Spatiotemporal Impact of Urbanization on Urban Heat Island and Urban Thermal Field Variance Index of Tianjin City, China

The rapid infrastructure development in densely populated areas has had several negative impacts. Increases in urbanization have led to increased LST, and urban ecological systems have been negatively affected. Urban heat islands (UHIs) can be mitigated by understanding how current and future LST phenomena are linked to changes in landscape composition and land use cover (LUC). This study investigated the multi-scale spatial analysis of LUC and LST in Tianjin using remote sensing and GIS data. We used Landsat data from 2005 to 2020 to examine the effects of LUC on LST in urban agglomeration. According to the Urban Thermal Field Variance Index (UTFVI), the city’s ecological evaluation was carried out. Results show that changes in LUC and other anthropogenic activities affect the spatial distribution of LST. For the study years (2004–2009), the estimated mean LST in Tianjin was 25.32 ◦C, 26.73 ◦C, 27.62 ◦C, and 27.93 ◦C. Between LST and urban areas with other infrastructures, and NDBI, significant positive correlation values were found about 0.53, 0.48, and 0.76 (p < 0.05), respectively. Temperatures would almost certainly increase by 3.87 ◦C to 7.26 ◦C as a result of decreased plant cover and increased settlements. These findings strongly imply a correlation between LST and the vegetation index. Between 2005 and 2020, the anticipated increase in LST of 3.39 ◦C is expected to harm urban environmental health. This study demonstrates how Tianjin and other cities can achieve ecological sustainability.


Introduction
Land use cover (LUC) change occurs primarily due to the alteration of natural vegetation at various scales due to logging, vast urban development, and agricultural expansion activities [1]. These changes contribute to various environmental problems, including biodiversity loss, increased greenhouse gas concentrations [2][3][4][5][6], fluctuation in land surface temperature (LST), and favored precipitation anomaly. Urbanization, associated with population growth, industrial/infrastructural development, and landscape composition, has become a global challenge. As a result of urbanization, LUC changes are one of the major causes of fluctuating LST in urban areas [6][7][8][9]. As a result of LST's adverse effects on the urban environment, the relationship between LUC and LST is gaining significant attention within the scientific community [10,11]. As the second-largest economy globally, China is experiencing rapid urbanization, with a dramatic expansion of built-up areas, residential neighborhoods, and new cities [12][13][14]. As a result, appropriate urbanization policies to modify land-use and landscape patterns are critical [15][16][17]. Understanding the relationship between urbanization and these patterns may help inform the development of more sustainable urban environmental management policies [14,[18][19][20]. The vast majority of people who relocate to new cities and their activities put a lot of strain on housing, infrastructure, food security, ecosystem services, and natural resources management [7,14,17,[21][22][23]].

Study Area
Tianjin, China, is a global beta second-tier city with over 13.86 million people [57] and an urban built-up area of approximately 1156 km 2 . It is located at 39°N and 117°E. (Figure 1). Tianjin has undergone long-term evolution in urban spaces such as other world metropolises, particularly in replacing its primary, secondary, and tertiary industries over the last twenty-five years. Changes in urban development from a single center to multiple centers [38]; and significant spatial differences in factors such as UHI-related green coverage, building height, building density, hardened pavements, and sky wideness [15,58]. Thermal comfort levels have decreased due to these changes. We chose Tianjin as an example of a mega-urban space for this study.

Data Source and Pre-Processing
For LST and LUC estimation, a series of remote-sensed images with a spatial resolution of 30 m were collected from Landsat-5 (Landsat Thematic Mapper-TM) and Landsat 8 (Operation Land Imager-OLI) over various periods between 2005 and 2020 [4,30,38,[59][60][61]. The primary objective was to use Google Earth Engine to select cloud-free images acquired during the same season of each sample year (Table 1). (GEE). Before LUC identification, a pre-treatment procedure was used to extract atmospheric effects from Landsat images [5,62,63]. For each land cover class, 500 ground control points were collected during a ground survey to assess the accuracy of the thematic layer. The US Geological Survey's web data portal provided the digital elevation model (DEM) data. Each Landsat scene has been enhanced using the histogram equalization technique to increase the image contrast [40,64].

Data Source and Pre-Processing
For LST and LUC estimation, a series of remote-sensed images with a spatial resolution of 30 m were collected from Landsat-5 (Landsat Thematic Mapper-TM) and Landsat 8 (Operation Land Imager-OLI) over various periods between 2005 and 2020 [4,30,38,[59][60][61]. The primary objective was to use Google Earth Engine to select cloud-free images acquired during the same season of each sample year (Table 1). (GEE). Before LUC identification, a pre-treatment procedure was used to extract atmospheric effects from Landsat images [5,62,63]. For each land cover class, 500 ground control points were collected during a ground survey to assess the accuracy of the thematic layer. The US Geological Survey's web data portal provided the digital elevation model (DEM) data. Each Landsat scene has been enhanced using the histogram equalization technique to increase the image contrast [40,64].

Experimental Design
The data sets were processed in Google Earth Engine (GEE) according to the specified methodology ( Figure 2) to create a false color composite (FCC). The following procedure involved determining the presence of an urban heat island (UHI) within the city to figure out the Normalized Difference Vegetation Index (NDVI), the Normalized Difference Vegetation Index (NDBI), and the land surface temperature (LST) [3,4,38,63,65]. The supervised image classification method was then used to classify each scene. Pearson correlation analysis was used to derive statistical inferences from the mean LST and the relative percentages of various land cover types, as well as the NDVI and NDBI values [34,56,66,67]. The Ecological Evaluation Index (EEI) for Tianjin City was mapped using UTFVI. ArcGIS Pro-2.5, Adobe illustrator, and R-studio generated all spatial maps and statistical illustrations.

Experimental Design
The data sets were processed in Google Earth Engine (GEE) according to the specified methodology ( Figure 2) to create a false color composite (FCC). The following procedure involved determining the presence of an urban heat island (UHI) within the city to figure out the Normalized Difference Vegetation Index (NDVI), the Normalized Difference Vegetation Index (NDBI), and the land surface temperature (LST) [3,4,38,63,65]. The supervised image classification method was then used to classify each scene. Pearson correlation analysis was used to derive statistical inferences from the mean LST and the relative percentages of various land cover types, as well as the NDVI and NDBI values [34,56,66,67]. The Ecological Evaluation Index (EEI) for Tianjin City was mapped using UTFVI. ArcGIS Pro-2.5, Adobe illustrator, and R-studio generated all spatial maps and statistical illustrations.

Computation of Land Use Cover Change (LUCC)
This study's land cover classification data was generated using an advanced supervised machine learning classification algorithm called random forest (RF) using 10-20 multiple DTs. RF is used to process spectral images in remote sensing-related fields [6,55]. Five different Landsat satellite images from the summer period (May-September) 2005 to 2020 were used to classify each reference year. Between May-September 2005 and 2010, Landsat-5 TM multispectral images were acquired, and Landsat-8 OLI multispectral images were developed in 2015 and 2020 [7,8,11,21]. For all spatial scenes, an image classifier was used to classify the LUC types into five classes: urban, water, cropland, forest, lowland [34,58,60]. A confusion matrix was generated from the classified image and signature

Computation of Land Use Cover Change (LUCC)
This study's land cover classification data was generated using an advanced supervised machine learning classification algorithm called random forest (RF) using 10-20 multiple DTs. RF is used to process spectral images in remote sensing-related fields [6,55]. Five different Landsat satellite images from the summer period (May-September) 2005 to 2020 were used to classify each reference year. Between May-September 2005 and 2010, Landsat-5 TM multispectral images were acquired, and Landsat-8 OLI multispectral images were developed in 2015 and 2020 [7,8,11,21]. For all spatial scenes, an image classifier was used to classify the LUC types into five classes: urban, water, cropland, forest, lowland [34,58,60]. A confusion matrix was generated from the classified image and signature file to evaluate accuracy. The image classification was carried out using the GEE platform's code editor.

NDVI and NDBI Calculation
The Normalized Difference Vegetation Index (NDVI) is the most frequently used vegetation index worldwide. Equation (1) is commonly used to verify the condition of vegetation by calculating the ratio of near-infrared (NIR) and red bands [6,28,53]. NDVI values range between −1 and +1, depending on the objects visible at ground level. Typically, green vegetation has a value between 0.2 and 0.8 [28,59]. where NIR corresponds to Band 4 (Landsat TM) and Band 5 (Landsat 8) and RED corresponds to Band 3 (Landsat TM) and Band 4 (for Landsat 8).
Normalized Difference Build-up Index (NDBI) can be calculated using the following formula (2) [68]. NDBI values range from −1 to +1. A negative value of NDBI corresponds to water bodies, whereas a positive value corresponds to developed areas. Vegetation has a low NDBI value [10,28,68,69].

Retrieval of Land Surface Temperature (LST)
LST can be derived from Landsat data with a thermal infrared band (Band-6) that has been radiometrically corrected. There is a procedure for converting digital numbers (DNs) at the satellite level that takes atmospheric absorption, re-emissive heat, and a collection of surface emission corrections into account [50]. The reflectance and brightness temperature of the top of the atmosphere (TOA) should be converted from the spectral radiance [9,22,29] and then calculated land surface temperature (LST) with the following equation (Equation (3)) [4,29,48] where, brightness temperature T B , λ is the effective wavelength (10.9 mm for band 10 in Landsat 8 data), σ is Boltzman constant (1.38 × 10 −23 J/K), h is Plank's constant (6.626 × 10 −34 Js), c is the velocity of light in a vacuum (2.998 × 10 +8 m/sec), and ε is emissivity. For this study, land surface temperature (LST) is expressed in degrees Celsius ( • C).
T is the LST raster value, while T min is the minimum temperature value.

The Urban Thermal Field Variance Index (UTFVI)
Tianjin City's urban heat island (UHI) was also studied using the UTFVI to understand its effects better.
The LST value is directly related to the amount of heat generated [4,65]. Thermal comfort indices can be used to assess the impact of urban heat islands (UHI) on the quality of life in the city. The UHI zones of Tianjin were evaluated for their environmental impact using UTFVI. Which was calculated using the following equation (Equation (5)) [15,44,61,71,72]: where, UTFVI is the Urban Thermal Field Variance Index; T s = LST ( • C); and T mean mean LST ( • C).

Accuracy Assessment of LUC
These land-use classifications' Kappa coefficient was calculated about 0.91%, 0.96%, 0.84% and 0.97%, respectively. For all four classified images, the kappa coefficient value was greater than 0.85, indicating the accuracy of the classification estimations and the reliability of the RF algorithm utilized in this investigation.    (Table 3).

Accuracy Assessment of LUC
These land-use classifications' Kappa coefficient was calculated about 0.91%, 0.96%, 0.84% and 0.97%, respectively. For all four classified images, the kappa coefficient value was greater than 0.85, indicating the accuracy of the classification estimations and the reliability of the RF algorithm utilized in this investigation.   Figure 4A). We categories Normalize Difference of Vegetation Index (NDVI) into four different classes (i) <0.25; (ii) 0.26-0.45; (iii) 0.46-0.59; and (iv) >0.60 to evaluate the vegetation index in the study area ( Figure 4A).   (Table 5). We classify the Normalized Difference of Build-up Index (NDBI) into four different categories ( Figure 4B); the total area in the category <−0.99 was estimated to be 32

Determination of Urban Heat Island (UHI)
Urban heat island (UHI) was categorized into five various categories: (i) no UHI, (ii) low, (iii) moderate, (iv) high, and (v) very high ( Figure 5A) for 2005 to 2020. No UHI area covered 294.29 km 2 which was 3.36%, and low UHI area was 2699.61 km 2 , moderate UHI area was 3710.06 km 2 , area covered in high UHI was 40.95 km 2 , and tiny patches were observed in highly high UHI in 2005. (Table 6). From 2005 to 2020, the non and low UHI area increased up to 12% and 10%, covering almost 1715.98 km 2 and 2662. 85  We also observed the severe changes in good and normal were 844.18 km 2 and 470.49 km 2 , which was 97% and 54% increment in these indices during the period 2005-2020. the worse and worst condition area grew by 421.35 km 2 (49%) and 360.78 km 2 (42%), respectively. During 2005 and 2020, Bad index area showed a massive decrease of about 1228.83 km 2 . Still, it turned to the most critical condition. It increased the proportion in the study region's worse and worst condition area, which shows the extreme heat condition due to rapid urbanization and development.   Similarly, high UHI and very high/extremely high UHI category areas were increased by 124.09 km 2 and 93.98 km 2 , representing an accumulative change of 0.8% and 0.63%, respectively, during 2005-2020.

Evaluation of Urban Thermal Field Variance Index (UTFVI)
UTFVI for the area was divided into five different categories according to the ecological indexing as (i) good, (ii) normal, (iii) bad, (iv) worse, and (v) worst in the period 2005 to 2020 ( Figure 5B). The area with the good and normal condition was 1166.09 km 2 and 4891.97 km 2 , 9.98% and 41.89% in 2005, and bad index area was 4557.55 km 2 (39.02%). The worse and worst situation observed were 743.68 km 2 (6.37%) and 320.00 km 2 (2.74%). In 2020, the good index area covered 2010.27 km 2 (16.02%), the normal area was 5362.46 km 2 (42.74%), the bad index area was 3328.72 km 2 (26.53%), and the area covered in worse condition was 1165.04 km 2 (9.29%). The most critical condition area was observed in the worst index, about 680.78 km 2 which was 5.43% of the whole study area. (Table 7). We also observed the severe changes in good and normal were 844.18 km 2 and 470.49 km 2 , which was 97% and 54% increment in these indices during the period 2005-2020. the worse and worst condition area grew by 421.35 km 2 (49%) and 360.78 km 2 (42%), respectively. During 2005 and 2020, Bad index area showed a massive decrease of about 1228.83 km 2 . Still, it turned to the most critical condition. It increased the proportion in the study region's worse and worst condition area, which shows the extreme heat condition due to rapid urbanization and development.

Statistical Analysis
Pearson's Correlation Analysis is used to determine the relationship between all LUC variables and different LST indices. The NDVI had a negative and statistically insignificant relationship with the LST. At the same time, the LST association between NDBI was positive and negligible. There was a strong positive significant correlation between urban areas, NDBI, and low land with LST ( Figure 6). This measure was also found a negative and insignificant relationship with water, cropland, and Forest.

Statistical Analysis
Pearson's Correlation Analysis is used to determine the relationship between all LUC variables and different LST indices. The NDVI had a negative and statistically insignificant relationship with the LST. At the same time, the LST association between NDBI was positive and negligible. There was a strong positive significant correlation between urban areas, NDBI, and low land with LST ( Figure 6). This measure was also found a negative and insignificant relationship with water, cropland, and Forest.
Similarly, the UHI and UTFVI had a significant positive association with urban Areas, NDBI, and low land, while the negative and insignificant association with water, cropland, Forest, and NDVI, including urban vegetation. The Pearson correlation coefficient confirms the positive and robust relationship between urban/constructed areas with LST, UHIs, and UTFVIs. As a result, it could be a hot climate in a metropolitan area if more infrastructure, highways, businesses, and industrial zones are built. Agricultural/cropland has a positive and significant relationship, while lowland has a negative and insignificant relationship with the LST.

Discussion
The objective of this study was to summarize the various factors that influence UHI density. A review of the elements was then carried out to identify the pros and cons. It was also concluded that planning and designing for UHI reduction are interrelated. Due Similarly, the UHI and UTFVI had a significant positive association with urban Areas, NDBI, and low land, while the negative and insignificant association with water, cropland, Forest, and NDVI, including urban vegetation. The Pearson correlation coefficient confirms the positive and robust relationship between urban/constructed areas with LST, UHIs, and UTFVIs. As a result, it could be a hot climate in a metropolitan area if more infrastructure, highways, businesses, and industrial zones are built. Agricultural/cropland has a positive and significant relationship, while lowland has a negative and insignificant relationship with the LST.

Discussion
The objective of this study was to summarize the various factors that influence UHI density. A review of the elements was then carried out to identify the pros and cons. It was also concluded that planning and designing for UHI reduction are interrelated. Due to the rapid growth of Tianjin city's urban population and industrialization, land use changes dramatically. The climate also changes due to the presence of more people. As a result of unregulated development and anthropogenic heat, the density of UHI increases. This study established that planning and designing for reduced UHI might be accomplished by modifying the features of the surrounding environment [4,[73][74][75]. Urbanization and industrialization in Tianjin produce higher temperatures in the city. Natural processes do not cause this phenomenon but result from human actions.

LUC, Impervious & Green Space versus Temperature
Land use cover (LUC) classification can assist in estimating the relationship between LST and a variety of land use indices [30,72,76]. Additionally, macro/anthropogenic climate [9], meso-and oscillating climate, the land-use landscape, urban expansion [22,77,78], spatial transitions, human pressures on the ecosystem [3,75,79,80], energy transition, general macro-and urbanized environments, and socioeconomic and biophysical conditions may all be associated with the UHI [9,15,75,81]. From 2005 to 2020, this study discovered a highly discordant and out-of-the-ordinary trend in LUC change. The rapid urbanization may explain this abrupt change in LUC. The extensive deforestation and cropland destruction associated with various development projects (housing and industrial development) may have resulted in a decline in total vegetation cover [5,58,65,73]. This rapid loss of vegetation cover (VC) has far-reaching implications for reducing natural cooling effects due to shading and plant evapotranspiration [34,40,48]. This study discovered a negative correlation between NDVI and LST and the presence of VC and its impact on UHI. Transpiration and evaporation of plants may become extinct as a result [12,16,76]. Grasslands and ornamental plants have a lower impact on LST reduction than other types of vegetation, such as forest/urban treebanks and gardens, according to earlier research [32,48]. Our findings indicate that LUC significantly affects LST in the urban environment, particularly in urban areas. The effect varies according to the proportion of green space, blue space (constructed areas), and water bodies [35,73], indicating that vegetation contributes to the city's cooling. Since buildings, streets, concrete and asphalt roads, and other components of the landscape [25,82], built-up areas contribute significantly to urban heat fluxes. Since natural evaporation from water surfaces helps cool the surrounding air, lowering overall air temperature in the region, the study's findings also show a positive linear relationship between LST and the built-up area [65,83,84]. This study found that UHI was not good for the NDVI [12,48,65]. Changes have been made to the city's downtown area and Tianjin's growth as a city. Many lands in these areas have been turned around by the government or private businesses to build new homes, businesses, and factories on top of it [25,53,83]. In certain situations, new industries were erected alongside secondary infrastructure on well-maintained cropland/forest areas, lowering the number of cropland and forest areas while raising LSTs in such places [83,85]. These areas were supposed to function as a buffer between urban and rural regions, absorbing surplus heat from vehicles and industry [11,52].
According to the findings, urban expansion is a key contributor to LST, which causes uneven thermal fluxes in metropolitan regions. This radiation/heat exchange is thought to be an essential contribution to UHI and to urban climate change. As a result, it can be stated that LUC has a significant impact on a location's surface radiant temperature and that human-induced LUC change is the key mechanism influencing LST rises in the urban micro-atmosphere.

LUC, Impervious & Green Space Vs. UTFVI
Spatial fluctuations in LST are significant climatic variables employed by remote sensing data to estimate the urban thermal environment using UTVFI. It has been shown that growing population density and transitions in the city's urban climate, notably in terms of urban development and vegetation cover loss, are the significant causes of oscillations in urban thermal conditions [4,7,65,77]. Land-use change is the driving factor of environmental deterioration and changes in urban EEI values in its aggregate spatiotemporal evaluation of LUC, NDVI, BUI, and UTFVI [28,70,74]. According to quantitative research, the increase in urban temperatures and changes in the urban and regional climate result from fast urban/built-up area expansion and various building materials and other key structures in city development [78].

Current Research Trends in UHI Mitigation and Adaptation
The world's natural resources are depleting yearly, while temperatures are rising. Most scientists are now attempting to avoid the production of this so that people can live in more pleasant environments [8,38,58]. The open and green surfaces are the most significant regulating element in decreasing UHIs. It is important to note that green areas do not have to be of a specific type. Green roofs roof gardens may now contribute to the urban environment [4,10,23].
Similarly, the location and shape of settlements are significant considerations for urban heat. To lessen urban heat, buildings must be placed in such a manner that they do not obstruct the breeze [43,58,69,82]. The study's findings will be shared with appropriate institutions/policymakers, and LUC Layout may be recommended for Tianjin. For example, additional green space should be added to metropolitan areas. According to the study's findings, smaller green patches are calmer than larger green ones. As a result, green zones should be developed in cities by constructing parks and tiny plant communities, effectively countering the rise in surface temperatures [12,18,31,36,40]. The critical issue that must be addressed in Tianjin's rural areas is preserving existing green spaces. Agricultural policy should be reviewed for Tianjin and its surroundings, and agricultural development should not be permitted in places with existing vegetation. High LSTs can be avoided by increasing green space in rural regions. Furthermore, any new settlements should be developed on impermeable or unoccupied land rather than green space. Green spaces in cities can bring long-term advantages if specific criteria are followed.

Conclusions
Environmental scientists and urban planners are currently concerned about the impact of the LST surge in urban areas, as it can cause serious health problems. According to scientific evidence, rapid growth raises the Earth's average surface temperature. This study looked into the issues of urban heat evaluation and land-use change in Tianjin, China. The UTFVI Index and Google Earth Engine (GEE), a web-based remote sensing tool, were used to analyze the region's thermal pattern on Landsat Time Series data estimated for LST's spatial distribution. The findings show that urbanization changed the spatial distribution of LST between 2005 and 2020 and that temperatures in Tianjin's central region soared. Simultaneously, temperatures were relatively low in areas far from densely populated areas. The city's eco-environmental changes were induced by changes in vegetation cover from 2005 to 2020, as demonstrated by the lowering NDVI trend. As a result, Tianjin was discovered to have the most severe urban heat island phenomenon and the worst ecological state, indicating that a more acceptable city layout and improved urban planning regulations should be implemented as soon as possible. The city's Urban Thermal Field Variance Index was used to determine Tianjin's ecological status (UTFVI). The worst Environmental Evaluation Index (EEI) has occurred in highly dense urban areas, contributing to a degraded and thermal eco-environment over time. According to the study, remote sensing techniques are appropriate for evaluating UHI and non-UHI sites. Still, high-resolution imagery with the same atmospheric and surface conditions is challenging to obtain and select. Tianjin has become the most severe urban heat island as a result of this environmental issue. The Urban Thermal Field Variance Index was used to calculate the city's worst ecological index. It demonstrated how, throughout time, the increasing density of dense metropolitan areas has contributed to the development of deteriorated and heated ecosystems. The study's limitations are that it only uses ground-based data and cannot be utilized to determine UHI zones. Institutional Review Board Statement: Not applicable.

Informed Consent Statement: Not applicable.
Data Availability Statement: On request, the authors will provide the data from this study.

Acknowledgments:
The author wishes to express his appreciation and gratitude to the anonymous reviewers and editors for their insightful comments and suggestions to improve the paper's quality.

Conflicts of Interest:
The authors declare no conflict of interest.