surface temperature and atmospheric temperature. The difference lies in that the former emphasizes the temperature difference between urban areas and suburban areas, while the measurement index of the latter is related to the temperature degree, building density, humidity, water body, green space distribution and other factors [22,23]. Since most of the current studies on the urban spatial thermal environment, and its climate change and environmental remediation at the global urban regional scale mainly focus on the SUHI effect, this paper mentions both the urban spatial thermal environment and SUHI. Unlike the SUHI, UTC is a person’s satisfaction with their thermal environment and is an indicator that quantifies how the ordinary person “feels”, based on the environmental conditions. Thus, the researchers have adopted the temperature and humidity index (THI) as an indicator of UTC, which quantifies thermal comfort based on the temperature and relative humidity data of meteorological stations [24,25]. However, these data are not widely used because of the limited number and location of observatories and the low accuracy of the interpolation algorithm.

The multi-time scale UTC and UHI research of urban agglomeration is frequently mentioned. In the previous study, the influencing factors were mostly limited to meteorological [26,27], economic [28,29] and statistical factors [30,31]. At the same time, the mechanism of urban thermal environment formation and urban underlying surface change is rarely mentioned, so the relationship between the LUCC and the urban thermal environments needed to be further studied. On the other hand, most of these studies put forward environment-oriented factors, which leads to the neglect of people’s feelings in the urban thermal environment. To enrich the void, this study aimed to establish an urban thermal environment indicator, oriented toward urban residents’ feelings, and assess the impact of LUCC on the urban thermal environment across the major urban agglomerations of China (the Beijing–Tianjin–Hebei Region, BTH; the Guangdong–Hong Kong–Macao Greater Bay Area, GBA; the Yangtze River Delta, YZD). In terms of methods, to describe the characteristics of the UTC and urban thermal environment comprehensively, the UTC index (UTCI) is introduced as a new indicator to measure the urban surface thermal environment in an urban agglomeration by adopting and transforming the THI model. The spatial continuity of LST and the spatial heterogeneity of high-temperature patches are comprehensively considered by analyzing the regional, temporal and spatial differences of the LST, which explain the UTC and UHI. Theoretically, based on the analysis of the urban agglomeration scale, this study is the first to explain the reasons and ways urban LUCC impacts UTC, which provides a strategy for planning urban sprawl and sustainable cities in urbanizing China.

2. Data and Methods

2.1. Overview of Study Area

Given the different scales of the urban agglomerations and single cities, there may present different spatiotemporal UTC and UHI patterns. In this study, the major urban agglomerations of China (BTH, GBA and YZD), which are situated in different climate zones, were chosen as the study areas (Figure 1). The study areas are ideal since urban agglomeration reflects the main feature of urban development in China, and the three urban agglomerations are influenced by extensive land use activities.

The BTH is characterized by a temperate, semi-humid climate with a mean annual temperature (MAT) of 4–13 °C and mean annual precipitation (MAP) of 300–800 mm. It includes Beijing, Tianjin and Shijiazhuang in Hebei Province. A northern subtropical humid climate characterizes the YZD in the southern part and a temperate humid climate in the counterpart, with a MAT of 10–19 °C and a MAP of 800–2000 mm, and it covers Shanghai, Jiangsu Province, Zhejiang Province and Anhui Province. The GBA is situated in the southern subtropical humid zone, with a MAP of 21–24 °C and a MAT of over 1500 mm, covering Hong Kong, Macao and Guangdong Province. In summer, the highest temperature of the three urban agglomerations can reach more than 40 °C [32].

The three urban agglomerations account for nearly half of the gross domestic product (GDP) and 29.7% of China’s total population, with approximately 5% of China’s area [33]. Moreover, these urban agglomerations’ rapid development and dramatic urbanization may potentially cause relevant problems.

2.2. Data

The data used in this study comprise LUCC, LST and relative humidity (RH) data. The land use data were obtained from the Aerospace Information Research Institute, Chinese Academy of Sciences, with a spatial resolution of 1 km.

The LST data were derived from MODIS eight-day composite products (version 6, MOD11A2), with a spatial resolution of 1 km. Firstly, the generalized split-window algorithm is used to estimate the MODIS LST from the images taken in the thermal bands 31 and 32. In addition, the acquisition of MODIS LST data is inevitably limited by atmospheric and surface conditions at observation time, since summer is also the rainy season in the study area. Thus, we randomly selected a specific day in June, July, August and September for each selected year, using the maximum value composite to synthesize these selected data in summer to obtain the maximum temperature in the selected years. This approach can help us to fill in all the data and reflect the high temperatures of summer every year. In this study, we chose 26 June, 20 July, 21 August and 14 September of 2005, 2010, 2015 and 2020 to obtain the annual summer maximum temperature after the maximum synthesis.

The RH data obtained from the National Earth System Science Data Center include the monthly mean relative humidity at a spatial resolution of 1 km. This study calculated the mean humidity from June to September of 2005, 2010, 2015 and 2020 to indicate the variability of humidity in the summers of the selected years.

2.3. Methods

This study used MODIS LST products, relative humidity (RH) data sets and land use data pertaining to 2020, 2015, 2010 and 2005 in BTH, GBA and YZD to evaluate the spatial and temporal variations of the LUCC, SUHI and UTC in the urban agglomerations. All the data from 2005 to 2020 are available. The large time span is chosen to study the change over a long time, and the interval of five years is sufficient to reflect the LUCC pattern. In addition, the large time span is conducive to reducing the relative error of the LUCC, which has a positive significance to the GWR accuracy. Based on the large scale of the study area, this span is the most complete and sufficient to reflect the phenomena and mechanisms of changes in the urban thermal environment during the rapid urbanization in China.

The framework of this study was as follows (Figure 2): First, to obtain the SUHII using the LUCC and LST data. Furthermore, to calculate the UTCI using the THI model, based on MODIS LST and RH data. Finally, the association between the UTCI and LUCC pattern was calculated using the geographically weighted regression (GWR) technique to assess the impacts of the LUCC on the UTC.

2.3.1. Modelling Surface Urban Heat Island Using SUHII

Remote sensing data are frequently used for the assessment of SUHI. The classical approach is to analyze urban/rural temperature differences. However, the differentiation between “urban” and “rural” remains unclear and confusing. In this study, to better model the impact of urban expansion and LUCC change in SUHII, we defined SUHII as the difference between the surface temperatures of built-up areas and green spaces. The SUHII formula of each grid unit is as in Equation (1):

$$SUHI{I}_i=T_{bui}-T_{gs},$$

(1)

where SUHII_i represents the SUHII of each grid unit in the built-up area; T_bui represents the surface temperature of the grid unit i in the built-up area; and T_gs is the average temperature of the green space grid unit.

$$SUHII=\frac{1}{n}{\displaystyle \sum _{i=1}^{n}SUHI{I}_i},$$

(2)

where SUHII_i represents the SUHII of the study area and n represents the total number of grid units in the built-up area. Following Lu’s classification standards [34], the SUHII is divided into five categories (

Table 1. SUHII classification standard [34].

SUHII	Levels
<0	No SUHII
0–2	Low SUHII
2–4	Moderate SUHII
4–6	High SUHII
>6	Extreme SUHII

).

2.3.2. Modelling Urban Thermal Comfort Using UTCI

The THI reflects the heat commute between the human body and the environment by temperature and humidity [35,36]. The formula is presented in Equation (3):

$$\mathrm{THI}=1.8t+32-0.55\left(1-f\right)\left(1.8t-26\right),$$

(3)

where THI is the temperature and humidity index, t is the temperature in Celsius degrees, and f is relative humidity (%).

In the classic approach, the THI originates from the temperature and relative humidity data from the meteorological stations. In this study, we use t_a obtained from the MODIS LST instead of t, and f_a obtained from the NESSDC China Humidity Dataset instead of f, to obtain the THI based on the remote sensing data and reanalysis data, which is the UTCI in this study. The formula is presented in Equation (4):

$$\mathrm{UTCI}=1.8t_a+32-0.55\left(1-f_a\right)\left(1.8t_a-26\right),$$

(4)

where UTCI is the urban thermal comfort index, t_a is the LST obtained from the MODIS products and f is relative humidity (%) obtained from the reanalysis data set.

According to the classification standard of the THI [37], combined with the distribution of the UTCI in this study, the human comfort evaluation system is formulated (

Table 2. THI classification Standard [37].

UTCI	Human Perception
<82	Very cool and very comfortable
82–87	Cool and comfortable
87–92	Hot and uncomfortable
91–97	Stuffy and uncomfortable
>97	Extremely hot and uncomfortable

). The thermal comfort level was divided into five categories, as human perception varies from comfortable to extremely uncomfortable. Given that the research focused on the period during summer, it also varies from cool to extremely hot.

2.3.3. Calculation of Association between UTCI and LUCC Pattern

Using geographically weighted regression (GWR) to analyze the relationships between the UTCI and LUCC patterns reveals not only the relationships between the UTCI and the LUCC, but also considers spatial non-stationarity. The general GWR is defined as follows in Equation (5):

$$y_i={\beta }_0\left(u_i,v_i\right)+{\displaystyle \sum _{j=1}^{}{\beta }_j\left(u_i,v_i\right)x_{ij}+{\epsilon }_i},$$

(5)

where $\left(u_i,v_i\right)$ represents the coordinates of location I; β₀ and β_j represent the intercept and coefficient of the independent variables; $y_i$ represents the dependent variable (i.e., UTCI, in this study); x_ij represents the jth independent variable at the ith location (i.e., the type of LUCC in this study); and ${\epsilon }_i$ represents the random error at the ith location.

3. Results

3.1. Analysis of Land Use Transfer Characteristics

From 2005 to 2020 (Figure 3), with the development of the social economy and urbanization, the construction land area of BTH increased by 44.26%, mainly concentrated in Beijing, Tianjin and Shijiazhuang, extending outward. The area of construction land in GBA increased by 39.10%, expanding outward, with Guangzhou and Shenzhen as the center. The construction land area in the YZD increased by 29.99%, initially expanding inland in the eastern coastal zone, and the area of construction land in the northern region increased rapidly. The construction land area ratio of the three major urban agglomerations is increasing, and the spatial distribution of the construction land is characterized by agglomeration, which is more common in developing countries.

Affected by geographical location, climatic conditions and other factors, the main land use types of the three regions are different. The main types of BTH and YZD are cultivated land, and those of GBA are forest land. However, the changes in these three regions are similar, mainly for the two types of land conversion into construction land and the proportion decreased year by year.

3.2. Analysis of SUHI Characteristics

The spatial–temporal pattern distribution of heat islands in the three urban agglomerations verifies that the LST results have high accuracy through the satellite–ground synchronous observation (Figure 4). Thus, the surface temperature of each period is obtained by the same method and flow path. The SUHII shows a high heat island intensity in the study areas, which are clustered and shredded through the standardized correction of the surface temperature and temperature profile analysis. Moreover, with the further development and spread of the city, these high-SUHII patches fuse and form larger SUHII patches, further fusing. Thus, in a visual interpretation, these patches tend to expand outwards with urbanization. The temperature difference between urban areas and surrounding areas is increasing, and the range of existing temperature differences tends to expand outwards.

From the perspective of spatial distribution, the SUHI of these three urban agglomerations is mainly distributed in the areas with the most concentrated construction land:

(1) Through calculation, we obtain the area of the heat island region where the SUHII index is greater than zero. The annual average rates of heat island area growth in the three regions are 1.01%, 1.41% and 1.09%, respectively. (2) The spatial distribution of the thermal landscape has changed, showing the characteristics of scattered-contiguous-diffusion transfer. (3) The development of functional relationships between cities affects the distribution of heat island intensity. With the enhancement of interurban interaction, heat islands between cities also present a state of adhesion, showing a shape like a “hump”.

3.3. Analysis of Urban Thermal Comfort Change Characteristics

Based on the normalized and hierarchical UTCI, many plots exceed the normal index in the three urban agglomerations in summer, and the size of the area and index has an increasing trend. From the perspective of spatial dimension, the high-index region of BTH urban agglomeration is mainly distributed in belt areas such as Beijing and Tianjin, and the grade of the ecological conservation area in the northwest is generally good. The high index areas of GBA are mainly distributed in Guangzhou City and Shenzhen and have a trend of expansion to the Pearl River region. By comparison, the high-index areas of the YZD are mainly distributed in the northern and eastern coastal areas of Anhui and have a trend of inland expansion.

By dividing the UTCI value into five levels according to the human senses, the thermal comfort level of the human body in the study areas was obtained (Figure 5). The annual distribution of thermal comfort (Figure 5) in Beijing, Tianjin and Hebei is clustered, and the discomfort zone is mainly distributed in the southern and eastern parts of the region. For the UTCI type, the main thermal comfort type in 2010 and 2020 is “worse” (Figure 6). The annual distribution of thermal comfort in Guangdong, Hong Kong and Macao was relatively concentrated on the two sides of the Pearl River. Meanwhile, in 2005, 2010 and 2015, the UTCI types in this region were mainly “normal” types, with the average area accounting for 37.72% of the total area. The annual distribution of thermal comfort in the YZD is concentrated. Meanwhile, the dominant type of the four years is “normal”, and the annual average area accounted for 36.75%. The comparison of the three urban agglomerations reveals that BTH faces the highest UTCI risk in terms of the proportion of thermal discomfort area or the proportion of thermal discomfort type.

3.3.1. Thermal Environment and Urban Thermal Comfort

The study of thermal environment, an important factor affecting UTCI, and its driving force are particularly critical for optimizing the structure of urban UTCI. Combined with an LST analysis and the spatial distribution, the LST index and UTCI level presented a high degree of similarity (Figure 6 and Figure 7), which explains that the LST’s contribution to the UTCI is positive. When other things are equal, an exorbitant surface temperature will increase the UTCI. Combined with an SUHII analysis, the SUHI effect is affected by the surface temperature, wind speed, building form and urban blue–green space. In some areas, the UTCI is high, but the heat island intensity is low, showing a difference between thermal comfort and the thermal environment. This difference reflects the different orientations of the LST and UTCI. The UTCI is more inclined to express human sensory responses to the thermal environment by introducing human sensory data such as humidity. However, the SUHII focuses on local high temperatures by calculating the extreme temperature differences in the same area. Therefore, studying the relationship between the LUCC and the UTCI provides a perspective on the response of human thermal comfort to urban surface changes.

3.3.2. Land Use Change and Urban Thermal Comfort

The GWR model was used to analyze the spatial non-stationarity relationship between the UTCI and the LUCC. The R² and adjusted R² are often used to assess the goodness of fit between the model and the data [38]. In this study, the result (

Table 3. GWR diagnostics.

Region	Year	Residual Squares	R2	Adjusted R2	AICc	Effective Number
BTH	2005	26.109	0.551	0.417	51.559	3.100
	2010	11.873	0.752	0.627	47.938	4.091
	2015	38.348	0.500	0.380	69.853	2.697
	2020	37.130	0.531	0.472	59.217	2.006
GBA	2005	23.158	0.182	0.090	48.675	2.008
	2010	21.624	0.342	0.268	47.815	2.008
	2015	44.804	0.371	0.301	55.843	2.008
	2020	42.886	0.395	0.328	55.197	2.008
YZD	2005	67.609	0.754	0.641	167.174	13.593
	2010	95.670	0.753	0.647	179.903	13.029
	2015	55.923	0.703	0.575	158.090	13.145
	2020	75.582	0.801	0.699	174.642	14.762

) shows that, in terms of the R² and adjusted R², the GWR model has better goodness of fit with the data and more significant overall regression in the regions of YZD and BTH, with the R² between 0.5 and 0.7 and adjusted R² between 0.4 and 0.6. GBA’s R² and adjusted R² are lower, between 0.2 and 0.3. The AICc performs better in assessing the model’s performance [39]. From the perspective of the AICc, the GWR model has better performance in BTH and GBA compared with the YZD areas, with a significant difference in the AICc values. In general, the goodness of fit can support the discussion of the relationship between the UTCI and LUCC in this paper, to a certain extent, and the performance of the GWR model varies in different regions.

The GWR analysis coefficient varies for land use types, geographical positions and time (Figure 8, Figure 9, Figure 10, Figure 11, Figure 12 and Figure 13). It means that different LUCC types have different correlations in controlling the UTCI.

In the GWR results of the BTH region (Figure 8), the built-up land positively contributes to the UTCI. In this study, a high UTCI in summer also means hot and uncomfortable. On the contrary, forest and grassland negatively contribute to UTCI, which means it will be more comfortable there than in the built-up land. The absolute value of built-up land is higher than that of grassland and forest, which is consistent with the result (Figure 5) that the uncomfortable thermal area in BTH is quite large. The proportion of the UTCI of each type of land has been analyzed in Figure 9; contrary to the forest, most built-up lands have “worse” or the “worst” thermal comfort levels.

In the GWR results of GBA (Figure 10), the positive contribution of built-up land and the negative contribution of grassland and forest are the same as that of BTH. However, the absolute value of built-up land is nearly equal to grasslands and forests, consistent with the result above (Figure 5). In 2005–2015, the “worse” and “worst” areas were less than those of BTH in almost all types of land, but the proportion increased in 2020 (Figure 11).

In the GWR results of YZD (Figure 12), the positive contribution of built-up land is much higher than the negative contribution of grassland and forest in certain areas. A further analysis (Figure 13) shows that the thermal environment of YZD is not the worst.

Overall, the built-up land contributes more positively to the UTCI than other LUCC types. Conversely, forest and grassland contribute negatively to UTCI. The contribution of different land types varies in different geographical positions, which their surroundings may influence.

4. Discussion

China is undergoing global warming in regional areas, and the temperature increase in China is substantially higher than the global warming level [40,41], suggesting the effects of other factors such as urbanization. The objective of this study is to investigate the association between the LUCC relevant to urbanization and the urban surface thermal environment, whereas other factors affecting the urban thermal environment are not discussed.

In the present study, the analysis of land use change shows that the rate of the built-up area increased by approximately 30% to 45% in 2005–2020 in our study areas, and the expansion of built-up areas mostly occurred over arable land. This increase in built-up land can be directly linked to the rapid growth of the economy and population [42]. Many cities in developing countries have reported that urbanization increased local warming. In the three urban agglomerations, the LST and SUHII are generally increasing, whereas the fluctuation of LST and SUHII locally is associated with the dynamics of increase and decrease in vegetation. Another significant result of this study is that the UTC has deteriorated along with the rise in LST and SUHII. In addition, the improvement effect of forest land on thermal comfort is more significant than that of water bodies. Furthermore, increasing water bodies in an extremely hot environment leads to increased humidity and reduced thermal comfort.

The results obtained in this study show that the changing LUCC pattern is closely related to the change in UTC, and the UTC is generally deteriorating. This is similar to the previous studies [32], whereas the previous studies did not evaluate the thermal comfort associated with LUCC, which is also the main novelty of this study. The findings of this study suggest that the provision of green and blue spaces in urban areas is useful in improving the urban thermal environment. However, providing blue spaces without regard to the human body’s thermal comfort perception can be useless.

Among various climate factors, the human body’s comfort is mainly influenced by temperature and humidity, which are directly involved in the body and the external environment of heat and moisture exchange. However, thermal comfort is also influenced by wind, air pressure, air oxygen and ultraviolet radiation, especially in mountainous and plateau areas. The study area is mainly located in the plains and eastern China coastal areas; thus, we do not consider these factors.

5. Conclusions

Based on the MODIS LST products, RH data and LUCC data, this study calculates and analyzes the spatial distribution and temporal transformation characteristics of the SUHI and UTC in the BTH, GHM and YRD urban agglomerations from 2005 to 2020 using the THI model based on remote sensing data. The following conclusions were reached:

(1)

By calculating the SUHII from 2005 to 2020, the three urban agglomerations have many heat island intensity regions, mostly clustered and zonally distributed. With the trend of the outward expansion of high-temperature areas in the urbanization process, urbanization is the main driving factor of the SUHI effect.

(2)

The calculation of the UTCI from 2005 to 2020 shows many uncomfortable areas in the three urban agglomerations in summer, with an area and exponential growth trend. From the spatial point of view, the high-index areas of Beijing, Tianjin, Hebei, Guangdong, Hong Kong and Macao expand to the coast, and the high-index areas of the YZD expand to the inland. In terms of time, the annual distribution of thermal comfort in BTH is discrete, and the annual distribution of thermal comfort in GBA is concentrated. Among the three urban agglomerations, BTH faces the highest UTCI risk.

(3)

From the spatial and temporal distribution of the SUHII, LUCC and UCTI, a high UTCI does not mean a high SUHII, but a high UTCI will bring a high UTCI effect. Land use change, especially the change intensity in construction land, increases the UTCI in each urban agglomeration to varying degrees. Among them, the change intensity in construction land has a significant indigenous effect on the change in thermal comfort. The area where the thermal comfort index increases correlated with the area where the construction land increases.

The main limitation of this study is that it only applies to urban agglomeration in plains, and the influence of topography on thermal comfort was not discussed. Moreover, this study focused on summer daytime, and the change in the urban thermal environment at other times is not mentioned. The indicators established and adopted in this study are utilized in the research of SUHI and UTC of other urban agglomerations.