Research on Summer Hourly Climate-Influencing Factors in Suburban Areas of Cities in CFA Zone—Taking Chengdu, China as an Example

Abstract

Elevated temperatures in urban centers have become a common problem in cities around the world. However, the climate problems in suburban areas are equally severe; there is an urgent need to find zero-carbon ways to mitigate this problem. Recent studies have revealed the thermal performance of vegetation, buildings, and water surfaces. They functioned differently regarding the climate at different periods of the day. Accordingly, this study synthesizes remote sensing technology and meteorology station observation data to deeply explore the differences in the role of each climate-influencing factor in the suburban areas of Chengdu. The land surface temperature (LST) and air temperature (T_a) were used as thermal environmental indicators, while the normalized difference vegetation index (NDVI), normalized difference water index (NDWI), normalized difference built-up index (NDBI), and altitude were used as environmental factors. The results showed that the relevant influences of the environmental factors on the climate in the sample areas were significantly affected by the time of the day. The NDVI (R² = 0.5884), NDBI (R² = 0.3012), and altitude (R² = 0.5638) all showed strong correlations with T_a during the night (20:00–7:00), which gradually weakened after sunrise, yet the NDWI showed a poorer cooling effect during the night, which gradually strengthened after sunrise, reaching a maximum at 15:00 (R² = 0.5012). One reason for this phenomenon was the daily weather changes. These findings facilitate the advancement of the understanding of the climate in suburban areas and provide clear directions for further thermal services targeted towards people in different urban areas.

1. Introduction

Cities play a crucial role in modern human settlements and are significant contributors to climate change [1]. The rapid urbanization and expansion of urban land areas have ongoing implications for the sustainability of urban ecosystems [2]. With the global population surpassing eight billion in 2022 [3], the human-induced urban heat island (UHI) and global warming are exacerbating the rapid deterioration of the urban climate. The adverse environmental impacts resulting from accelerated urbanization may surpass those caused by human activities during the Industrial Revolution in the 19th century [4]. Despite regional variations in the global climate, the urban heat island and high urban temperatures have become ubiquitous characteristics of the deteriorating urban climate worldwide [5]. In recent years, the UHI [6,7] has been of increasing concern to scholars because it not only affects biodiversity [8], vegetation activity [9], and the local climate but also has many negative impacts on humans, such as causing droughts and fires, affecting the air quality [10,11,12], increasing energy consumption, and even increasing mortality rates [13]. Among them, an urban area with a higher surface temperature than the surrounding suburbs is known as a surface urban heat island (SUHI) [14], and one of the key parameters in SUHI studies in recent years is the SUHI intensity [15], which is characterized by the surface temperature. In recent years, studies on the urban heat island have focused on the inner city [16,17]; however, little research has been conducted to characterize the climate changes in suburban areas. Studies have shown that, for the same urban temperature, different SUHI intensities can occur when varying the rural cover reference [18]. Therefore, the exploration of climate change in suburban areas is imperative to further explore the drivers of the urban heat island.

There are several factors influencing the urban climate. Recent studies have revealed the thermal adjusting effects of vegetation, buildings, and water surfaces [19]. These findings were acquired using various technical methods. They confirmed that the urban temperatures were higher than the surrounding natural ones via field measurement (FDMT) [20]; a similar phenomenon was found via remote sensing (RMSS) [21] and meteorology station observation (MSO) [22]. Nevertheless, vegetation showed little cooling effect on spaces at certain heights (above 20 m) [23]. Water surfaces had similar impacts on the urban climate and subjective thermal perceptions [24] and even some shallow-level water surfaces [25]. These studies were conducted by simulation (SIMU), RMSS, and FDMT [19]. Gray spaces are also key to urban heat stress. Their warming degrees could be reduced by the properties of certain materials [26]. However, the real magnitude of the afternoon temperature drop caused by the albedo increase is close to 0.09 °C per 0.1 rise in the albedo [27]. The role of urban green, blue, and gray infrastructure, represented by vegetation, water bodies, and anthropogenic city structures, in the urban heat island is still controversial [28,29]. Still, the factors associated with them have been gradually used as important technological tools to mitigate the urban heat island and enhance the resilience of cities and their suburban ecosystems [30]. They have been shown to have good social and ecological value in general [31].

Overall, a wide range of urban elements are influential in meteorology adjustments. These facts have been revealed by studies using several of the abovementioned methods [19]. However, their shortages and the conflicts between different studies cannot be ignored [32]. A few investigations, for instance, found that water bodies were insignificant for cooling in some conditions [25]. These shortcomings might result from the limitations of a single method, since each study usually uses one approach. Therefore, it is necessary to continue the exploration with methodological enhancements.

This study attempted to combine several different methods to conduct a comprehensive exploration of the climate influences in urban suburbs. The interactive compensation between different methods could avoid the limitations of studies using a single one. In turn, it could provide a data reference and theoretical support for the construction of urban spatial strategies based on the concept of the sustainable development of cities and society.

2. Methodology

2.1. Site Selection

Chengdu (Cfa) is a city located in the center of Sichuan, Southwest China. It is a plain formed by the impacts of the river flow, with an area of around 14,335 km² (Figure 1). It is dominated by plains, plateaus, and some hills, with altitude differences of more than 2000 m [33]. Although located in the temperate zone, the unique geography frequently causes extreme weather [34]. Local residents suffer from thermal discomfort resulting from the hot–humid meteorology, especially in summer (Figure 2).

As an important large city in the west of China, Chengdu is divided into various administration zones, including 12 districts and three counties. Among them, Wuhou District, Jinjiang District, Qingyang District, Jinniu District, and Chenghua District are the core urban areas of Chengdu [36], which have the highest population densities [37] and housing expenses [38]. Nonetheless, outskirt areas have been developing quickly in recent decades; the whole city has significant sprawl. This has resulted in more serious urban climate degradation [39]. In the summer of 2022, continuous hot weather (Figure 3) led to a major gap in the energy supply in Chengdu, resulting in the near paralysis of the city’s normal production and living order [40]. In order to cope with the rapidly growing urban heat island and energy problems, the establishment of an urban spatial strategy based on the concept of the sustainable development of the city and society has become a consensus among local governments. The study of the climate factors affecting cooling in Chengdu will help to deepen the understanding of the region’s climate and provide references and suggestions to improve the quality of life of the current residents, reduce the urban energy consumption, and minimize the urban heat island effect.

2.2. Data Resources

2.2.1. Data Background

Earlier studies about the UHI acquired data through SIMU and/or FDMT; these studies mainly focused on exploring the effects of factors such as vegetation [41], water bodies [42], the population [43] and buildings on air temperature (T_a) change. The scale of these studies was relatively small, and the period was short. Therefore, they could not fully reflect the differences in the effects of climate change factors at the urban scale and were not able to describe the differences in the effects of the climate at the urban scale.

MSO is a traditional method of regional climatic data acquisition. It acquires long-term and abundant meteorology data, which are available for UHI research at a macro scale [41]. The meteorology data from the meteorological stations are usually collected hourly [44]. Nevertheless, the stations are usually constructed at fixed sites around the city. They are representative of the meteorology at certain locations only [45]. This would limit the data available for a wider range of exploration.

In large-scale studies, RMSS is often utilized as well. RMSS uses sensors on satellites to collect thermal and geographical data on land surfaces. Full data across scanning ranges could be collected [46]. Therefore, the climate of the whole city with certain periodic spans could be studied. This method suits research at the macro scale, making it similar to MSO. The collected data are available for long-term meteorology research, such as regional climate change [47]. However, satellites lack flexibility. They collect daily data for a fixed period of time, only once a day. Hourly data for one full day are challenging to acquire [48].

MSO and RMSS are popular measures for urban climate studies. They have both benefits and shortages [19]. Thus, a study using a single method is not able to achieve a comprehensive investigation. This study attempts to utilize these two methods to implement a complex investigation. The shortages of either of them can be avoided in this case.

2.2.2. Data Selection

A variety of meteorological parameters are available in thermal environmental studies, i.e., the air temperature (T_a), land surface temperature (LST), air velocity (V_a), etc. [19]. The T_a and LST are often used to evaluate UHI effects [49]. Various tools are used for data acquisition, such as the Testo 480 for FDMT and HOBO for MSO [50]. The temperature difference is available, indicating the level of the UHI, which is defined as the UHI intensity (UHII). A few studies have utilized several thermal indices (e.g., T_a and LST) for comprehensive evaluation [51]. This study selected them for a comparative investigation.

The urban climate is a complex phenomenon influenced by a variety of factors. Apart from unchangeable geographical elements, the local population [52], land use and land cover (LULC) [53], longitude and latitude [54], altitude [55], urban spatial form [56], natural climate [57], green infrastructure (vegetation) [58], blue infrastructure (water surfaces) [24], and gray infrastructure (buildings) [26] are climatically influential. There are a variety of parameters proposed for evaluation. A geographic information system (GIS) can digitalize them in regard to their properties. Vegetation, anthropogenic city structures, and water surfaces are indicated by the normalized difference vegetation index (NDVI) [59], normalized difference built-up index (NDBI) [60], and normalized difference water index (NDWI) [61]. They are effective for the study of microclimates [26]. In addition, the altitude of the meteorological stations in the sample cities varies considerably (448.4–698.5 m), which might be meteorologically influential. This study considered the altitude as another affecting factor.

In total, there were six parameters involved. T_a and LST were available as climate indices, whereas the rest were their affecting factors.

2.3. Climatic Indicators

2.3.1. Thermal Indices from Meteorological Stations

Urban climate research previously focused on the main urban area. It compared the temperatures between cities and surrounding suburban areas, revealing the countryside being cooler than the central districts [62]. This has been common knowledge since the advancement of the UHI around three centuries ago [63]. For a different exploration and to obtain different findings, this study avoided the main urban areas, which have been widely proven to be warmer; outskirts with insignificant building density variations were selected as samples for this climate investigation. These towns or counties were located outside the main area of Chengdu. Based on the existing administrative divisions and the meteorological station distribution [64] in Chengdu, each district is equipped with a meteorological station that provides hourly meteorological data for the entire day, corresponding to its administrative district. In total, 13 meteorological stations (Figure 4) were available for data issuing (1—Shuangliu, 2—Dayi, 3—Chongzhou, 4—Pengzhou, 5—Xinjin, 6—Xindu, 7—Wenjiang, 8—Pujiang, 9—Qionglai, 10—Pixian, 11—Dujiangyan, 12—Jintang, 13—Longquanyi). There are various types of meteorology parameters collected by each station, including T_a, V_a, and RH. T_a directly influences subjective thermal sensations [65], and it is often used in weather reports. This study selected the hourly T_a of summer (July and August) from these variables for climatic analyses. There were a few stations (No. 2, No. 5, No. 8, No. 9) far from the metropolitan area of Chengdu, some stations (No. 6, No. 13, No. 11, No. 3, No. 4, No. 12) were further distanced from it, and other stations (No. 1, No. 7, No. 10) were close to it. All districts were within the climatic zone of the Sichuan Basin Climate; thus, the local climates were mostly affected by the regional environments, rather than geographical factors.

2.3.2. Thermal Indices through Remote Sensing

The Landsat 8 satellite (NASA & USGS, CA, USA) is a commonly used data collection tool for climate research. It collects data for any full city. The LST acquired by it is often used as a climatic index in studies. Generally, the LST is found to be linearly correlated with land cover parameters, e.g., a negative correlation between the NDVI and LST [66]. This study utilized the Landsat 8 to acquire the LST of selected sites in July and August of both 2019 and 2020 (summer in the city). The LST is a complex index with diverse thermal parameters. Its calculation process and relevant parameters are expressed in Equation (1):

$${\mathrm{T}}_{\mathrm{S}}={\displaystyle \frac{\left[\mathrm{a}\left(1-\mathrm{C}-\mathrm{D}\right)+\left(\mathrm{b}\left(1-\mathrm{C}-\mathrm{D}\right)+\mathrm{C}+\mathrm{D}\right){\mathrm{T}}_{\mathrm{i}}-{\mathrm{DT}}_{\mathrm{a}}\right]}{\mathrm{C}}}$$

(1)

where ${\mathrm{T}}_{\mathrm{a}}$ and ${\mathrm{T}}_{\mathrm{i}}$ refer to the air temperature and brightness temperature, C and D are the parameters, and the detailed calculation process of ${\mathrm{T}}_{\mathrm{i}}$ is expressed as follows (Equation (2)):

$${\mathrm{T}}_{\mathrm{i}}={\displaystyle \frac{{\mathrm{K}}_2}{\ln \left(\frac{{\mathrm{K}}_1}{{\mathrm{L}}_{\mathsf{\lambda }}}+1\right)}}$$

(2)

where ${\mathrm{L}}_{\mathsf{\lambda }}$ refers to the spectral radiation value, and ${\mathrm{K}}_1$ and ${\mathrm{K}}_2$ are thermal conversion coefficients.

The mean temperatures of the atmosphere in middle-latitude areas were linearly correlated with that of land nearby spaces; the detailed calculation process of ${\mathrm{T}}_{\mathrm{a}}$ is expressed as in Equation (3):

$${\mathrm{T}}_{\mathrm{a}}=16.0110+0.92621{\mathrm{T}}_{\mathrm{o}}$$

(3)

where ${\mathrm{T}}_{\mathrm{o}}$ is the surface layer temperature.

Equations (4) and (5) are available to calculate parameters C and D:

$${\mathrm{C}}_{\mathrm{i}}={\mathsf{\epsilon }}_{\mathrm{i}}{\mathsf{\tau }}_{\mathrm{i}}$$

(4)

$${\mathrm{D}}_{\mathrm{i}}=\left(1-{\mathsf{\tau }}_{\mathrm{i}}\right)\left[1+\left(1-{\mathsf{\epsilon }}_{\mathrm{i}}\right){\mathsf{\tau }}_{\mathrm{i}}\right]$$

(5)

where ${\mathsf{\epsilon }}_{\mathrm{i}}$ is the land surface emissivity of band i, while ${\mathsf{\tau }}_{\mathrm{i}}$ is the atmospheric transmissivity of band i. They can be provided by the official website of NASA. This study selected the NDVI threshold method for the land surface relative emissivity [67].

The distributions of LST data can be acquired by grids in certain ranges. This study defined grids in cubes with dimensions of 10 km [67]. There were 131 groups of data defined to evaluate the urban climate at the outskirts of Chengdu (Figure 5 and Figure 6). This study selected a radius of 300 m around each point position for mean LST calculation [68].

2.4. Factors Affecting Urban Climate

LULC significantly influences the climate. Three factors regarding LULC are relatively important: vegetation, anthropogenic city structures, and water surfaces. They are defined as green, gray, and blue infrastructure. RMSS proposes several parameters to indicate them: the NDVI, NDBI, and NDWI for vegetation, anthropogenic city structures, and water, respectively.

The NDVI indicates the plant conditions on the land surface at any point. It is valued between −1 and 1; a higher value means denser coverage, which is captured by color sensing [69]. Usually, it can be calculated via Equation (6):

$$\mathrm{NDVI}={\displaystyle \frac{\mathrm{NIR}-\mathrm{Red}}{\mathrm{NIR}+\mathrm{Red}}}$$

(6)

NIR refers to the near-infrared band, while Red is the red band. The NDVI digitally defines the condition of vegetation on the land surface. It is difficult to precisely distinguish the detailed land use, e.g., buildings or trees. Generally, vegetated sites have NDVI values above 0, sites without plants have negative NDVI values (−1–0), and this parameter is often revealed to be negatively correlated with the LST.

The NDBI is a physical parameter that indicates building cover information. It has a similar processing procedure to the NDVI. This study chose to evaluate anthropogenic city structures (gray infrastructure (where buildings are the focus of this study)) on land surfaces. Equation (7) is available to calculate this:

$$\mathrm{NDBI}={\displaystyle \frac{\mathrm{MIR}-\mathrm{NIR}}{\mathrm{MIR}+\mathrm{NIR}}}$$

(7)

MIR refers to the middle-infrared band. As a parameter indicating buildings (a key factor of the UHI), the NDBI is linearly correlated with thermal stress [70].

The NDWI is available to indicate large open water surfaces. The sizes and qualities of water surfaces can be reflected by it. The values of the NDWI be calculated via the following Equation (8):

$$\mathrm{NDWI}={\displaystyle \frac{\mathrm{Green}-\mathrm{NIR}}{\mathrm{Green}+\mathrm{NIR}}}$$

(8)

where NIR and Green refer to the near-infrared band and the green band. The NDWI is usually negatively relevant to the heat level as the natural cooling property of water [71].

Thermal indices in microscopes were graded with dimensions of 10 km. A total of 131 sites were selected. This study obtained comprehensive geographical information for Chengdu according to a radius of 300 m. Data were acquired through the ArcGIS 2022 platform (Figure 7). The central points of the 13 meteorological stations were available as circles with a radius of 300 m. The mean values of the NDVI, NDWI, and NDBI within the circles were available, defining the physical environments of the sites (Figure 8).

In addition to the physical environment, the local geography is also influential, and the altitudes of the 13 meteorological stations were included in this study (Figure 9).

2.5. Data Analyses

Urban climate research can obtain results through data analyses. There are various statistical models available for prediction. Among them, Pearson correlation analysis is used to evaluate the correlations between different factors [21]. Numerous factors affect the climate of a meteorological station, and we can screen out the factors with significant influences as independent variables for regression analysis. As a trend prediction method, multiple linear regression (MLR) [72] is frequently used to measure the correlations between one dependent variable and several groups of independent variables. Different meteorological stations have different climate conditions, and T_a can be used as a dependent variable. The hourly T_a selected in this study for a total of 124 days from 2019 to 2020 is affected by several independent variables, including the NDVI, NDWI, NDBI, altitude, time, etc., and they were all included in the MLR models for analyses in SPSS 26 [72]. The thermal effects of various meteorological stations at different times of the day were indicated by R² and p values (evaluated by R² values and p < 0.05 to indicate whether they are significantly correlated). The research methodology flow chart is shown in the Appendix A (Figure A1).

3. Results for Data from Remote Sensing

3.1. Data Description

The three-dimensional scatterplot of the NDBI–NDVI–NDWI, established from 131 sets of normalized index values in the suburbs of Chengdu, is depicted in Figure 10. In the three-perimeter space, the scatterplot exhibits a predominantly high–left and low–right shape. The high–left area primarily represents built-up land pixels, while the low–right area mainly consists of water bodies and vegetation pixels. Notably, a strong correlation among the three indices is observed in the figure. Specifically, the NDBI demonstrates a negative correlation with both the NDVI and NDWI; furthermore, it can be inferred that regions with higher NDBI values correspond to lower values of the NDVI and NDWI. Additionally, there exists a positive correlation between the NDVI and NDWI within this context.

Figure 11 illustrates the statistical distribution of the data for 131 selected points and 13 meteorological stations in the suburbs of Chengdu (13 sites were calculated separately for each year: 2019 and 2020). The standard deviations at these 131 points for the NDVI, NDBI, and NDWI were 0.193, 0.110, and 0.145, respectively. Correspondingly, for the 13 meteorological stations, they were 0.241 (NDVI), 0.091 (NDBI), and 0.119 (NDWI), respectively. Furthermore, a negative correlation was observed between the NDBI values and both the NDVI and NDWI values overall. Additionally, the dispersion of the overall NDVI in the urban suburbs resembled that of the meteorological stations. The standard deviation analysis conducted on each normalized index at the meteorological station revealed that the NDBI exhibited the lowest degree of dispersion while having the lowest sample richness. Meanwhile, the NDVI had significantly higher dispersion levels, which were approximately 2.7 times greater than those of the NDBI dataset, and it had the best sample richness among the three datasets. Thus, the subsequent study of the 13 stations can be used as a proxy for the results of the entire large-scale study of the urban suburb.

3.2. Data Analyses

Linear correlations between the LST and various parameters are illustrated in Figure 12. Each one shows a significant effect, either negative (NDVI, R² = 0.4283 & NDWI, R² = 0.3251) or positive (NDBI, R² = 0.3638). At point 69 (LST = 43.44), which has the highest LST value, NDVI = 0.1876, NDWI = −0.7996, NDBI = 0.0653; meanwhile, at point 55 (LST = 28.82), which has the lowest LST value, NDVI = 0.6248, NDWI = −0.0222, NDBI = −0.2528. Among these factors, the NDVI showed a stronger influence than the other factors. An increase in either the NDVI or NDWI by one point caused LST reductions of 9.01 °C and 10.42 °C, respectively, whereas an increase in the NDBI by one point resulted in an increase in the LST of 14.59 °C (Figure 12). Since there exists a significant correlation between the air temperature and surface temperature, there also exists a strong correlation between the air temperature and urban green, blue, and gray infrastructure at the urban scale. Meanwhile, vegetation and water effectively cool down the thermal environment, while buildings contribute to warming it up.

4. Results for Data from Meteorological Stations

4.1. Data Description

These fluctuations exhibited similar patterns, with an increase observed early in the morning (around 7:00), reaching a peak at 17:00, and then gradually decreasing until the following morning. Station 12 had the highest value, with a maximum of 29.6 °C (16:00) and a minimum of 22.9 °C (5:00). Stations 6 and 13 were relatively hot, reaching nearly 29 °C in the afternoon. Stations 8 and 11 were cooler than most of the other ones. It could be seen from the image that the data issued by the stations had certain regularity. The ones located far from the city center had lower values. However, they rarely continuously fall as the distance increases (

Table 1. Average temperatures of meteorological stations and distance from main urban area.

ID	Name	Mean Temperature (°C)	Temperature Difference (°C)	Distance from City (km)
1	Shuangliu	26.13	0.52	17.12
2	Dayi	25.31	−0.30	53.57
3	Chongzhou	25.20	−0.41	35.64
4	Pengzhou	25.97	0.36	38.10
5	Xinjin	25.79	0.18	33.79
6	Xindu	26.28	0.67	17.31
7	Wenjiang	25.29	−0.32	21.93
8	Pujiang	25.07	−0.54	76.22
9	Qionglai	25.41	−0.20	58.09
10	Pixian	25.64	0.03	24.93
11	Dujiangyan	24.54	−1.07	53.94
12	Jintang	26.32	0.71	40.52
13	Longquanyi	26.01	0.40	21.09
Temperature difference = mean temperature from each site—average value (28.05 °C)

). This conflicts with traditional findings about the urban climate, such as the UHI footprint [73], which might result from the fact that part of the periphery of the Chengdu core area has been designated as an urban green corridor, and the ecological environment varies greatly from one point to another (Figure 13).

The LST data of the 13 meteorological stations were associated with T_a in linear regression for the validation test (Figure 14). They showed a significant positive correlation (R² = 0.672). Therefore, both the meteorology data of the stations and the satellite exhibited similar patterns. This suggests that both the LST and T_a are valid indicators of the local climate. The T_a data from the stations were used for further analyses.

4.2. Linear Analyses for Station Data

The values of T_a from the meteorological stations were correlated with various influential factors, including LULC, the time of the day, and the altitude, using Pearson’s correlation. The significance values for Ta and each environmental factor were found to be less than 0.01. Among these factors, the time of day had the most significant impact on the meteorological conditions as the temperature fluctuates throughout the day. Furthermore, the NDVI, NDWI, and altitude showed a negative correlation with Ta, indicating their potential to mitigate urban heat by cooling down the city. Conversely, the NDBI exhibited a positive correlation with Ta and was found to significantly contribute to increased temperatures in the city. These findings are consistent with previous studies [74].

4.3. Analyses against Each Factor

The previous section has proven that both environmental factors and time are significantly influential on the meteorology. For further exploration, this study associated the hourly meteorological data with each environmental factor independently by linear regression. The meteorology index (hourly) was available as the dependent variable. Key information from the analyses (R² and significance) is presented in

Table 2. Correlation coefficients and significance (R²) between hourly temperature and local factors.

Time	R2				Coefficient				Sig.
	NDVI	NDWI	NDBI	Altitude	NDVI	NDWI	NDBI	Altitude	NDVI	NDWI	NDBI	Altitude
0	0.5867	0.1714	0.2499	0.5456	−1.880	−2.057	3.268	−5.109	0.000	0.035	0.009	0.000
1	0.5534	0.1478	0.2364	0.5237	−1.772	−1.854	3.085	−4.859	0.000	0.052	0.012	0.000
2	0.5342	0.1574	0.2089	0.5071	−1.748	−1.921	2.911	−4.799	0.000	0.045	0.019	0.000
3	0.5361	0.1535	0.2149	0.5109	−1.623	−1.758	2.738	−4.466	0.000	0.048	0.017	0.000
4	0.5506	0.1409	0.2161	0.5468	−1.634	−1.673	1.060	−4.589	0.000	0.059	0.017	0.000
5	0.5234	0.1336	0.2155	0.561	−1.588	−1.624	2.714	−4.633	0.000	0.066	0.017	0.000
6	0.5048	0.1329	0.2115	0.5682	−1.484	−1.542	2.560	−4.439	0.000	0.067	0.018	0.000
7	0.5201	0.1574	0.2191	0.5266	−1.520	−1.693	2.629	−4.312	0.000	0.045	0.016	0.000
8	0.4831	0.1222	0.2376	0.2384	−1.593	−1.622	2.976	−3.155	0.000	0.080	0.012	0.011
9	0.4891	0.2486	0.2496	0.1251	−1.388	−2.002	2.640	−1.978	0.000	0.010	0.009	0.076
10	0.3913	0.3357	0.202	0.1395	−1.195	−2.242	2.288	−2.012	0.001	0.002	0.021	0.060
11	0.2621	0.2494	0.191	0.2277	−1.138	−2.248	2.588	−2.990	0.008	0.009	0.026	0.014
12	0.2918	0.393	0.1119	0.3357	−1.121	−2.634	1.850	−3.389	0.004	0.001	0.095	0.002
13	0.1986	0.4332	0.0732	0.3077	−0.935	−2.796	1.512	−3.281	0.023	0.000	0.181	0.003
14	0.1662	0.4719	0.0578	0.3179	−0.924	−3.153	1.452	−3.603	0.039	0.000	0.237	0.003
15	0.1318	0.5012	0.0365	0.3384	−0.889	−3.512	1.308	−4.017	0.068	0.000	0.350	0.002
16	0.1618	0.4872	0.0596	0.3791	−1.006	−3.534	1.626	−4.339	0.042	0.000	0.229	0.001
17	0.2444	0.4463	0.0899	0.4362	−1.292	−3.536	2.088	−4.866	0.010	0.000	0.137	0.000
18	0.2996	0.3925	0.119	0.4641	−1.411	−3.269	2.368	−4.950	0.004	0.001	0.084	0.000
19	0.3364	0.3221	0.1795	0.4901	−1.375	−2.724	2.676	−4.677	0.002	0.002	0.031	0.000
20	0.4533	0.1659	0.266	0.5449	−1.638	−2.007	3.344	−5.063	0.000	0.039	0.007	0.000
21	0.5387	0.1253	0.2835	0.528	−1.932	−1.886	3.733	−5.391	0.000	0.076	0.005	0.000
22	0.5754	0.133	0.3012	0.559	−1.974	−1.921	3.804	−5.483	0.000	0.067	0.004	0.000
23	0.5884	0.1443	0.2734	0.5638	−1.929	−1.934	3.503	−5.321	0.000	0.056	0.006	0.000

. The meteorological effects of every environmental factor varied with the time of the day, differing during the daytime and at night.

A couple of environmental factors were meteorologically significant during the day. The NDWI showed more significant effects during the afternoon (15:00, R² = 0.5012, p < 0.05). However, its impacts were reduced continuously after this, and little thermal influence emerged in the nighttime. This conflicted with the NDVI. During the day, the linear correlation between the NDVI and hourly T_a was not as strong as that at night. Midnight (0:00 and 23:00) witnessed the strongest statistical relationships (R² > 0.57, p < 0.00) between the NDVI and T_a. Despite being less important at noon (R² < 0.2 from 13:00 to 16:00), the NDVI showed high importance for most of the day. In other words, vegetation might be climatically influential throughout the day. The varying significance of the NDBI and the altitude was similar to that of the NDVI, and the highest R² values were seen at around 23:00 (nearly 0.3 and 0.56).

The R² of the four parameters fluctuated significantly all day. They were compared, as illustrated in Figure 15. Generally, the NDVI and the altitude had higher R² values, which were 0.41 and 0.43, respectively, on average. The mean values for the NDWI (0.26) and NDBI (0.19) were lower. This means that the local meteorology of Chengdu is significantly influenced by vegetation and the altitude.

The hourly significance of all four physical parameters is expressed in Figure 16. They fluctuated over time regularly. The NDVI, NDBI, and altitude presented similar regularity. They were thermally effective in the evening, which was shown by the lower daytime R² values. Their values peaked in the middle of the night (nearly 0.6, 0.3, and 0.57, respectively). In contrast, the NDWI had little meteorological effect in the evening; however, it had an impact on cooling in the afternoon (0.4 to 0.5 between 12:00 and 18:00).

Further, since the daily weather changes over time, this study assumed a certain statistical correlation between the hourly meteorology and R² values. Hence, the R² of each variable was linearly associated with the hourly weather index. Figure 17 illustrates the linear relationship between the R² and hourly mean T_a. There were three parameters in which the R² was negatively correlated with the mean T_a, namely the NDVI (R² = 0.6928), NDBI (R² = 0.7823), and altitude (R² = 0.1608). The thermal effects of vegetation, buildings, and the altitude were relatively significant in cool periods. Nevertheless, the NDBI had a relatively low R² for the full day. Moreover, the negative correlation between the R² and the altitude was insignificant (R² = 0.1608). Therefore, the current weather was more significantly influential on the meteorological effects of vegetation. The R² of the NDWI (R² = 0.7217) was positively correlated with the hourly mean T_a. This suggests that the cooling effect of water bodies is more pronounced and climatologically more important at higher temperatures.

5. Discussion

This study has explored the impact of urban green, blue, and gray infrastructure on the urban heat island at different times of the day, especially on the cooling process in urban suburbs. Urban temperature variations depend on the properties of the biophysical components of the surface [75], which means that different land cover types can be associated with the thermal environment in different ways [76]. As mentioned in Section 2.4, the normalized indices can provide quantitative references for the green, blue, and gray urban infrastructure at the urban scale. Sundborg interpreted the differences in the urban climate as a function of the urban energy balance, where different urban climate influences are involved in the process of the urban energy balance, and an imbalance in energy exchange will eventually lead to the urban heat island phenomenon [77]. This study further showed the roles of different factors in influencing climate changes over time and altering the urban energy balance. In suburban areas, urban green infrastructure, urban gray infrastructure, and the altitude were more significant at night, while urban blue infrastructure had a stronger impact in the late afternoon. These phenomena resulted from diverse causes; they are discussed as follows.

5.1. Discussion of the Impact of Each Factor

Urban green infrastructure, indicated by the NDVI, includes parks, street trees, green roofs, and other vegetated sites [78]. Areas with this element are effective for cooling in various aspects. The canopying effects of plants obstruct sunlight radiation through absorption and reflection [79], which reduces heat radiation on canopied spaces. Generally, larger vegetation canopies and richer vegetation species can improve shading and alter the urban surface roughness, enhancing the convective heat exchange efficiency and the cooling capacity of plants [80,81]. In addition, plants can regulate the ambient temperature by increasing the latent heat through stomatal transpiration [79], but, in warmer climates, many plants will undergo “C4” photosynthetic metabolism, i.e., keeping stomata closed during the day and opening them at night. In this situation, transpiration and cooling will almost disappear during the daytime. Nevertheless, under warm conditions, plants are more efficient in photosynthesis. They open the stomata for transpiration during the evening and close them during the daytime [82]. There is little biological effect during the daytime. Further, the climate is influential in terms of biological effects. In continuous warm and/or drought weather, plants keep the stomata closed to reduce water and moisture loss. As a result, they have insignificant daily cooling effects. Overall, vegetation is able to adjust the microclimate through both physical shading and biological effects. They are meteorologically effective throughout the day [83]. Overall, green infrastructure is the only factor that has a significant cooling effect throughout the day due to a combination of physical shading and biological transpiration mechanisms [83], and it has gained a strong nighttime cooling capacity in the suburbs of Chengdu through its biological qualities [84].

Urban blue infrastructure refers to various water surfaces, including rivers, lakes, and the sea [85]. They achieve effective cooling due to the physical properties of water surfaces, which cause heat exchange with surface environments. This results from the high specific heat capacity and enthalpy characteristics of water [79]. First, the water converts the absorbed sensible heat into latent heat and releases water vapor through evaporation, a process that exhibits a clear diurnal cycle [86]. The water gradually warms up in the morning by absorbing solar radiation energy and starts to evaporate, and then the surface temperature of the water body and the vapor pressure difference between the water and the air reach the highest point in the afternoon, when the evaporation produced by evapotranspiration reaches the most extreme value of the day. At the same time, the cooling capacity of the water body is at the highest peak during the diurnal cycle [87]. Secondly, the difference in the specific heat capacity between land and water further contributes to the diurnal difference in water cooling [88]. During the daytime, solar radiation contributes to the rapid warming of the surface with a low specific heat capacity, so that the air rises from the land to the water surface to form circulation, and the airflow is blown from the water to the land. During the nighttime, as the land cools down more rapidly than the water, the airflow is blown from the land to the water, and the different wind directions will significantly affect the ambient temperature of the surrounding environment [89]. Therefore, blue infrastructure can effectively absorb long-wave radiation to cool the surface and consume short-wave radiation through evaporation. It realizes energy transfer by conduction, convection, and advection to regulate the ambient temperature, and the extreme value of the cooling intensity occurs in the hot afternoon [87]. Blue infrastructure plays a major role in daytime cooling in Chengdu’s suburbs, where water flows are plentiful.

Urban gray infrastructure is one of the main factors causing the surface temperature to increase because it is mainly covered by impervious surfaces, so the sensible heat exchange is significant [90]. In this study, the NDBI had a significantly greater effect on the temperature rise at night than during the day (Figure 16), which is a good indication that the shading effect of buildings plays a role in reducing the temperature in the suburban areas of the city, where the building density is not high. However, the influencing role of the NDBI in this study was smaller than those of other influencing factors, which is in contrast to previous studies in which gray infrastructure was the main factor involved in warming [26,91], and this phenomenon may be related to the location of each meteorological observatory. Firstly, Chengdu has weak solar radiation all year round, a small temperature difference between morning and evening, and high humidity, which makes the warming effect of building surfaces during the daytime less obvious than that in cities with high solar radiation [18]. Secondly, the location of the meteorological station in the study was near abundant trees and fewer buildings, which also weakened the warming effect of buildings. This suggested that the warming effect of buildings is weaker during the day than at night in the suburbs of Chengdu, which may be related to the fact that a moderate number of buildings have a shading effect during the day. The geographical factor, altitude, showed an important microclimatic impact during the evening. However, as an unchangeable factor, urban designers are unable to use this for thermal comfort adjustment.

5.2. Limitations and Further Work

The present study has revealed the effects of urban green, blue, and gray infrastructure on the urban heat island from a comprehensive perspective based on the urban scale. It is clarified that the NDVI, NWDI, and NDBI can effectively characterize different land cover types, in which the NDBI has a lower impact on the suburban climate than the NDVI and NDWI, which is different between the suburbs and cities. In addition, the study used meteorological station data to further clarify that the impact of relevant environmental factors on the urban–suburban climate continues to change with time and under weather changes.

This study also proposed a research method that combines meteorology station observation data and remote sensing image data for the study of the large-scale urban heat island and urban cooling. This method compensates for the deficiencies in the spatial and temporal continuity of the traditional independent meteorology station observation data and remote sensing image data and provides new ideas and methods for future urban-scale climate studies.

However, the study had limitations in sample data acquisition, such as the inadequate representation of the environmental diversity surrounding the individual meteorological stations in a single city, which may have impeded the accurate depiction of the gray infrastructure’s impact on environmental interventions. Furthermore, this study was limited to July and August in 2019 and 2020, thereby constraining the generalizability of its findings.

In future investigations, it is recommended to obtain data from a broader range of meteorological stations with diverse environmental characteristics for further research purposes. Additionally, extending the temporal scope of the study would be beneficial.

6. Conclusions

This study was based on remote sensing data and meteorological station data at the periphery of the core area of a city in Southwest China, exploring the impact of various urban environmental factors on the urban climate. Comprehensive analyses of the meteorological station and remote sensing data revealed that the climate at each location was influenced by a variety of environmental factors and that their effects varied according to the time of the day. The main findings are listed as follows.

• Urban LST data from RMSS were found to be significantly and positively correlated with T_a from meteorological stations (R² = 0.672, Sig < 0.005).
• Urban green, blue, and gray infrastructure were key elements affecting the urban climate. Green (R² = 0.4283) and blue (R² = 0.3251) infrastructure were effective for cooling, while gray infrastructure (R² = 0.3638) contributed to warming. This could be explained by their correlation trends (negative or positive).
• Environmental factors had different influences on the urban microclimate due to the daily time differences, e.g., the NDVI and NDWI were the most significant at 23:00 and 15:00, respectively.
• The role of environmental factors in influencing the climate at the periphery of the city was affected by weather variations, with the NDVI (R² = 0.6928) and NDBI (R² = 0.7823) decreasing gradually with the temperature increase, and the NDWI (R² = 0.7217) showed the opposite.

This study provides an examination of intervention models for some of the main factors affecting the urban climate regarding spatial data from the periphery of the urban core. The results show that the effects of different influences vary with time and with the weather conditions. This research’s results will provide a data reference and theoretical support for the construction of urban space strategies based on the concept of urban and social sustainability.