research paper on urban heat island

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• My Account Login
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 21 October 2020

Development of a holistic urban heat island evaluation methodology

• Valentino Sangiorgio 1 ,
• Francesco Fiorito 1 , 2 &
• Mattheos Santamouris 2  

Scientific Reports volume  10 , Article number:  17913 ( 2020 ) Cite this article

11k Accesses

45 Citations

30 Altmetric

Metrics details

• Atmospheric science
• Climate change
• Environmental impact
• Natural hazards

Urban Heat Island (UHI) phenomenon concerns the development of higher ambient temperatures in urban districts compared to the surrounding rural areas. Several studies investigated the influence of individual parameters in the UHI phenomenon, on the other hand, an exhaustive study that quantifies the influence of each parameter in the resulting UHI is missing in the related literature. This paper proposes a new index aimed at quantifying the hazard of the absolute maximum UHI intensity in urban districts during the Summer season by taking all the parameters influencing the phenomenon into account. In addition, for the first time, the influence of each parameter has been quantified. City albedo and the presence of greenery represent the most important characteristics with an influence of 29% and 21%. Population density, width of streets, canyon orientation and building height has a medium influence of 12%, 10%, 9% and 8% respectively. The remaining parameters have an overall influence of 11%. These results are achieved by exploiting three synergistically related techniques: the Analytic Hierarchy Processes to analyse the parameters involved in the UHI phenomenon; a state-of-the-art technique to acquire a large set of data; and an optimization procedure involving a involving a Jackknife resampling approach to calibrate the index by exploiting the effective UHI intensity measured in a total of 41 urban districts and 35 European Cities.

Similar content being viewed by others

Identification of reliable locations for wind power generation through a global analysis of wind droughts

Enrico G. A. Antonini, Edgar Virgüez, … Ken Caldeira

Comparing the carbon footprints of urban and conventional agriculture

Jason K. Hawes, Benjamin P. Goldstein, … Nevin Cohen

Public perceptions and support of climate intervention technologies across the Global North and Global South

Chad M. Baum, Livia Fritz, … Benjamin K. Sovacool

Introduction

Global warming and the associated rise in extreme temperatures substantially increase the possibility of heat waves jeopardising the safety of some vulnerable population classes. In this context, the Urban Heat Island (UHI) phenomenon aggravates warm temperatures in urban districts together with the related risks 1 . To this aim, the scientific and technical communities are interested in studying the phenomenon in depth, in order to obtain new procedures to quantify the potential Urban Heat Island Intensity (UHII) and predict the hazard. The hazard of the phenomenon is connected to many classes of parameters including: (1) meteorological variables such as synoptic weather and climate conditions, (2) characteristics of the city , intended as urban layout and materials' characteristics, (3) anthropogenic heat , related to population density (4) and city canyons whose influence depend on urban layout. In addition, the measured UHII depends on the selection of the reference rural measuring station and in particular on the characteristics of the urban districts where the station is located 2 .

Several authors demonstrated that UHI has a serious impact on safety of some vulnerable population classes, and buildings energy consumption principally during the Summer period 3 . Indeed, the phenomenon highly increases the cooling energy consumption and the corresponding peak electricity demand of the cities 4 . Consequently, the UHI can be associated with an important increase of urban pollutants concentration, it is related with the tropospheric ozone 5 and the city's carbon footprint 6 . Finally, the phenomenon seriously affects comfort, health and increases mortality problems 7 , 8 , 9 .

In the building sector, scientist and technicians are interested in developing new techniques and procedures to evaluate the average UHII, and maximum UHII in order to forecast and mitigate this effect, decrease energy demand of building stock and eradicate the energy poverty 10 . This objective is framed within the European and global challenge of innovation for the built environment to assumes a minimization of the energy consumption of buildings and mitigate of the urban heat island and the local climate change.

On the other hand, the classical evaluation of the absolute maximum UHII can be complex and resource consuming. In addition, data acquisition strategy varies considerably among the reported studies making complex an interstudy comparison. In particular, the necessary data acquisition presents critical points: (1) high duration of the experimental phase; (2) high number of measuring stations used; (3) data acquirement depending on the seasons; (4) difficulty in finding historical data to improve the efficacy of monitoring phase in order to observe and support climatological studies of local climate variations 4 , 11 . In this context, an index calibrated to forecast the absolute maximum UHII phenomenon would be of great support for technicians of city government and for civil protection in order to analyse the territory and define mitigation strategies.

In the related literature there are some attempts to obtain a concise equation to associate some parameters related to the urban districts with the increase of the heat island effect. The first attempt regards the dependencies of UHII with the immediate environment of the measurement site such as the ratio between building height and width of street (H/W) 12 . Successively, more in dept investigation focused on the sky view factor (SVF) as discussed in the review study of Unger 13 . The ratio H/W and the SVF sky view factor have the advantages to be very easy to be used by practitioners. On the other hand, H/W and SVF efficiency is local and does not include other important factors. Indeed, in recent studies, the challenge is integrating other types of variables such as common surface types in cities or greenery. Because of the complex processes governing the urban climate, involving qualitative and quantitative data, the related literature exhibits few attempts to find an equation to estimate the potential UHI. In addition, typically such approaches must be calibrated for each new city and for this reason they have very limited applicability 14 , 15 , 16 . An effective attempt to involve multi parameters has been developed by Theeuwes et al. 17 . Unfortunately, the authors themselves specified that their approach is not able to consider important urban properties, such as building materials or anthropogenic heat. To sum up, a calibration of an index that manages to consider several parameters including meteorological variables, characteristics of the city , anthropogenic heat , and city canyons would be extremely useful in the scientific and technical community, but it has not yet been obtained in previous researches.

Among the existing methodologies to calibrate an intensity index, the Multi-Criteria Decision-Making (MCDM) approaches are particularly effective to consider both qualitative and quantitative data in the analysis 18 . In particular, the Analytic Hierarchy Processes (AHP) has been widely used in the literature to understand hazard and risk phenomena as specified in the review work of De Almeida et al. 19 . In addition, optimization procedures can be applied to calibrate the index and support the MCDM approaches 20 , 21 . To this aim, such methodology can be suitable in calibrating an index of potential absolute maximum UHII, but this approach has not yet been attempted in the related literature.

In this paper, the aims of the current study are twofold: for the first time an exhaustive analysis is performed to quantify the influence of parameters affecting the UHII during the Summer season. In addition, the work proposes the calibration of a novel index involving many parameters devoted to quantifying the potential absolute maximum Urban Heat Island Intensity ( I UHII ) in urban districts in order to predict the local hazard of the phenomenon.

This ambitious objective is achieved by employing four synergistically related techniques: (1) the AHP to analyse, structure the problem and define the index; (2) a large data acquisition process to obtain an exhaustive dataset; (3) an optimization procedure to calibrate the index based on a large data acquisition of 41 urban districts; (4) two validation test involving a Jackknife approach to identify the stability of the solution and the Absolute Error of the proposed model.

Influence of parameters affecting absolute max UHII

For the first time, we achieved the quantification of the influence of every parameter for the creation of the UHI phenomenon in urban districts. In particular, the weights of eleven parameters or criteria i (with i  = 1,…,11) are defined and calibrated by using an AHP and Optimization-based calibration.

The resulting influence is expressed in percentage and the description of results follows the parameters classification proposed in the introduction.

The most important parameters belong to the characteristics of the city class and in particular to the Land Cover Types: (1) “Albedo” ( i  = 5) represents the ability of urban districts surfaces to reflect solar radiation and has an influence of 29%; (2) the “Greenery” ( i  = 6) percentage presents an influence of 21%.

Another important class regards the presence of city canyons in urban districts. Indeed, the average “width of street” ( i  = 9), “building height” ( i  = 8) and the “canyon orientation” ( i  = 10) affect the phenomenon of 10%, 8% and 9% respectively, with a further increase of 5% for the “irregularity of the city” streets ( i  = 11).

The anthropogenic heat is a class that affects the phenomenon with an intermediate level. The “population density” ( i  = 7) is the only associated parameter of this class with 12% of influence.

Finally, the meteorological variables have the least influence of all the parameters: (1) “Clear sky days” ( i  = 4) with a value of 2%, (2) “Windless days” ( i  = 1) with 2%, (3) “Average max Summer temperature” ( i  = 2) and “average Summer thermal excursion” ( i  = 3) of 1% each one.

Figure  1 shows the influence of every parameters expressed in percentage and displayed in a pie chart.

Iinfluence of each parameter in the absolute max UHII phenomenon.

The novel index aimed at quantifying potential absolute max UHII in urban districts

We also developed a novel index I UHII to predict the potential UHII in urban districts based on the same eleven parameters, hereafter named criteria in accordance with AHP 22 . In addition, for each criterion i a set of intensity ranges j (with j  = 1,…, n i ) is defined to characterize its intensity levels.

Our results show that I UHII is easy to be applied by practitioners thanks to a simple and effective equation:

were v i and w ij are the weights associated to the criteria i and to the intensity ranges j respectively.

Suitable tabulated weights (Supplementary Table S1 ) are obtained to use the Eq. ( 1 ) where every single parameter (or criterion) is associated to a weight v i and every intensity ranges is associated to a specific weight w ij .

These tabulated weights are calibrated by exploiting a set of 41 urban districts ( \(Ud=1,\dots ,41\) ) for which the real max UHII is obtained from an exhaustive bibliographic analysis. A total of 35 European cities have been involved to contemplate different climate zone, different cities typologies in order to get the calibrated weights effective in different contexts.

Validation results

To validate the index, a comparison of the prediction (obtained with the index \({{\varvec{I}}}_{{\varvec{U}}{\varvec{H}}{\varvec{I}}{\varvec{I}}}\) ) with all the 41 independent values is carried out. In addition, a Jackknife approach (a one-by-one removal of the 41 urban districts) is used in order to evaluate the stability of the solution. This validation shows that the calibration is robust and, even if the input data changes due to the one-by-one removal, the results do not vary significantly. Moreover, Fig.  2 shows the statistical graph (Boxplot) of each parameter influence in the absolute max UHII phenomenon. In particular, the coloured boxes represent the distribution of the weights (expressed in percentage) and the black horizontal line inside the boxes denotes the median of the sample. While the box contains all the results within the 25th and 75th percentile of the population, the vertical dotted line contains all the results which are not considered outliers.

Boxplot of the influence of each parameter (or criterion) in the absolute max UHII phenomenon.

In addition, in order to explicitly prove the effectiveness of the proposed calibration by a common metrics, a second analysis identifies the Absolute Error and Relative Error of \({{\varvec{I}}}_{{\varvec{U}}{\varvec{H}}{\varvec{I}}{\varvec{I}}}\) in comparison with the effective absolute max UHII obtained from the bibliography. Also in this case, the Jackknife approach is used to perform a resampling of the input data and verify the robustness of the solution. Figure  3 shows the values of the absolute error and the mean relative error (represented in the same graph by the symbol “ × ”) obtained for every removal of the 41 urban districts. It is worth noting that the average values of both absolute and relative errors do not change significantly despite the resampling of input data, proving the robustness of the obtained results.

Analysis of the Absolute Error and Relative Error (average values represented by the symbol “ × ”) for each urban district removal with the Jackknife approach.

This additional analysis demonstrates that the proposed model is able to identify the potential absolute max UHII with an average accuracy of about 1 °C ( Mean Absolute Error  = 0.9 °C) and an average mean relative error less than fifteen percent ( Mean Relative Error  = 14.5%). In particular, in more than the 70% of the investigated urban district the Relative Error is less than 10%. On the other and, in the Urban district where the absolute max UHII is small, often there is a higher error (above 15%) even if the Absolute Error remain about 0.9 °C.

These obtained error can be considered acceptable to the effective application of the index. Indeed, it is in line with the reliability that characterizes the UHII measurements with standard fixed station. Furthermore, the classical approach to measure UHII can be affected by an inaccuracy due to the choice of the rural station measurement (to be compared with the city station measurement), as it is discussed by Oke et al. 2 .

A third analysis, based on the Pearson correlation coefficient (ρ), is used to verify the importance of all the considered parameters involved in the proposed AHP framework and consequently in the index I UHII . In particular, the Optimization-based calibration showed that the meteorological variables have the minor influence of all the parameters on the development of the absolute max UHII. On the other and, the four considered meteorological drivers are appropriately correlated and then are important to describe the phenomenon as emphasized by the Pearson correlation coefficient ( Windless days ρ  =  0.308 , Average max Summer temperature ρ  =  0.272 and Average Summer thermal excursion ρ  =  0.238 and Clear sky days ρ  =  0.240) .

This third analysis provides the scatter plot of every parameter (criterion i ) in relation with the absolute max UHII, together with the corresponding correlation coefficients. Supplementary Fig. S1 shows the complete overview of the relationships between variables and the investigated phenomenon. This analysis confirms that most of the involved parameters have a medium or high correlation and none of the involved parameters has very low and a null correlation. Hence, the related criteria can be considered important in the absolute max UHII manifestation.

In addition, in order to show how the proposed method assess the absolute max UHII in different cities, an additional scatter plot with observed values versus predicted values is showed in Supplementary Fig. S2 . The Pearson coefficient ( ρ  =  0.716 ) exhibits a good correlation providing a further information about the effectiveness of the calibration.

The present study provides a comprehensive analysis of all the parameters involved in the absolute max UHI phenomenon, an effective data collection procedure and a useful Index, calibrated and validated, to forecast the absolute max UHII in urban districts. This discussion focuses on the principal two contribution of the proposed work (the analysis of the involved parameters and the new index) in comparison with previous studies. Afterwards, the limitations are discussed, and the application potential of the proposed method are emphasized by showing the case of Milan (Italy).

The first contribution of this work regards an exhaustive analysis of the parameters involved in the absolute max UHII by classifying them in criteria, intensity ranges and variables. In the related literature these parameters have been often studied individually or by considering all the parameters associated with a specific “ class” : meteorological variables , characteristics of the city , anthropogenic heat , and city canyons .

To provide some examples, Zhao et al. 23 and Lokoshchenko 24 investigate the contributions of local meteorological variables in the UHI. In particular, Zhao et al. 23 focus on the influence of humid and dry climates with specific investigation on the precipitation effect. However, Lokoshchenko 24 investigates UHII changes over the past hundred years principally studying the correlation of the phenomenon with the urban dry island and including a connection with the growth of population density.

The characteristics of the city are principally investigated in studies involving the use of remote sensing to evaluate the UHII 25 , 26 . These studies confirm the importance of vegetation and albedo connected with the built environment and only marginally consider other classes of parameters including average air temperature and anthropogenic heat fluxes.

On the contrary in other researches , anthropogenic heat is a topic of great interest, and many studies are completely focused on the development of new approaches for the evaluation of heat flux database, with high spatial resolution 27 , 28 , 29 .

In addition, more local analysis are developed to study city canyons. To provide an example an exhaustive review about the sky view factor in city canyons is discussed in Unger 13 .

Such literature framework suggests that studies of the UHI are principally focused on the investigation of few parameters. More in details, some existing studies that consider a greater number of parameters are typically dedicated on the investigation of specific regions or cities and do not propose a quantification of their influence. For instance, Lokoshchenko 30 investigates many meteorological parameters such as: air temperature, soil temperature, precipitation, wind speed, but also green areas, albedo and density of population. However, Lokoshchenko 30 focusses the attention on the UHI in Moscow and do not quantify the impact of the considered parameters in the phenomenon. In this context the review papers of Santamouris 4 , 10 , demonstrate that a complete analysis on the parameters influencing the absolute maximum UHII is missing in related literature. To this reason, in comparison with these previous studies, the proposed work provides an important new contribution by quantifying the influence of the parameters involved in the phenomenon on the basis of a scientific approach never applied before in this field. Indeed, this research contemplates the use of multi-criteria methods in synergy with mathematical optimization and a Jackknife resampling approach to investigate the absolute maximum UHII.

The second topic of the proposed work regards the achieving of a new index to evaluate the hazard of the absolute max UHI phenomenon in urban district during the Summer season. A similar goal is proposed in the study of Theeuwes et al. 17 . Such study calibrates a diagnostic equation of the max UHI by using many parameters including meteorological variables, such as incoming shortwave radiation, daily thermal excursion and wind speed. The Theeuwes’s study utilized weather stations set-up by hobby meteorologists located in gardens. However, such approach gives a limited variability in urban climate zones where measurements are taken. In addition, the obtained diagnostic equation does not consider several other urban properties such as building materials, anthropogenic heat and characteristics of city and canyons. Alternatively, other studies applied statistical models and linear regression for the calibration including parameters related to the characteristics of city in addition to the meteorological variables 15 , 16 . Unfortunately, these last statistical approaches require a retuning for each city as it is remarked in the study of Theeuwes et al. 17 .

To sum up, in the existing approaches, an easy-to-use concise index, able to consider many parameters influencing the phenomenon, and applicable in different cities, has never been obtained. On the contrary, our research deals with the calibration and validation of an index involving both qualitative and quantitative data in the analysis. In addition, in order to overcome the drawback of retuning, the proposed index is calibrated by exploiting a comparison with the effective absolute max UHII recorded by standard measuring stations in 41 urban districts of 32 different cities located in the European continent. The effective annual max UHII is obtained by a wide literature analysis considering only data obtained by standard measuring stations and published in peer review journals, books, and research reports. In particular, the weights of criteria of the proposed index are calibrated by solving an optimization problem to minimize the differences between the results of the Index of Potential UHII ( I UHII ) and the effective annual max UHII. The mathematical optimization problem exploits the results of the AHP and the dataset to calibrate the index and quantify the influence of every parameter. Finally, two validation tests are performed involving a Jackknife approach to identify the stability of the solution and the evaluation of the Absolute Error of the proposed model.

The shortcoming and limitation of the proposed procedure regards the applicability for the winter season or in arctic climate. Some existing researches studied the mechanisms of the UHI in this context, showing some differences with the Summer UHI, such as the relevant influence of the soil temperature and the improving of the hazard for low temperatures. In addition, also the effect of UHI on society is different, indeed in arctic climate, the positive effects are proven due to the reduction of the need for heating 31 , 32 . However, currently the related literature does not provide enough data to perform an exhaustive calibration even in this cold weather context. For this reason, the current study is limited on the evaluation of the absolute max UHII in Summer. Moreover, the presented study does not consider some parameters typically influencing the UHI in cold climate.

On the other hand, the proposed empirical approach has many advantages in comparison with the approaches mentioned above. It is easy to apply for practitioners and the necessary data for application are available without complex acquisition procedures or unopened access databases. In particular, a comprehensive dataset to forecast the max UHII should contain at least the following information: (1) meteorological variables that can be extracted from existing meteorological European database 33 and Copernicus data 34 ; (3) characteristics of the city acquirable by exploiting satellite data processing and image analysis; (4) anthropogenic heat and city canyons that can be extracted from European Environment Agency 35 , Google Maps and Technical cartographies (when available).

The resulting index is able to outcome a hazard value to forecast the absolute max UHII in the urban districts by using eleven input data including meteorological variables, characteristics of the city (in terms of albedo and greenery), anthropogenic heat and city canyons.

To provide an example, there are not exhaustive studies on the heat island effect for the whole city Milan in in related literature. If a practitioner wanted to know the potential Hazard of all the districts of Milan, the proposed index can be applied as showed as follows.

The collection procedure allows to easily achieve the input data of all the 88 urban districts of the city of Milan. Note that information are acquired by considering an area within 500 m from the centre of gravity of considered districts.

For the case of Milan, data regarding meteorological variables are the same for the whole city, instead the characteristics of the city , anthropogenic heat and city canyons have a great variability.

The proposed index I UHII is applied to forecast the hazard of the phenomenon for every one of the 88 urban districts by using Eq. ( 1 ) and the tabulated weight of Supplementary Table S1 . Figure  4 shows the application of the index to generate an hazard map of the city of Milan for the Absolute Max UHII. The central areas of Milan, (near the cathedral) are the most subject to suffer the UHI caused by an intensely developed urban fabric. The only two exceptions concern the central districts with large parks (Sempione Park and the Garden of Porta Venezia). On the contrary, outside the central area, the UHI effect decreases principally thanks to greater green areas and a minor canyon effect produced by different urban layout (e.g. lower buildings, wider and greener streets). To sum up, the proposed approach can be useful to analyse the UHI phenomenon in single urban district or in hole cities. It is a useful tool for researcher to forecast the UHI hazard during in Summer season. In addition, also practitioners, administrations and governments can exploit this tool for urban planning, to understand the needs and effectively design new green areas or mitigation strategies of UHI as showed in the case of Milano.

Example of the application of the I UHII to create a hazard map of the city districts of Milan. The Map is generated by using SketchUp Make 2016 v16.1.1451 ( https://www.sketchup.com/ ) and Adobe Photoshop 2020 v21.1.0 ( https://www.adobe.com ) starting from the imagery of the “Territorial information system” of Milano ( https://geoportale.comune.milano.it/sit/ ).

Conclusions

The obtained potential UHII index can be a useful tool for evaluate effective urban planning strategies at the urban scale and identify the best intervention for individual building aimed at the minimization of the phenomenon. Beyond this, the proposed index could also be used by researchers, architects, engineers and interested stakeholders to obtain maps of Hazard at different scales (urban, regional or national) and have valuable assistance in identifying energy retrofit interventions in order to improve the energy performances of buildings and mitigate the UHI.

In comparison with the existing studies, the main novelty of the paper is threefold. For the first time, the influence of each parameter involved in the UHII is quantified. Different techniques of multi-criteria decision methods, state of the art data acquisition procedure, mathematical optimization and validation are simultaneously employed in a structured and synergic way for analysing the UHII. The second novel issue regards the proposal of a quantitative index to evaluate the potential absolute max UHII. This index overcomes the limitations of the necessary experimental data acquisition described in literature to obtain a preliminary local analysis of the UHII phenomenon 4 , 11 .

Thirdly, the research reaches the ambitious result of calibrating the index by analysing 41 urban districts in 35 different cities, including the effective UHII registered and derived from a wide literature analysis.

The project involved a combination of interdisciplinary skills including meteorology, knowledge of UHII phenomenon, statistics, optimization and multi-criteria analysis, data mining, satellite data processing and image analysis in order to apply three synergistic techniques (AHP, a large-scale data acquisition process, an optimization and validation procedure).

Future research will evaluate a vulnerability index and an exposure index to the UHII phenomenon in order to obtain an overall risk index of the phenomenon. In addition, the proposed index will be integrated in Spatial Decision Support Systems for the large-scale risk assessment useful to set effective mitigation strategies.

AHP to analyse the problem

The AHP methodology is used to define the UHII index. By exploiting the information extracted from the related literature, it is possible to define criteria and sub-criteria and determine the index (An exhaustive description of the AHP method applied to define the index is presented in the supplementary methodology).

The AHP step 1 in consists in the Structure of the Problem to determine an index useful to quantify potential UHII in urban district. In particular, the goal is defined as the phenomenon Absolute Max Urban Heat Island Intensity . To this aim, eleven criteria i (with i  = 1,…,11) involved in the generation of the UHI 36 are defined and grouped into four macro-criteria in accordance with the classes of parameters defined in the introduction. For each criterion a set of intensity ranges j (with j  = 1,…, n i ) is defined to characterize its intensity levels. The eleven criteria , and related intensity ranges involved in the UHI phenomena are structured in a hierarchical flowchart that is showed in Fig.  5 and discussed in the supplementary methodology.

Structure of the Problem : weights, criteria and intensity ranges to determine an index able to quantify the potential absolute max Urban Heat Island Intensity in urban district. The chart is drawn by using Microsoft Visio 2007 v 12.0.4581.1014 ( https://www.microsoft.com ).

The AHP step 2 is used to individually analyses each aspect of the defined UHII problem (Fig.  5 ) in order to weight the parameters involved. The criteria and intensity ranges weights are defined as follows:

v i is the weight associated with each i th criterion

w ij is the weight associated with each j th intensity range related to the i th criterion.

eleven judgment matrices A are evaluated in order to identify the tabulated weights of intensity ranges w ij .

After weighting, the potential Absolute Max Urban Heat Island Intensity Index ( I me ) can be defined. This operation coincides with the AHP step 3 of the summary of priority. The Eq. ( 1 ) is obtained by multiplying each criteria weight by the intensity range weight and adding the results, as in the classical AHP procedure.

The data acquisition

This section explains the data acquisition process. State of the art techniques including bibliographic research, data mining for data extraction from existing databases, satellite data processing and image analysis are applied in order to achieve an effective dataset, useful for calibrating and validating the index. In addition, modern and updated databases are used, and out random cross-checks are carried out to verify the data reliability. It is essential to specify that data regarding meteorological, city, urban district and anthropogenic parameters refers to the specific temporal period of the acquisition of the max UHII registered in the associated urban districts. The complete set of data, including the rural station geographic coordinates and the acquisition period, is showed in the Supplementary Table S2 . In particular, the data acquisition process can be summarized in the identification of: the effective UHII , meteorological variables , characteristics of the city , anthropogenic heat , and city canyons. The acquired data regards different climatic context, dimension and characteristics of the city to achieve an effective calibration and validation.

Experimental data on the effective UHII is acquired for 41 different cities in 10 different countries of the European Continent. The information is collected from 15 scientific articles published in peer review journals, books, and research reports (Table 1 ). Among more than hundreds investigated studies, only articles providing sufficient information on the spatial area considered, the type of experiments and the used equipment are taken into account. Many of the considered studies reported experiments and data for more than one district, thus in total the paper includes data from 41 observations.

In addition, only studies aiming to quantify absolute maximum UHII by using fixed standard measuring equipment/station are considered while articles investigating UHI characteristics using mobile surveys techniques are rejected. To sum up, the considered studies can be considered reliable and homogeneous if they provide: (1) the absolute maximum UHII; (2) UHII acquisition performed by standard fixed measurement stations; (3) geographical coordinates of the urban station; (4) geographical coordinates of the rural stations; (5) data acquired in a period of time in which also climate data are available in the used meteorological databases.

Moreover, Fig.  6 shows the data about specific urban districts, the related geographic position and the UHII measures. For every urban district a specific identification code \(Ud=1,\dots ,41\) is assigned.

The meteorological variables are extracted from the European database of Tutiempo Network 33 and Clara 50 . This database is comprehensive since it has stored over the years many historical meteorological data of the main cities of the European continent. In addition, in order to verify the data reliability, a cross check with the database of Urbansis and Copernicus 34 and weatherspark 51 is exploited.

The characteristics of the cities are acquired by exploiting satellite data processing and image analysis. The goal of this data acquisition technique is twofold: (1) identifying the average of the land cover albedo on the basis of the surface material typology including water, greenery, red brick roof, white plaster or pigment, gravel or concrete; (2) quantifying the percentage of greenery in relation to the evapotranspiration. In particular, an area within 500 m from the standard measuring station is considered and high-resolution imagery of Landsat 7 and Landsat 8 satellite are used.

Supplementary Fig. S3 schematizes a part of the image analysis process for the urban district Ud  = 32 (Rome, Roma 3, Italy). In particular, land cover type (Albedo) of the urban district is obtained by using the typical urban materials albedo initially investigated by Oke 12 and reported in Van Hove et al. 38 and Bradley et al. 52 . To provide an example, the most common values for albedo detected in the investigated urban districts are: green areas and grass (0.205), forests (0.150) asphalt (0.125), water (0.500), white plaster (0.500) and red brick (0.300). The results of the analysis are showed in the Supplementary Table S2 .

The anthropogenic heat is represented by the single criterion “population density”. Data regarding the population density can be obtained by exploiting the Eurostat 53 and integrating some missing data of small Netherland cities by a specific search on local web-sites.

The last information to be acquired regards the city canyons : these data are both quantitative and qualitative. The quantitative data regard the average width of streets and building height. The width of streets can be obtained by technical cartography when available, or from a suitable image analysis. Indeed, once extracted the asphalt cover from the area within 500 m from the measuring stations, it is possible identify the average width of streets by averaging the measurable amplitudes. The building height is obtained by exploiting the classification of Oke 54 including the five Urban Climate Zones (UCZ) associated to the number of floors of the considered buildings. A qualitative evaluation is carried out by identifying the UCZ through Google street view or the 3D view of Google Earth when available. In addition, some qualitative analyses can be performed to evaluate the canyon orientation, by considering the directions of the main streets of the urban district, and the irregularity of the city.

UHIIs extracted from a literature review. The figure was created by using Google My Maps 2020 ( https://www.google.com/maps/d/ .), Microsoft Excel v16.33 ( https://www.microsoft.com ) and Adobe Photoshop 2020 v21.1.0 ( https://www.adobe.com ).

Calibration and validation

In this section, the dataset obtained from the previous phases is used to achieve the calibration and validation of the index. Starting from the definition of the index in Eq. ( 1 ), a mathematical programming problem is formulated in order to obtain calibrated weights by an optimization procedure by exploiting 41 urban districts. Afterwards two validation tests involving a Jackknife resampling procedure are carried out to identify the stability of the solution and evaluate the Absolute Error and the Mean Relative Error of the proposed model. In addition, Pearson correlation coefficient of the involved criteria with the absolute maximum UHII is calculated, in order to verify that all the eleven criteria are sufficiently correlated and then important for the analysis.

In this section, the results obtained from the previous phases are used to achieve the calibration of the index.

In order to obtain the weights v i and calibrate the index of UHII, defined in Eq. ( 1 ), the \(K\) =41 urban districts ( \(Ud=1,\dots ,K\) ) are considered in eight cities of the European continent. The index of \({{I}_{UHII}}^{Ud}\) associated with the urban districts \(Ud=1,\dots ,K\) is written in function of the column vector of weights v  = [ v 1 , v 2 , v 3 , v 4 , v 5, v 6, v 7, v 8, v 9, v 10, v 11 ] T as follows:

where j are the alternatives associated with the Ud -th urban districts and weights w ij are identified by exploiting the data showed in Supplementary Table S2 and the tabulated weight of Table S1 .

In addition, the effective UHII (of every urban districts \(Ud\) ) useful for calibration is hereafter expressed with \({UHII}^{Ud}\) ( \(Ud=1,\dots ,K\) ).

Subsequently, it is possible to define the function F ( v ), which evaluates the difference between the values of \({{I}_{UHII}}^{Ud}\) and the effective max \({UHII}^{Ud}\) (reported in Fig.  5 ) calculated for each examined urban districts Ud  = 1,…, K :

To calibrate vector v, it is assumed that the index of potential \({{I}_{UHII}}^{Ud}\) should be as close as possible to the index of effective max Urban Heat Island Intensity \({UHII}^{Ud}\) for the considered urban districts.

F ( v ) represents the sum of the differences between the proposed index and the effective intensity of the registered phenomenon. Consequently, F ( v ) is set as the objective function to be minimized by satisfying a set of constraints on vector v :

Constraints (5) are set to avoid negative or null values of the weights. Now, in order to calculate vector v , the following Mathematical Programming (MP) problem is formulated:

A generalized reduced gradient method 55 is applied to solve the MP problem by using a Multi-start of a population of 1000 and a convergence of 0.001.

To validate the index the following two tests and one correlation analysis are carried out:

Firstly, a specific test is used to evaluate how the final results would have changed if the set case study considered for validation were different by using the Jackknife approach. In particular, a one-by-one removal of the 41 urban districts is performed and every removal the MP problem is solved to provides the vector of weights v . Finally, a statistical variation of the vector v is obtained on the basis of the resampling of input data.

Secondly, an additional test identifies the Absolute Error and the Mean Relative Error of \({{I}_{UHII}}^{Ud}({\varvec{v}})\) in comparison with the effective absolute max UHII obtained from the bibliography. Also in this case, a Jackknife resampling is used to obtain the statistical variation of the absolute for every one-by-one removal of the 41 urban districts.

Thirdly, the Pearson correlation coefficient is evaluated to measure the linear correlation between the criteria and the UHII. This analysis allows understanding the presence of a predictive relationship between the considered parameters and the investigated phenomenon. In particular, the absence of low and null correlation coefficients can justify the presence of all the eleven parameters in both the AHP framework and the index evaluation.

Santamouris, M. Energy and Climate in the Urban Built Environment (James and James Science Publishers, London, 2001).

Google Scholar  

Oke, T. R., Johnson, G. T., Steyn, D. G. & Watson, I. D. Simulation of surface urban heat islands under ‘ideal’ conditions at night—Part 2: Diagnosis and causation. Bound. Layer Meteorol. 56 , 339–358 (1991).

Article   ADS   Google Scholar  

Santamouris, M. Cooling the cities—A review of reflective and green roof mitigation technologies to fight heat island and improve comfort in urban environments. Sol. Energy 103 , 682–703 (2014).

Santamouris, M. Analyzing the heat island magnitude and characteristics in one hundred Asian and Australian cities and regions. Sci. Total Environ. 512 , 582–598 (2015).

Article   ADS   PubMed   CAS   Google Scholar  

Stathopoulou, E., Mihalakakou, G., Santamouris, M. & Bagiorgas, H. S. On the impact of temperature on tropospheric ozone concentration levels in urban environments. J. Earth Syst. Sci. 117 , 227–236 (2008).

Article   ADS   CAS   Google Scholar  

Santamouris, M. Heat island research in Europe—The state of the art. Adv. Build. Energy Res. 1 , 123–150 (2007).

Article   Google Scholar  

Pantavou, A., Theoharatos, G., Mavrakis, A. & Santamouris, M. Evaluating thermal comfort conditions and health responses during an extremely hot summer in Athens. Build. Environ. 2 , 339–344 (2011).

Santamouris, M. & Kolokotsa, D. On the impact of urban overheating and extreme climatic conditions on housing energy comfort and environmental quality of vulnerable population in Europe. Energy Build. 98 , 125–133 (2015).

Sakka, A., Santamouris, M., Livada, I., Nicol, F. & Wilson, M. On the thermal performance of low income housing during heat waves. Energy Build. 49 , 69–77 (2012).

Santamouris, M. Innovating to zero the building sector in Europe: Minimising the energy consumption, eradication of the energy poverty and mitigating the local climate change. Sol. Energy 128 , 61–94 (2016).

Stewart, I. D. A systematic review and scientific critique of methodology in modern urban heat island literature. Int. J. Climatol. 31 , 200–217 (2011).

Oke, T. R. Boundary Layer Climates Routledge (Mathuem & CO, London, 1987).

Unger, J. Intra-urban relationship between surface geometry and urban heat island: Review and new approach. Clim. Res. 27 , 253–264 (2004).

Arnfield, A. J. Two decades of urban climate research: A review of turbulence, exchanges of energy and water, and the urban heat island. Int. J. Climatol. J. R. Meteorol. Soc. 23 , 1–26 (2003).

Hoffmann, P., Krueger, O. & Schlünzen, K. H. A statistical model for the urban heat island and its application to a climate change scenario. Int. J. Climatol. 32 , 1238–1248 (2012).

Szymanowski, M. & Kryza, M. Local regression models for spatial interpolation of urban heat island—An example from Wrocław, SW Poland. Theor. Appl. Climatol. 108 , 53–71 (2012).

Theeuwes, N. E., Steeneveld, G. J., Ronda, R. J. & Holtslag, A. A. A diagnostic equation for the daily maximum urban heat island effect for cities in northwestern Europe. Int. J. Climatol. 37 , 443–454 (2017).

Sangiorgio, V., Martiradonna, S., Uva, G., & Fatiguso, F. An information system for masonry building monitoring. In 2017 IEEE International Conference on Service Operations and Logistics, and Informatics (SOLI) . 1 , 230–235 (2017).

De Almeida, A. T., Alencar, M. H., Garcez, T. V. & Ferreira, R. J. P. A. systematic literature review of multicriteria and multi-objective models applied in risk management. IMA J. Manag. Math. 28 , 153–184 (2017).

Article   MathSciNet   MATH   Google Scholar  

Sangiorgio, V., Uva, G. & Fatiguso, F. Optimized AHP to overcome limits in weight calculation: Building performance application. J. Constr. Eng. Manag. 144 , 04017101 (2017).

Sangiorgio, V., Uva, G., Fatiguso, F. & Adam, J. M. A new index to evaluate exposure and potential damage to RC building structures in coastal areas. Eng. Fail. Anal. 100 , 439–455 (2019).

Saaty, T. L. Decision making with the analytic hierarchy process. Int. J. Serv. Sci. 1 , 83–98 (2008).

Zhao, L., Lee, X., Smith, R. B. & Oleson, K. Strong contributions of local background climate to urban heat islands. Nature 511 , 216–219 (2014).

Article   ADS   CAS   PubMed   Google Scholar  

Lokoshchenko, M. A. Urban heat island and urban dry island in Moscow and their centennial changes. J. Appl. Meteorol. Climatol. 56 , 2729–2745 (2017).

Hung, T., Uchihama, D., Ochi, S. & Yasuoka, Y. Assessment with satellite data of the urban heat island effects in Asian mega cities. Int. J. Appl. Earth Observ. Geo Inf. 8 , 34–48 (2006).

Bonafoni, S., Baldinelli, G. & Verducci, P. Sustainable strategies for smart cities: Analysis of the town development effect on surface urban heat island through remote sensing methodologies. Sustain. Cities Soc. 29 , 211–218 (2017).

Dong, Y., Varquez, A. C. G. & Kanda, M. Global anthropogenic heat flux database with high spatial resolution. Atmos. Environ. 150 , 276–294 (2017).

Flanner, M. G. Integrating anthropogenic heat flux with global climate models. Geophys. Res. Lett. 36 , 1–5 (2009).

Yang, W. et al. A new global anthropogenic heat estimation based on high-resolution nighttime light data. Sci. Data 4 , 170116 (2017).

Article   PubMed   PubMed Central   Google Scholar  

Lokoshchenko, M. A. Urban ‘heat island’ in Moscow. Urban Clim. 10 , 550–562 (2014).

Fortuniak, K., Kłysik, K. & Wibig, J. Urban–rural contrasts of meteorological parameters in Łódź. Theor. Appl. Climatol. 84 , 91–101 (2006).

Konstantinov, P., Varentsov, M. & Esau, I. A high density urban temperature network deployed in several cities of Eurasian Arctic. Environ. Res. Lett. 13 , 075007 (2018).

Tutiempo Network, S.L. Climate data: Europe. https://www.tutiempo.net/ (2020)

ECMRWF (European Centre for Medium-Range Weather Forecasts). Climate change service. https://climate.copernicus.eu/ (2020).

EEA Web Team. European Environmental Agency. https://www.eea.europa.eu/ (2020).

Rizwan, A. M., Dennis, L. Y. & Chunho, L. I. U. A review on the generation, determination and mitigation of Urban Heat Island. J. Environ. Sci. 20 , 120–128 (2008).

Article   CAS   Google Scholar  

Varentsov, M., Wouters, H., Platonov, V. & Konstantinov, P. Megacity-induced mesoclimatic effects in the lower atmosphere: A modeling study for multiple summers over Moscow Russia. Atmosphere 9 , 1–24 (2018).

Van Hove, L. W. A., et al.  Exploring the urban heat island intensity of Dutch cities: Assessment based on a literature review, recent meteorological observation and datasets provide by hobby meteorologists. No. 2170. Alterra (2011).

Giovannini, L., Zardi, D. & de Franceschi, M. Analysis of the urban thermal fingerprint of the city of Trento in the Alps. J. Appl. Meteorol. Climatol. 50 , 1145–1162 (2011).

Guattari, C., Evangelisti, L. & Balaras, C. A. On the assessment of urban heat island phenomenon and its effects on building energy performance: A case study of Rome (Italy). Energy Build. 158 , 605–615 (2018).

Marando, F., Salvatori, E., Sebastiani, A., Fusaro, L. & Manes, F. Regulating ecosystem services and green infrastructure: Assessment of urban heat island effect mitigation in the municipality of Rome, Italy. Ecol. Model. 392 , 92–102 (2019).

Milelli, M. Urban heat island effects over Torino. COSMO Newsl. 16 , 1–10 (2016).

Busato, F., Lazzarin, R. M. & Noro, M. Three years of study of the Urban Heat Island in Padua: Experimental results. Sustain. Cities Soc. 10 , 251–258 (2014).

Petralli, M. et al. Ruolo delle aree verdi nella mitigazione dell’isola di calore urbana: Uno studio nella città di Firenze. Riv. Ital. Agrometeorol. 1 , 51–58 (2006).

Kolokotroni, M. & Giridharan, R. Urban heat island intensity in London: An investigation of the impact of physical characteristics on changes in outdoor air temperature during summer. Sol. Energy 82 , 986–998 (2008).

Krüger, E. & Emmanuel, R. Accounting for atmospheric stability conditions in urban heat island studies: The case of Glasgow, UK. Landsc. Urban Plan. 117 , 112–121 (2013).

Tayanc, M. & Toros, H. Urbanization effects on regional climate change in the case of four large cities of Turkey. Clim. Change. 35 , 501–524 (1997).

Cheval, S., Dumitrescu, A. & Bell, A. The urban heat island of Bucharest during the extreme high temperatures of July 2007. Theoret. Appl. Climatol. 97 , 391–401 (2009).

Lemonsu, A. & Masson, V. Simulation of a summer urban breeze over Paris. Bound. Layer Meteorol. 104 , 463–490 (2002).

Karlsson, K. G. et al . CLARA-A1: CM SAF Clouds, Albedo and Radiation dataset from AVHRR data–Edition 1–Monthly Means/Daily Means/Pentad Means/Monthly Histograms, Satellite Application Facility on climate Monitoring (2012).

Weatherspark Climate data: World. https://it.weatherspark.com (2020).

Bradley, A. V., Thornes, J. E., Chapman, L., Unwin, D. & Roy, M. Modelling spatial and temporal road thermal climatology in rural and urban areas using a GIS. Climate Res. 22 , 41–55 (2002).

Eurostat - European statistics. https://ec.europa.eu/info/departments/eurostat-european-statistics (2018).

Oke, T.R. Initial guidance to obtain representative meteorological observations at urban sites. Instruments and observing methods, no. 81. Canada. World Meteorological Organization . (2006).

Lasdon, L. S., Fox, R. L. & Ratner, M. W. Nonlinear optimization using the generalized reduced gradient method. Revue française d’automatique, informatique, recherche opérationnelle. Rec. Opér. 8 , 73–103 (1974).

Download references

Author information

Authors and affiliations.

DICATECH, Politecnico di Bari, Via Edoardo Orabona 4, Bari, Italy

Valentino Sangiorgio & Francesco Fiorito

High Performance Architecture, School of Built Environment, University of New South Wales, Sydney, NSW, 2052, Australia

Francesco Fiorito & Mattheos Santamouris

You can also search for this author in PubMed   Google Scholar

Contributions

V.S. made the following contribution: Conceptualization, Methodology development, Formal analysis, Investigation, Writing the Original Draft, Visualization and data presentation. Francesco Fiorito made the following contribution: Conceptualization, Methodology development, Writing Review, Editing and Supervision. M.S. made the following contribution: Conceptualization, Methodology development, Writing Review, Editing and Supervision.

Corresponding author

Correspondence to Valentino Sangiorgio .

Ethics declarations

Competing interests.

The authors declare no competing interests.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary figures., supplementary methodology., supplementary tables., rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Sangiorgio, V., Fiorito, F. & Santamouris, M. Development of a holistic urban heat island evaluation methodology. Sci Rep 10 , 17913 (2020). https://doi.org/10.1038/s41598-020-75018-4

Download citation

Received : 06 April 2020

Accepted : 08 October 2020

Published : 21 October 2020

DOI : https://doi.org/10.1038/s41598-020-75018-4

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

This article is cited by

Normalized urban heat island (uhi) indicators: classifying the temporal variation of uhi for building energy simulation (bes) applications.

• Manan Singh
• Ryan Sharston

Building Simulation (2023)

Assessment of the impact of the different settlement patterns on the summer land surface temperature: Elazığ

• Yaşar Menteş
• Sevgi Yilmaz
• Hasan Yilmaz

Environmental Science and Pollution Research (2022)

Intra-urban microclimate investigation in urban heat island through a novel mobile monitoring system

• Ioannis Kousis
• Ilaria Pigliautile
• Anna Laura Pisello

Scientific Reports (2021)

By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here .

Loading metrics

Open Access

Peer-reviewed

Research Article

Contrary to expectation: The surface urban heat island intensity is increasing in population shrinking region while decreasing in population growing region-A comparative analysis from China

Roles Conceptualization, Data curation, Methodology, Software, Writing – original draft

Affiliation School of Earth and Environmental science, The University of Queensland, Queensland, Australia

Roles Formal analysis, Methodology, Writing – review & editing

* E-mail: [email protected]

Affiliation School of Geographical Sciences, Southwest University, Chongqing, China

• Luofu Liu, 

• Published: March 18, 2024
• https://doi.org/10.1371/journal.pone.0300635
• Reader Comments

Exploring the complex relationship between population change and surface urban heat island (SUHI) effect has important practical significance for the ecological transformation development of shrinking cities in the context of the prevalence of urban shrinkage and the global climate change. This paper compares the population change and SUHI effect between population shrinking region (Northeast Region, NR) and population growing region (Yangtze River Delta, YRD) in China, and explores their differences in driving mechanisms, using GIS spatial analysis and Geodetector model. Our results indicated that there are significant differences in population changes and SUHI intensity between these two regions. About 72.22% of the cities in the NR were shrinking, while their SUHI intensities increased by an average of 1.69°C. On the contrary, the urban population in the YRD shows a linear growth trend, while their SUHI intensities decreased by 0.11°C on average. The results of bivariate Moran’s I index also indicated that the spatial correlation between the urban population changes and the SUHI intensity changes are not significant in the above regions. Furthermore, there are significant differences in the primary drivers of SUHI variations between these two regions. In the NR, underlying surface changes, including the changes of green coverage and built-up areas, are the most important driving factors. However, atmospheric environment changes, such as carbon dioxide emission and sulfur dioxide emission, are the key drivers in the YRD. Northam’s theory of three-stage urbanization and environmental Kuznets curve hypothesis are powerful to explain these differences.

Citation: Liu L, Zhang W (2024) Contrary to expectation: The surface urban heat island intensity is increasing in population shrinking region while decreasing in population growing region-A comparative analysis from China. PLoS ONE 19(3): e0300635. https://doi.org/10.1371/journal.pone.0300635

Editor: Baojie He, Chongqing University, CHINA

Received: November 20, 2023; Accepted: March 2, 2024; Published: March 18, 2024

Copyright: © 2024 Liu, Zhang. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: All relevant data are within the manuscript and its Supporting information files.

Funding: This research was funded by the Fundamental Research Funds for the Central Universities (SWU2009435). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

1. Introduction

Both growth and shrinkage are essential historical stages in the process of urbanization [ 1 , 2 ]. However, the growth-oriented mode occupied the mainstream of urban development theory since the Industrial Revolution in the middle of the 18th century [ 3 ]. Urban shrinkage has usually been ignored in this context because the antithesis of growth, such as population loss, economic slowdown and land vacancy, are often negatively labeled as “failure”, “recession” and “ignominy”. In recent years, urban shrinkage becomes a hot topic due to the prevalence of shrinking cities. Nearly a quarter of the world’s cities are experiencing long-term population decline because of the combined effect of globalization, deindustrialization, suburbanization and population aging [ 4 ]. Cities in America’s rust belt, such as Detroit and Pittsburgh, have lost even more than half of its population between 1960 and 2010 [ 5 , 6 ]. Growing attentions have been paid on urban shrinkage because it is recognized as a global phenomenon and affects the fate of many cities decisively [ 7 , 8 ].

Urban heat island (UHI) effect is the phenomenon that urban areas are warmer than their rural hinterlands [ 9 ]. In recent years, growing greenhouse gas emissions and accelerating urbanization process have caused the rising of global temperature [ 10 ], and then triggered the increasing of extreme heat events and the aggravation of UHI effect. Against this background, the health threats of extreme heat events, UHI effect and their synergy on urban residents are growing [ 11 ]. For example, the 2003 heat wave in France caused about 15,000 excess deaths [ 12 ]. Heat waves have been listed as the most dangerous natural disasters in many places like the United States, Europe and Australia [ 13 ]. The mitigation of UHI effect has become a hot topic globally.

Urban shrinkage was usually regarded as a threat or a challenge in earlier studies [ 14 ]. Recently, growing scholars begin to explore the opportunities brought by urban shrinkage, especially for the potential ecological opportunities. Haase et al. [ 15 ] believed urban shrinkage offers great potential for rebuilding urban green space and they developed a matrix approach to link population shrinkage and ecosystem services. Zeng et al. [ 16 ] found that urban shrinkage helps the improvement of CO 2 emission efficiency. Unfortunately, limited literatures discussed the relationship between population shrinkage and SUHI effect, and most of them are case studies that focused on individual cities. For example, Emmanuel and Krueger [ 17 ] discussed the variation of the UHI intensities in the urban shrinkage and growth processes in Glasgow, UK. Jang [ 18 ] explored the impacts of demolition programs for abandoned houses on thermal comfort in Daegu, South Korea. Cai et al. [ 19 ] analyzed the impact of short-term population loss on urban thermal environment in Wuhan, China. These case studies provide important references for subsequent research. Nevertheless, it is difficult to obtain universal conclusions from a small number of case studies considering the diversity of shrinking cities. Peng et al. [ 20 ] explored the evolution characteristics of SUHI effect in 180 shrinking cities in China. But they did not comprehensively compare the SUHI effects of shrinking and growing cities, nor did they quantify its driving mechanisms. In addition, urban shrinkage and SUHI effect are closely related to the regional background of each city. However, previous studies have not analyzed the relationship between population shrinkage and SUHI effect from the regional perspective. In summary, exploring the relationship between urban shrinkage and SUHI effect at the regional scale is helpful to fill these research gaps. Its significances are as follows: (1) it helps to reveal the influence of population shrinkage on the urban eco-environment, and thus provides some empirical evidences for the ecological transformation development of shrinking cities [ 15 , 16 ]. (2) It is helpful to unravel the complex relationship between population change and SUHI effect [ 19 , 20 ], so as to provide useful assistance for the mitigation of SUHI effect during the population shrinkage process. (3) Most previous studies on urban shrinkage and UHI effect were focused on individual cities [ 5 , 17 , 19 ]. The regional perspective presented in this study is hopeful to make a good contribution to the literatures.

In this paper, a population shrinking region (Northeast Region) and a growing region (Yangtze River Delta) in China were selected to explore the relationships between urban population shrinkage and SUHI intensity. This paper seeks to address the following questions: (1) are there any differences in SUHI effects between a typical shrinking region and a growing region? (2)Which factors are the key drivers of their respective SUHI effects? (3) How to mitigate the SUHI effect in the context of regional population change?

2. Material and methods

2.1. study area.

China has experienced an unprecedented process of rapid urbanization since the 1980s. Its urban population increased from 191 million in 1980 to 914 million in 2021; and its urban built-up area has also increased from 7,438 km 2 in 1980 to 62,421 km 2 in 2021 [ 21 ]. However, China’s rapid urbanization has also given rise to a series of eco-environmental problems, including UHI effect [ 22 ]. The UHI effect in China is continuously increasing due to the combined effect of multiple factors like climate change, built-up area expansion, population agglomeration and air pollution, which poses a serious threat to the health of urban residents. The results of previous studies indicated that the heat-related deaths in China increased fourfold between 1990 and 2019 [ 23 ].

Shrinking cities have emerged in China in the past few years for multiple reasons like population aging, fewer children and economic fluctuation [ 24 ]. The Northeast region (NR) is the most typical shrinking region because it has the most shrinking cities in China [ 25 ]. From 2010 to 2020, the total population of the NR decreased by 11.3 million [ 21 ]. On the contrary, the Yangtze River Delta (YRD) is a population growing region. From 2010 to 2020, the total population of the YRD increased by 19.62 million [ 21 ]. YRD is one of the most developed regions in China. It contributes nearly a quarter of China’s economic output with less than 4% of China’s territory. To sum up, NR and YRD are the two most representative cases in China from the perspective of regional population change. It is helpful to clarify the complex relationship between regional population change and SUHI effect by the comparative analysis of these two typical regions.

2.2. Data sources and pre-processing

• (1) MODIS Land Surface Temperature (LST) data MOD11A2 LST product with a spatial resolution of 1km was used to analyze the SUHI effect in China. The original remote sensing image of this product was produced by the Moderate-resolution Imaging Spectroradiometer (MODIS), and this product can be downloaded on the official website ( https://ladsweb.modaps.eosdis.nasa.gov/ ) of the National Aeronautics and Space Administration (NASA). MODIS Reprojection Tool (MRT) provided by the Land Processes Distributed Active Archive Center (LP DAAC) ( https://lpdaac.usgs.gov/lpdaac/tools/ ) was utilized for the preprocessing of MODIS products like trans-projection, splicing and clipping. The value of 0 in the quality assurance layer was utilized to filter the MOD11A2 pixels for accuracy guarantee. The time resolution of MOD11A2 LST product is 8 days, so the maximum value compositing method [ 13 ] was used to generate annual LST data for 2010–2020.
• (2) Meteorological data Annual spatial interpolation data of multiple meteorological elements, including sunshine hours, average wind speed, relative humidity and evapotranspiration, were obtained from the Resource and environment science and data center, Chinese Academy of Sciences ( https://www.resdc.cn/ ). This data was produced from the daily observation data of corresponding meteorological elements at more than 2400 meteorological stations in China, using Anuspl software for spatial interpolation. The spatial resolution of this data is 1 km.
• (3) Digital Elevation Model (DEM) data The Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Global Digital Elevation Model (GDEM) V002 dataset used in this paper was obtained from the official website of NASA ( https://search.earthdata.nasa.gov/search/ ). Its spatial resolution is 30 m. ArcGIS 10.5 software was utilized to obtain the average elevation of each city.
• (4) Normalized difference vegetation index (NDVI) data Monthly NDVI data (2010–2020) with a spatial resolution of 1km was obtained from the National Earth System Science Data Center, National Science & Technology Infrastructure of China ( http://www.geodata.cn ). Maximum value compositing method was used to generate annual NDVI data for 2010–2020.
• (5) Land cover type data Annual land cover type data for 2010–2020 with a spatial resolution of 30 m was obtained from Wuhan University. Its overall accuracy reached 79.31%. The land cover types of this data include cropland, forest, shrub, grassland, water, snow and ice, barren, impervious, and wetland. More details about this data can be found in Yang and Huang [ 26 ].
• (6) Fine particulate matter data Annual particular matter with a diameter smaller than 2.5 microns (PM 2.5 ) data for 2010–2020 was obtained from the Atmospheric Composition Analysis Group, Washington University in St. Louis ( https://sites.wustl.edu/acag/datasets/surface-pm2-5/ ). Its version is V5.GL.03, and its spatial resolution is 0.01° × 0.01°. More details about this data can be found in Aaron et al. [ 27 ].
• (7) Carbon dioxide emission data Carbon dioxide (CO 2 ) emission data for 2010–2020 was derived from the Open-Data Inventory for Anthropogenic Carbon dioxide (ODIAC), which is produced by the center for global environment research, national institute for environmental studies ( https://db.cger.nies.go.jp/dataset/ODIAC/ ). The spatial resolution of this data is 1 km, and its version is ODIAC2022. More details about this data can be found in Oda et al. [ 28 ].
• (8) Statistical data Statistical data at the city scale was used to construct multiple variables in this paper, including urban population, urban built-up area, GDP, volume of sulphur dioxide emission, number of industrial enterprises above the designated size, etc. These statistics were obtained from the China City Statistical Yearbook and the China Urban Construction Statistical Yearbook.

2.3 Methodology

2.3.1. calculation of suhi intensity..

According to the concept of UHI effect, SUHI intensity was defined as the temperature difference between urban area and its surrounding rural area [ 29 , 30 ]. Two land cover types, urban impervious surface and rural cropland, were selected as typical urban and rural area respectively in this research [ 31 ]. The reasons for selecting arable land as the rural reference sites are as follows: (1) Arable land is a typical landscape in rural areas. (2) Compared with arable land, forest and wetland are less affected by humans. However, the surface temperatures of forest and wetland were significantly lower than others, so usually they were eliminated from normal rural areas for avoiding the overestimate of the SUHI effect [ 29 , 31 ]. (3) The study area of this paper is located in the northeast and eastern regions of China, while the grassland and bare land are concentrated in the western region of China. Therefore, they are unsuitable for this research.

• PPT PowerPoint slide
• PNG larger image
• TIFF original image

Abbreviation. ISP: impervious surface parcel; LST: land surface temperature; SUHI: surface urban heat island.

https://doi.org/10.1371/journal.pone.0300635.g001

In addition, k i in formula ( 2 ) was used to represent the dynamic trend of urban population too. Cities with a value of k i less than 0 were classified as shrinking cities.

2.3.2. Statistics test method.

Parametric test and nonparametric test tools were utilized to test the significance of regional difference in population change and SUHI effect between NR and YRD. Specifically, the Independent-samples T test was selected for parameter testing; and Mann-Whitney U test, Kolmogorov-Smirnov Z test and Wald-Wolfowitz Runs test were selected for nonparametric testing. SPSS 25.0 software was used to complete these statistics tests.

2.3.3. Spatial correlation analysis.

2.3.4. Geodetector model.

2.3.5 Driving mechanism analysis.

The forming mechanism of SUHI effect is complicated. Numerous social, economic and environmental factors are contributed to SUHI effect. This research focuses on the following socio-economic factors considering its research objectives ( Fig 2 ).

• (1) Population change. Urban population change exerts multiple effects on SUHI intensity. First, residents will directly increase the urban heat through human metabolism and energy consumption in their daily life. These anthropogenic heat emissions are important contributors to SUHI effect. Previous studies have proved that even temporary urban population reduction can reduce SUHI intensity immediately [ 33 , 34 ]. Secondly, urban residents will indirectly affect the SUHI intensity through their impact on land cover change [ 35 , 36 ].
• (2) Underlying surface change. Compared with rural regions, the space in the city is very limited. Consequently, drastic land cover change is an inevitable part of urbanization process. The rapid increase in impervious surface raises the temperature of urban regions because its heat capacity is small, while heat conductivity is large [ 37 , 38 ]. Meanwhile, forest and wetland have important ecological functions of local microclimate regulation. Unfortunately, they are significantly shrinking during the urbanization process [ 39 ]. In short, the drastic change of underlying surface is one of the major contributors to SUHI effect.
• (3) Socio-economic change. The socioeconomic development has a dual impact on SUHI effect [ 40 ]. On the negative side, industrial production increases energy consumption and heat emission remarkably [ 41 ]; the development of real estate industry accelerates the transformation of natural surface to artificial surface in urban regions. These socio-economic processes will exacerbate the SUHI effect [ 42 ]. On the positive side, the rapid economic development also contributes to the growth of fiscal revenue, which provides important financial guarantee for technological progress and environmental improvement, helping to mitigate the SUHI effect.
• (4) Atmospheric environment change. SUHI effect is a local meteorological phenomenon [ 43 ], and there is a complex interaction between atmospheric environment and SUHI effect [ 44 ]. Previous studies have proved that LST is positively correlated with O 3 concentration and negatively correlated with PM 2.5 , PM 10 , SO 2 , NO 2 and CO concentration [ 45 ]. The improvement of atmospheric environment is helpful to reduce the urban-rural difference in incident radiation and thus mitigate the SUHI effect [ 46 ].

https://doi.org/10.1371/journal.pone.0300635.g002

Finally, 21 socio-economic indicators were preliminarily selected as the driving factors of SUHI effect ( Table 1 ). SPSS 25.0 software were utilized for pre-regression, and indicators with Variance Inflation Factor (VIF) bigger than 10 or failed the significance test were removed.

https://doi.org/10.1371/journal.pone.0300635.t001

3.1. Regional population change

The results presented in Fig 3 indicated that the change trends of urban population in the Northeast region (NR) and the Yangtze River Delta (YRD) are quite different. The urban population of the YRD increased from 61.83 million in 2010 to 72.93 million in 2020, which shows a linear growth trend. However, the NR exhibits an inverted U-shaped trend of population change. Its urban population peaked in 2013 (36.78 million) and has continued to decline in subsequent years.

https://doi.org/10.1371/journal.pone.0300635.g003

As for city scale, the number of shrinking cities in the NR is 26, accounting for 72.22% of its total number of cities. In contrast, there are only three shrinking cities in the YRD, accounting for 7.32% of its total number of cities. The result of Levene test indicated that the samples of these two regions meet the hypothesis testing of the heteroscedasticity (F = 6.233, p value = 0.015<0.05), and the corresponding results of Independent-samples T test indicated that there were significant differences in population change between these two regions (t = -4.307, Sig. (2-tailed) = 0.000 < 0.1). In addition, all the p values of three nonparametric tests, including Mann-Whitney U test, Kolmogorov-Smirnov Z test and Wald-Wolfowitz Runs test, were less than 0.01. These results manifest there are significant differences in the central location, cumulative frequency distribution curve and overall distribution of samples from these two regions.

3.2. SUHI intensity

The change trends of SUHI intensity in the NR and the YRD are quite different too. In the NR, the SUHI intensities at the city scale increased by 1.69°C on average during 2010–2020. Cities with rapid SUHI intensity growth were concentrated in the central part of the NR. The SUHI intensities of all the three provincial capitals, namely Harbin, Changchun and Shenyang, were increased by more than 2°C. On the contrary, the SUHI intensities in the YRD decreased by 0.11°C on average during 2010–2020. The cities with a significant increase in SUHI intensity were scattered in the peripheral areas, while the SUHI variations of provincial capital cities were not sharp.

The result of Levene test indicated that the samples of these two regions meet the hypothesis testing of the homogeneity variance (F = 1.240, p value = 0.269>0.05), and the results of following Independent-samples T test indicated that the differences of SUHI intensity changes between these two regions were not significant (t = 1.227, Sig. (2-tailed) = 0.224 >0.1). In addition, the p values of Mann-Whitney U test, Kolmogorov-Smirnov Z test and Wald-Wolfowitz Runs test were greater than 0.05, which means the differences in the central location, cumulative frequency distribution curve and overall distribution of samples from these two regions are not significant.

3.3 Spatial correlation

As is shown in Fig 4 , the global bivariate Moran’s I indexes of urban population change and SUHI intensity change of two regions are low and fail to pass the significance test (p value>0.05). These results indicate that the spatial correlation between urban population change and the variation of SUHI intensity is not significant in either population shrinking region or growing region.

https://doi.org/10.1371/journal.pone.0300635.g004

3.4 Driving mechanism of SUHI variations

As is shown in Fig 5 , the primary drivers of SUHI variations in NR and YRD are different. In NR, the change of underlying surface is the most important driving factor, including green coverage and built-up area. In the case of Harbin, the capital of Heilongjiang province, its urban population decreased by 5.02% during 2010–2020, but its built-up area increased from 386.91 km 2 to 473.00 km 2 , with a growth rate of 22.25%. Its per capita road surface area also increased from 7.91 m 2 to 16.01 m 2 , with a growth rate of 102.34%. Correspondingly, its green coverage rate in built-up district decreased from 38.38% to 34.22%; and the per capita public recreational green space increased from 10.07 m 2 to 10.19 m 2 , with a growth rate of only 1.24%. Consequently, the SUHI intensity of Harbin increased from 1.12°C to 5.69°C during 2010–2020. Shenyang, the capital of Liaoning province, is similar to Harbin. From 2010 to 2020, the built-up area of Shenyang increased by 37.62%, but its green coverage rate in built-up district decreased by 2.83%, resulting in an increase of 2.58°C in its SUHI intensity.

Notes: (a) Values in brackets are q statistics; (b) The full name of each indicator can be found in Table 1 .

https://doi.org/10.1371/journal.pone.0300635.g005

Atmospheric environment changes, such as carbon dioxide emission and sulfur dioxide emission, are the primary driving factor for the SUHI variation in the YRD, which is different from the NR ( Fig 5 ). For instance, Nanjing is the capital of Jiangsu province and one of the traditional “big four stoves” in China. Its urban population and built-up area increased by 1.59 million and 249.64 km 2 respectively during 2010–2020. However, the SUHI intensity of Nanjing decreased from 4.60°C to 2.76°C over that period. In recent years, Nanjing has fallen out of the list of “big four stoves” for its improvement of thermal environment [ 59 ]. The upgrading of industrial structure and the improvement of atmospheric environment are the critical contributors for the improvement of thermal environment in Nanjing. From 2010 to 2020, Secondary industry as percentage to GDP in Nanjing dropped from 45.37% to 35.19%, while the portion of tertiary industry increased from 51.85% to 62.81%. The atmospheric environment of Nanjing has been significantly improved in this context. Its industrial sulfur dioxide emissions decreased from 115,507 tons in 2010 to 9709 tons in 2020, with a reduction rate of 91.59%; meanwhile, its carbon dioxide emissions and industrial dust emissions also decreased by 39.51% and 37.23%, respectively.

As is shown in Fig 6 , there is a significant interaction effect between various socio-economic driving factors on the SUHI variations in NR and YRD, and most interaction types are dual-factor enhancement (Enhance, bi-) and nonlinear enhancement (Enhance, nonlinear-). These results indicate that the explanatory powers of drivers have been significantly improved (bigger q statistics) through interaction. In the NR, the interaction of urban population and green coverage rate has the highest explanatory power, and its q statistics rises to 0.7104. While in the YRD, the interaction of industrial enterprise quantity and carbon dioxide emission has the highest explanatory power, and its q statistics rises to 0.7265. In a word, the comprehensive effect of various socio-economic factors leads to the variation of SUHI intensity.

https://doi.org/10.1371/journal.pone.0300635.g006

4. Discussion

This paper finds an interesting phenomenon: the SUHI effect is increasing in population shrinking region, while decreasing in growing region. Population agglomeration caused by rural-urban migration is one of the most important characteristic of urbanization [ 60 , 61 ]. Dense population usually induces a series of eco-environmental problems in urban regions, including SUHI effect [ 62 ]. Therefore, it is often assumed that population agglomeration and urban expansion tend to intensify SUHI effect [ 63 , 64 ]. Then, population shrinkage helps the mitigation of SUHI effect becomes a logical corollary. However, the results of this paper do not support this corollary. Our results indicate that the relationship between population change and SUHI effect is not significant, and population shrinkage will not naturally leads to the mitigation of SUHI effect. It’s true that population shrinkage provides rare opportunities for the ecological transformation in urban areas [ 65 , 66 ], but more efforts are needed to turn these opportunities into reality. We agreed with the viewpoint that shrinking cities are not mirror images of growing cities [ 67 ]. Urban shrinkage is a complex and unique process, which does not bring about immediate and natural improvement of the urban ecological environment.

The results of driving mechanism analysis indicate that there are significant difference in the primary driving factors of SUHI variations in NR and YRD. Underlying surface change is the primary driver of the SUHI variation in the NR with the characteristic of population shrinkage. Growth-oriented has been the mainstream mode of urban development for quite a long time [ 68 , 69 ]. In this context, how to push shrinking cities back to growth is the priority task for urban policymakers. These quick cures for shrinking cities, such as strengthening resource exploitation, accelerating infrastructure construction and increasing industrial investment, are natural choices for government decision-makers [ 70 ]. Therefore, urban shrinkage may leads to the urban spatial expansion and increasing resource consumption in the short term, thus cut down the expenditures in urban eco-environment construction, results in the further deterioration of urban environment [ 71 , 72 ]. However, atmospheric environment change is the primary driver of the SUHI variation in the YRD with the characteristic of population growing. The YRD is one of the most developed and urbanized regions in China. Shanghai is the core city of the YRD region. In 2021, the urbanization rate of Shanghai reached 89.31%, ranking first in China; its per capita GDP is 175,420 yuan/person and ranking second in China. According to Northam’s theory of three-stage urbanization, the core feature of urbanization in the YRD is no longer the population migration from rural area to urban area, but the transformation of the leading industry from secondary industry to tertiary industry [ 73 , 74 ]. The transformation and upgrading of industrial structure and economic growth not only reduce environmental pollution, but also provide more financial funds for the improvement of urban habitat environment, thus promoting the mitigation of SUHI effect. This logic is consistent with the theory of environmental Kuznets curve (EKC) [ 75 , 76 ].

The significances and marginal contributions of this study are as follows: (1) it provides some new empirical evidences to the pending issues as regards the ecological transformation process of shrinking cities. Our results indicated that there is no definite conclusion between urban population change and SUHI intensity. For shrinking cities, population decline helps to cut down the individual consumption in transportation, resources and energy, thereby reducing urban anthropogenic heat emissions and contributing to the mitigation of SUHI effect [ 33 , 34 ]. However, previous studies also found that the influences of economic size and development intensity on SUHI effect are greater than population change [ 40 , 77 ]. If decision-makers still choose the extensive development path with the characteristics of dramatic spatial expansion and massive resource consumption, it will not only offset the benefits of SUHI mitigation brought by population shrinkage, but also leads to the continuous deterioration of the urban thermal environment. (2) It emphasizes the explanatory power of the historical urbanization process on SUHI effect. A large number of literatures discussed the impacts of some static driving factors on SUHI effect, such as underlying surface, climatic conditions, urban form and air pollution [ 29 , 37 , 78 ]. Regrettably, the impacts of socio-economic dynamic process on SUHI effect were rarely mentioned [ 54 ]. Although the SUHI effect is a physical geographical phenomenon, but in essence, it is a response of local thermal environment to several social-economic processes in the context of urbanization development [ 79 ], such as the evolution processes of urban population, industrial structure and resident consumption structure. Therefore, incorporating the SUHI effect into the urbanization process is an essential step to explore its deeper driving mechanism. This paper has made some preliminary attempts for this topic. But obviously, more efforts are needed to unravel the complex relationship between urbanization processes and SUHI effect. (3) It highlights the importance of regional scale analysis. It is widely accepted that urban shrinkage and SUHI effect are closely related to the individual characteristics of a specific city, such as natural conditions, demographics and industrial structure [ 2 , 80 ]. Therefore, most previous studies were focused on city scale [ 81 , 82 ]. However, growing literatures found that urban shrinkage and SUHI effect were spatially correlated [ 83 , 84 ], which means the influences of regional macro background are significant too. Considering that the analysis results of urban shrinkage and SUHI effect at the city scale were greatly affected by the individual characteristics of each city, regional scale analysis is a hopeful option to surpass the limitations of traditional city scale analysis and reveal more general rules about urban shrinkage and SUHI effect.

There are some limitations in this research. (1) The difference of average surface temperatures between urban and rural areas is used to characterize SUHI intensity in this research. Although this processing mode is consistent with the traditional concept of SUHI effect [ 29 ], but the regional average value is too simple to reflect the range, amplitude and spatial pattern of urban thermal environment comprehensively and accurately. In addition, recent literature discovered that the amplitude and direction of SUHI intensity trends were significantly influenced by non-urban reference selections [ 85 ]. Therefore, the calculation of SUHI intensity still needs improvement in the future. (2) Annual LST data was utilized to analyze SUHI effect in this research. However, SUHI effect has significant diurnal and seasonal variations [ 86 ]. Ignoring these variations may introduce some uncertainty to the conclusions of this research. (3) This paper only discussed the interactions between two variables on SUHI effect, but failed to reveal the interactions among multiple driving factors.

5. Conclusion

This research compares the population change and SUHI effect between population shrinking region (NR) and growing region (YRD) in China, and explores their differences in driving mechanisms. The results indicated that there are significant differences in population changes and SUHI intensity between these two regions. About 72.22% of the cities in the NR were shrinking, while their SUHI intensities increased by an average of 1.69°C. On the contrary, the urban population in the YRD shows a linear growth trend and only 3 cities are shrinking, but their SUHI intensities decreased by 0.11°C on average. The results of bivariate Moran’s I index also indicated that the spatial correlation between the urban population changes and the variations of SUHI intensity are not significant in both population shrinking and growing regions. Furthermore, there are significant differences in the primary drivers of SUHI variations between these two regions. In the NR, underlying surface changes, including the changes of green coverage and built-up area, are the most important driving factors. However, atmospheric environment changes, such as carbon dioxide emission and sulfur dioxide emission, are the key drivers in the YRD. Northam’s theory of three-stage urbanization and environmental Kuznets curve hypothesis are powerful to explain these differences.

This research can be deepened from the following aspects: (1) considering the complicated relationship between population change and SUHI effect, more models are recommended to explore their nonlinear multiple response relationships in the future, such as system dynamic model, intermediate effect model and structural equation model. (2) More case regions around the world should be selected to further verify the conclusions of this paper in view of the diversity of global urban shrinkage and SUHI effect.

Supporting information

https://doi.org/10.1371/journal.pone.0300635.s001

• View Article
• Google Scholar
• PubMed/NCBI
• 21. National Bureau of Statistics of China (NBSC). China Statistical Yearbook. Beijing: China Statistics Press; 2022.

Urban Heat Island studies: Current status in India and a comparison with the International studies

• Published: 04 March 2020
• Volume 129 , article number  85 , ( 2020 )

Cite this article

• K Veena 1 , 2 ,
• K M Parammasivam 1 &
• T N Venkatesh 2  

2087 Accesses

38 Citations

6 Altmetric

Explore all metrics

Urbanization has resulted in many critical issues like increase in pollution levels, sudden climatic changes and the rise of temperature in the urban area, that is the formation of Urban Heat Islands (UHI). As the density of population rises, most of the land areas are being converted into cities and cities grows very rapidly. Due to the UHI effect, the cities are becoming hotter day by day. In India, all the metropolitan cities are victims of UHI effect and the severity of heat formation, necessitates research in this area. The present paper evaluates the trends of UHI studies in Indian cities and its out reach till 2018. Heat Island classification, methods of studying UHI in India and their limitation are discussed. Eventually a comparison of new trends of UHI studies in the world and where India lacks its growth in UHI research are included in this paper. One of the findings is that numerical modelling studies are very limited in India in this field and more focus in this area is required.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price includes VAT (Russian Federation)

Instant access to the full article PDF.

Rent this article via DeepDyve

Institutional subscriptions

Similar content being viewed by others

Systematic review of the efficacy of data-driven urban building energy models during extreme heat in cities: Current trends and future outlook

Nilabhra Mondal, Prashant Anand, … Mattheos Santamouris

The impact of Iran’s urban heritage on sustainability, climate change and carbon zero

Fatemehalsadat Afsahhosseini

World Urbanisation: Trends and Patterns

Ahmad S, Matloob H, Warsi T R and Khan M D A 2011 The evaluation of heat island effect in various cities in Ganga–Yamuna Doab region using MODIS land surface temperature product; Asian J. Environ. Sci. 6(2) 185–190.

Google Scholar  

Ali Gazi M A and Mondal I 2018 Urban Heat Island and its effect on Dweller of Kolkata Metropolitan area using geospatial techniques; Int. J. Comput. Sci. Eng. 6 741–753.

Ambinakudige S 2011 Remote sensing of land cover’s effect on surface temperatures: A case study of the urban heat Island in Bangalore, India; Applied GIS 7(1) 1–12.

Amirtham L R 2016 Urbanization and its impact on Urban Heat Island intensity in Chennai Metropolitan area India; Indian J. Sci. Technol. 9(5) .

Aslam M Y, Krishna K R, Beig G, Tinmaker M I R and Chate D M 2017 Seasonal variation of Urban Heat Island and its impact on air-quality using SAFAR observations at Delhi, India; Am. J. Clim. Change 6(02) 294–305.

Article   Google Scholar  

Babazadeh M and Kumar P 2015 Estimation of the Urban Heat Island in local climate change and vulnerability assessment for air quality in Delhi; Eur. Sci. J. , 55–65.

Baby M M and Arya G 2016 A Study of Urban Heat Island and its mapping; Int. J. Sci. Eng. Res. 4(3) 45–47.

Benali A, Carvalho A, Nunes J, Carvalhais N and Santos A 2012 Estimating air surface temperature in Portugal using MODIS LST data; Remote Sens. Environ., 108–121.

Blocken B 2015 Computational Fluid Dynamics for urban physics: Importance scales possibilities limitations and ten tips and tricks towards accurate and reliable simulations; Build. Environ., 1–27.

Borbora J and Das A K 2014 Summertime Urban Heat Island study for Guwahati City, India; Sustain. Cities Soc. 11 61–66.

Chakraborty T, Sarangi C and Tripathi S N 2016 Understanding diurnality and inter-seasonality of a sub-tropical Urban Heat Island; Bound.-Layer Meteorol. , https://doi.org/10.1007/s10546-016-0223-0 .

Deosthali V 2000 Impact of rapid urban growth on heat and moisture Islands in Pune City, India; Atmos. Environ. 34 2745–2754.

Devi S S 2006 Urban Heat Islands and environmental impact; 6th Symposium on the Urban Environment and Forum on Managing our Physical and Natural Resources.

Gopalakrishnan S G, Rama Krishna T V B P S and Sharan M 2003 Some signatures of Urban Heat Patches in southern India; Proc. Indian Natl. Sci. Acad. 5 603–614.

Goswami A, Mohammad P and Sattar A 2016 A temporal study of Urban Heat Island (UHI) – A evaluation of Ahmedabad city, Gujarat; International Conference on Climate Change Mitigation and Technologies for Adaptation, pp. 1–5.

Grover A and Singh R B 2015 Analysis of Urban Heat Island (UHI) in relation to Normalized Difference Vegetation Index (NDVI): A comparative study of Delhi and Mumbai; Environments 2 125–138.

Gupta A, Dey V and Goel D A 2009 Mitigation of thermal pollution to enhance urban air quality through Remote Sensing and GIS; 10th ESRI India User Conference.

Gupta R 2012 Temporal and spatial variations of Urban Heat Island effect in Jaipur City using satellite data; Environ. Urban. Asia 3(2) 359–374.

Jain G and Sarkar S 2017 Urban Heat Island: Causes effects and mitigating strategies; Imp. J. Interdiscip. Res. 3(5) 2098–2103.

Joshi R, Raval H, Pathak M, Prajapati S, Patel A, Singh V and Kalubarme M H 2015 Urban Heat Island characterization and isotherm mapping using geo-informatics technology in Ahmedabad City, Gujarat State, India; Int. J. Geosci. 6 274–285.

Kannamma D and Sundaram D A M 2015 Significance of microclimatic study in Urban Canyons towards ambient urban space design; J. Today’s Ideas – Tomorrow’s Technol. 3(1) 95–109.

Kikon N, Singh P, Singh S K and Vyas A 2016 Assessment of urban heat islands (UHI) of Noida City, India using multi-temporal satellite data; Sustain. Cities Soc. 22 19–28.

Kratzer A 1937 Das Stadtklima.

Kratzer A 1956 The Climate of Cities (Das Stadtklima).

Kumar R and Singh S 2016 Case study for change detection analysis and land surface temperature retrieval in Uttrakhand Region and their correlation; Imp. J. Interdiscip. Res. 2(9) 1821–1825.

Kumar K S, Udaya Bhaskar P and Kumari K P 2017a Assessment and mapping of urban heat island using field data in the new capital region of Andhra Pradesh, India; Indian J. Sci. Technol. 10(11) 1–8.

Kumar R, Mishra V, Buzan J, Kumar R, Shindell D and Huber M 2017b Dominant control of agriculture and irrigation on Urban Heat Island in India; Sci. Rep. , 1–10.

Landsberg H E 1981 The Urban Climate; Int. Geophys. , 28 .

Luke H 1818 The climate of London deduced from meteorological observations made at different places in the Neighbourhood of the Metropolis; 1p.

Marques Filho E P, Cassol M A, Karam H and Rizza U 2013 The vertical structure of Tropical Urban Heat Island with LES; Am. J. Environ. Eng. 3 24–31.

Mehrotra S 2018 Urban informal housing and surface Urban Heat Island intensity: Exploring spatial association in the city of Mumbai; Environ. Urban. Asia 9(2 ) 1–20.

Mills G 2008 Luke Howard and The Climate of London; Weather 63(6) 153–157.

Mills G 2009 Luke Howard, Tim Oke and the study of urban climates; Eighth Symposium on the Urban Environment.

Mills G 2014 Urban climatology: History, status and prospects; Urban Clim. 10 479–489.

Mirzaei P A and Haghighat F 2010 Approaches to study Urban Heat Island – Abilities and limitations; Build. Environ. 45 2192–2201.

Mohan M, Kandya A and Battiprolu A 2011 Urban Heat Island Effect over national capital region of India: A study using the temperature trends; J. Environ. Prot. 2 465–472.

Mohan M, Kikegawa Y, Gurjar B R, Bhati S and Reddy Kolli N 2013 Assessment of Urban Heat Island effect for different land use–land cover from micrometeorological measurements and remote sensing data for megacity Delhi; Theor. Appl. Climatol . 112 647–658.

Mohan M, Kikegawa Y, Gurjar B R, Bhati S, Kandya A and Ogawa K 2012 Urban Heat Island assessment for a Tropical Urban Airshed in India; Atmos. Clim. Sci. 2 127–138.

Nakata-osaki C M, Lucas de Souza L C and Rodrigues D S 2015 A GIS extension model to calculate Urban Heat Island intensity based on urban geometry; Comput. Urban. Plan. Urban. Manag ., 1–16.

Nesarikar-Patki P and Raykar-Alange P 2012 Study of inuence of land cover on Urban Heat Islands in Pune using remote sensing; IOSR J. Mech. Civil Eng. 3 39–43.

Oke T R 1968 Sustainable urban development: Minimising Urban Heat Island effect and imperviousness factor; pp. 1–20.

Oke T R 1973 City size and the Urban Heat Island; Atmos. Environ. 7 769–779.

Oke T R 2006 Towards better scientific communication in urban climate; Theor. Appl. Climatol. 84 179–190.

Oke T R, Mills G, Christen A and Voogt J A 2017 Urban Climates.

Pandey P, Kumar D, Prakash A, Masih J, Singh M, Kumar S, Jain V K and Kumar K 2012 A study of Urban Heat Island and its association with particulate matter during winter months over Delhi; Sci. Total Environ. 414 494–507.

Ramachandra T V and Kumar U 2010 Greater Bangalore: Emerging Urban Heat Island; GIS Develop. 14(1) 1–16.

Sahu S, Pawar A and Mukherjee M 2014 Urban Heat Island investigation: Techniques and methods so far; Architecture – Time Space & People , pp. 16–22.

Sastry M, Majumder M, Kumar P and Kiran Kumar D E V S 2013 Sustainable urban development: Minimising Urban Heat Island effect and imperviousness factor; Sanei Working Paper Series, 13–03.

Sharma D and Bharat A 2009 Urban Heat Island effect – causes impacts methods of measurement and mitigation options; Institute of Town Planners, India 6(2) 69–77.

Sharma N and Pandey P 2015 Study of Urban Heat Island in Bathinda City, Punjab; 16th ESRI India User Conference, pp. 1–14.

Shastri H, Barik B, Ghosh S, Venkataraman C and Sadavarte P 2017 Flip flop of Day–night and Summer–winter Surface Urban Heat Island intensity in India; Sci. Rep. – Nature 1–8.

Singh P and Kikon Noyingbeni Verma P 2017 Impact of land use change and urbanization on Urban Heat Island in Lucknow city, Central India: A remote sensing based estimate; Sustain. Cities Soc. 32 100–114.

Swain D, Roberts G J, Dash J, Vinoj V, Lekshmi K and Tripathy S 2016 Impact of rapid urbanization on the microclimate of Indian cities: A case study for the city of Bhubaneswar; Proc. SPIE 9877 1–7.

Thomas G, Sherin A P, Ansar S and Zachariah E J 2014 Analysis of Urban Heat Island in Kochi, India, using a modified local climate zone classification; Procedia Environ. Sci. 21 3–13.

Toparlar Y, Blocken B, Maiheu B and van Heijst G 2017 A review on the CFD analysis of urban microclimate; Renew. Sust. Energ. Rev., 1–28.

Vogt J V, Viau A A and Paquet P F 1997 Mapping regional air temperature fields using satellite-derived surface skin temperature; Int. J. Climatol. , 1559–1579.

Zhang N, Wang X and Peng Z 2013 Large-Eddy simulation of mesoscale circulations forced by inhomogeneous Urban Heat Island; Bound.-Layer Meteorol. , 179–194.

Download references

Acknowledgement

The first author is financially supported by CSIR (Grant No. Ack.No.141410/2K15/1) through Direct SRF Scheme and it is gratefully acknowledged.

Author information

Authors and affiliations.

Madras Institute of Technology, Chennai, 600 044, India

K Veena & K M Parammasivam

CTFD Division, CSIR-National Aerospace Laboratories, Bengaluru, 560 017, India

K Veena & T N Venkatesh

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to K Veena .

Additional information

Communicated by Amit Kumar Patra

Rights and permissions

Reprints and permissions

About this article

Veena, K., Parammasivam, K.M. & Venkatesh, T.N. Urban Heat Island studies: Current status in India and a comparison with the International studies. J Earth Syst Sci 129 , 85 (2020). https://doi.org/10.1007/s12040-020-1351-y

Download citation

Received : 20 May 2019

Revised : 17 October 2019

Accepted : 14 December 2019

Published : 04 March 2020

DOI : https://doi.org/10.1007/s12040-020-1351-y

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Urbanization
• Urban Heat Islands
• numerical modelling
• Find a journal
• Publish with us
• Track your research

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• Int J Environ Res Public Health

The Impact of Urbanization on Urban Heat Island: Predictive Approach Using Google Earth Engine and CA-Markov Modelling (2005–2050) of Tianjin City, China

Nadeem ullah.

1 School of Architecture, Tianjin University, Tianjin 300272, China

Muhammad Amir Siddique

Mengyue ding, sara grigoryan, irshad ahmad khan.

2 Centre for Research in Agricultural Genomics (CRAG) CSIC-IRTA-UAB-UB, Campus UAB, Bellaterra, 08193 Barcelona, Spain

Zhihao Kang

Shangen tsou, tianlin zhang, yazhuo zhang.

3 School of Civil Engineering, Tianjin University, Tianjin 300272, China

Associated Data

On request, the authors will provide the data from this study.

Urbanization has adverse environmental effects, such as rising surface temperatures. This study analyzes the relationship between the urban heat island (UHI) intensity and Tianjin city’s land cover characteristics. The land use cover change (LUCC) effects on the green areas and the land surface temperature (LST) were also studied. The land cover characteristics were divided into five categories: a built-up area, an agricultural area, a bare area, a forest, and water. The LST was calculated using the thermal bands of spatial images taken from 2005 to 2020. The increase in the built-up area was mainly caused by the agricultural area decreasing by 11.90%. The average land surface temperature of the study area increased from 23.50 to 36.51 °C, and the region moved to a high temperature that the built-up area’s temperature increased by 1.5%. Still, the increase in vegetation cover was negative. From 2020 to 2050, the land surface temperature is expected to increase by 9.5 °C. The high-temperature areas moved into an aerial distribution, and the direction of urbanization determined their path. Urban heat island mitigation is best achieved through forests and water, and managers of urban areas should avoid developing bare land since they may suffer from degradation. The increase in the land surface temperature caused by the land cover change proves that the site is becoming more urbanized. The findings of this study provide valuable information on the various aspects of urbanization in Tianjin and other regions. In addition, future research should look into the public health issues associated with rapid urbanization.

1. Introduction

The rapid growth of urban areas worldwide has been observed over the past few decades [ 1 ]. The main factors contributing to urbanization are the lack of economic development and the increasing population [ 2 ]. Despite the slow growth of the global population, it is still expected that the number of people will continue to increase by around 2030 [ 1 ]. According to estimates, the world’s urban area is expected to grow by over a million kilometers by 2030 [ 1 , 3 , 4 , 5 , 6 ]. Urbanization is most prevalent in developing countries due to rapid economic development. China is one of the most prominent in the world regarding urbanization. It has been estimated that the country’s urban land area expanded at an annual rate of 13.3% [ 7 ].

Urbanization positively impacts people’s lives, as it allows them to improve their living standards and reduce their energy consumption. It can also help mitigate climate change by reducing vehicle miles travelled and greenhouse gas emissions [ 8 ]. Unfortunately, there are still negative impacts of urbanization. Due to the human activities that have occurred in the past few decades, the city has expanded. This process has caused both positive and negative effects [ 9 ].

Urbanization is a complex process involving multiple modelling variables and mechanisms involved in its development. The various aspects of this process must be thoroughly studied to understand its effects. One of the most effective ways to predict an urban area’s characteristics is through Land Use Cover change (LUCC) analysis [ 8 , 10 , 11 ]. A comprehensive simulation of the urban development process is necessary in today’s world [ 12 , 13 ]. With the help of spatial data, such as land area and development characteristics, urban models can be used to study the patterns of urbanization. These models can also simulate the conditions affecting the city’s development. Urban models use mathematical equations to describe the urban system [ 14 , 15 ]. They can also deal with the various factors that affect the development of a city. The study results are based on the interactions between different strategies and aspects [ 9 , 16 ]. Urban models are becoming more effective at predicting future changes in the LUCC due to the complexity of the process. They can use the available data and conditions to model the different factors affecting the city’s development. Numerous studies have been conducted on the use of LUCC in policy formulation and decision making.

However, the application of cellular automata and the Markov process is relatively rare. Tianjin is considered one of the most prominent cities in China that has experienced sustained urbanization, industrialization, and urbanization in China [ 17 , 18 ]. As a result of its ongoing development, many cities are expected to continue to grow [ 19 ]. It is essential that the cities’ LUCC change be studied and analyzed to determine its future trend [ 20 , 21 ]. This study was conducted to comprehensively analyze the various factors that have affected the city’s development. The study examined the LUCC change in Tianjin from 1995 to 2015. It first created five maps with different classifications at different points in time. The analysis revealed that many areas were converted into built-up areas. The model was then analyzed to create a set of dynamic variables for the Cellular Automata Model (CA) [ 11 , 22 , 23 ]. These variables were then used to project the LUCC change in the city from 2025 to 2050.

2. Materials and Methods

2.1. study area.

The city of Tianjin, the largest city on China’s northern coast, is straddled at 38°34′ N to 40°15′ N and 116°43′ E to 118°04′ E ( Figure 1 ), having a thousand square kilometers. It is regarded as the fifth-largest city in the country after Shanghai, Beijing, Guangzhou, and Shenzhen [ 24 ]. With a warm, temperate, semi-humid monsoonal climate, it is characterized by four distinct seasons during the year [ 25 , 26 ]. Over the past few years, Tianjin has experienced massive urbanization, with its population increasing from 12.99 million in 2010 to 13.86 million in 2021 [ 1 ].The city of Tianjin has a gross domestic product of about 240 billion yuan, making it one of the most prominent economic centers in China’s northern region [ 25 , 27 , 28 ]. It is an international port city and has experienced rapid urbanization over the past few decades. Due to rapid urbanization, large areas of land, such as forests, farmland, and meadows, have been converted into built-up areas [ 29 , 30 ].

Geographic location and characterization of the study area: ( A ) People’s Republic of China; ( B ) Beijing–Tianjin–Hebei (TBH); and ( C ) multispectral satellite image of Tianjin city.

2.2. Acquisition of Spatial Dataset

The United States Geological Survey (USGS) provided cloud-free images of the study area, which were taken from path 170 and series 053, through its website ( http://earthexplorer.com ) [ 6 , 9 , 10 , 13 , 17 , 25 , 31 ]. Due to the varying time of day and night in the study area, the data collected by the Landsat 5 Thematic Mapper (TM) and the Landsat 7 Enhanced Thematic Mapper (ETM) were used to create the LUCC map [ 13 , 18 , 32 , 33 , 34 ]. The data collected by the two satellites ( Table 1 ) were also used to calculate the Normalized Difference of Vegetation Index (NDVI) and Land Surface Temperature (LST).

The Landsat data used in this study are outlined in detail.

2.3. Methodology

An integrated workflow template ( Figure 2 ) was used to perform a series of steps. We began by processing the information sets in GEE to create a false colour positive (FCC) [ 10 , 11 , 25 , 35 , 36 ]. A georeferenced map of the outer boundaries of Tianjin was used to extract and mask the study area from all spatial ideas. The support Vector Machine (SVM) classification method was applied to improve the supervised classification results obtained from Landsat imagery [ 13 , 30 , 37 , 38 ]. Then, the LST was calculated to determine the time zones in the city [ 17 ]. A Pearson correlation analysis was performed based on the land cover, average LST, and percentage of greened and non-greened areas from 2005, 2010, 2015, and 2020 [ 8 , 31 , 39 , 40 ]. The CA-Markov model was used to forecast future trends for LUCC and LST in 2035 and 2050 [ 41 ]. All spatial statistical analyses and maps were created using ArcGIS 10.7, and ggplot2, corrplot and psych packages used in RStudio [ 42 ].

Flowchart of the methodology for the present study.

2.4. Land Use Cover Change (LUCC) Calculation

Landsat imagery (Landsat-5 TM & Landsat-8 OLI) was used to map the LUCC of Tianjin city for a four-time frame (2005, 2010, 2015, and 2020). The Support Vector Machine (SVM) classification algorithm in GEE was used to classify land use and areas [ 43 , 44 ]. Five types of LUCC were identified: built-up land, cropland, lowland, forest, and water body ( Figure 3 A). Built-up land included artificial structures such as buildings, roads, and other impervious surfaces. Water included rice fields, reservoirs, and rivers [ 28 , 35 , 45 ].

( A ) Shows the land use land cover area, and ( B ) proportional changes of LUCC during 2005-2020.

At specified intervals, GEE was used to assess the accuracy of the classification results. Field reference points were collected using a Google Earth explorer, which collected field reference average of 250 points for 2005, 2010, 2015, and 2020.

The classification accuracy of the signatures and images was evaluated by creating a confusion matrix consisting of rows and columns that refer to the categories derived from the image. The matrix rows are labelled with the reference values, while the columns represent the categories identified using the same criteria. The total number of entries that formed the main diagonal was then divided by the number of pixels. The Kappa coefficient was calculated using Equations (1)–(3) [ 1 , 2 , 46 , 47 ]:

where r = the number of rows in the error matrix; P ij = The proportion of pixels in a row “ i ” and column “ j ”; and P i = the fraction of the marginal sum of row “ i ”.

2.5. Calculation of Land Surface Temperature (LST)

The Landsat-8 thermal infrared sensor (TIRS) of bands 10 and 11 and the OLI sensor of bands 2–5 were used individually to convert the raw image into a radiance spectral image (SR) by following the equations ( Table 2 ) step by step.

Stepwise process for Land surface temperature (LST) determination.

where λ is the effective wavelength (10.9 mm for a thermal band in Landsat 8 data), σ is the Boltz–Mann constant (1.38 × 10 −23 J/K), h is the Plank constant (6.626 × 10 −34 Js), and c is the speed of light in vacuum (2.998 × 10 −8 m/sec).

2.6. CA-Markov Prediction Model Analysis

This model uses a stochastic Markov probability matrix to predict the transition from one state to another [ 14 , 43 , 50 ]. The study aims to analyze the various effects of urbanization on the land use and development of the city of Tianjin using a computer model known as a Markov chain model. This model was used to predict land use and development trends [ 13 , 36 , 51 ]. A conditional probability formula was used to estimate trend lines from Equations (4)–(6).

Because of Markov chain and cellular automata modelling, LUCC and LST’s future scenarios are calculated by projecting 2035 and 2050 using Terrset’s land use change modeler (LCM) (Clark Labs TerrSet 18.31).

3.1. Changes in LUCC between 2005 and 2020

According to the LUCC distribution values for 2005, 2010, 2015, and 2020, the built-up area in cities has increased ( Figure 3 ). Built-up area increased from 15.46% in 2005 to 17.80% in 2010, 19.56% in 2020, and 22.72% in 2050. In 2005, the study area included 18.43% of the lowlands; this number decreased to 12.52% by 2010, 11.89% by 2015, and 10.21% by 2020. Arable land increased rapidly from 26.10% in 2005 to 28.95% in 2020, while other land decreased from 26.47% to 26.10%. Increasing migration from villages to cities has led to an expansion of cultivated land outside prime locations. The cultivated area decreased from 10.72% in 2005 to 7.98% in 2020. Water covered 1.21% of the site in 2005, 0.92% in 2010, 0.87% in 2015, and 0.68% in 2020. An assessment of land use changes during 2005–2020 showed that farmland in the northeastern study area was converted to urban areas (mainly industrial areas). Between 2005 and 2020, built-up urban land and cropland increased by 15.45% and 1.64%, respectively, while lowland land decreased by 13.73%. These results show that about 11.45% of the lowlands have been converted into built-up areas. The LUCC changes were classified into five categories LUC with corresponding definitions ( Table 3 ).

Land use cover (LUC) statistics in 2035 and 2050.

The results of all studies show that urban built-up has changed significantly over two decades. In recent decades, Tianjin has gone from a village to a city residential settlement. This transition happens between agricultural land to residential areas. Urban growth and LST are sensitive to accuracy assessment [ 29 ]. According to [ 52 , 53 ], a method was defined for assessing the accuracy of the classification of maps. According to the LUCC maps, the overall accuracy was 84.39% in 2005, 90.43% in 2010, and 94.11% in 2020. Kappa coefficients for the LUCC maps were 0.79, 0.87, and 0.92. The kappa coefficient should be greater than 0.75 or 0.80 to show compatibility between the classification and the reference data [ 54 ]. The United States Geological Survey (USGS) recommends using Landsat satellite images for LUCC mapping if the accuracy level is 85% [ 55 ]. Our accuracy evaluation results are consistent with those recommended in the literature.

3.2. Relationship between LUCC and LST

LST is significantly affected by land use changes (LUCC). The number and distribution of hotspots increase with LUCC types (especially urban expansion) [ 56 ]. A map of LST distribution was created using Landsat TM/ETM+/OLI imagery for the study area ( Figure 4 ). There were temperature variations from 21 °C to 43 °C in 2005, 21.8 °C to 44.3 °C in 2010, 22.1 °C to 44.9 °C in 2010, and 22.5 °C to 45.9 °C in 2020. During 2005–2020, built-up urban areas had the highest average temperatures, followed by lowland, cropland, vegetation, and water. In 2005, all LUCC categories had the most elevated average temperatures ( Figure 5 A). In 2005, urban built-up areas had an average LST of 38.43 °C, 38.99 °C in 2010, 41.86 °C in 2015, and 44.80 °C in 2020. During 2005–2020, the temperature in built-up urban areas LST decreased by 4.12 °C but increased by a maximum of 6.82 °C from 2010 to 2020. Lowlands had the second-highest LST for all LUCC categories during the study period. The LST for wasteland decreased by 3.38 °C from 2005 to 2010 but did not change significantly between 2010 and 2020. The LST for cropland was 31.04 °C in 2015 and increased to 31.98 °C, 32.63 °C, and 33.75 °C in 2010, 2015, and 2020, respectively.

( A ) Maps for land use land cover (LUCC) classes: (i) water, (ii) vegetation, (iii) forest, (iv) urban, (v) barren land, (vi) cropland; ( B ) land surface temperature (LST) was divided into five thermal categories: (i) <20 °C, (ii) 20–25 °C, (iii) 25–30 °C, (iv) 30–35 °C, and (v) >35 °C of Tianjin between the study period of 2005 to 2020, at 5-year intervals.

( A ) Distribution of land surface temperature, and ( B ) proportional changes of LST during 2004–2019.

From 2005 to 2020, the temperature of cropland LST decreased by 3.85 °C, while it increased by 15.35% in developed areas. Vegetated areas recorded a decrease of 5.32 °C between 2005 and 2010 LST but an increase of 6.83 °C between 1999 and 2015. All LUCC categories recorded the lowest LST in 2010, and the average temperature in water bodies and vegetated areas was the lowest overall. According to statistics from LST for 2005–2020, the maximum difference between urban areas and water bodies is 14.35 °C.

LST has remained relatively stable between 2005 and 2020. These areas are also referred to as lowlands. Compared to urban areas, lowlands have a higher LST value. Our study came to similar conclusions. High LST values may be found in these areas due to the soil composition (sand, clay, etc.). The average daily air temperature may influence the LST values at the satellite imagery data on the day the satellite imagery was taken rather than the spatial values of the land use classes. To verify that LST results calculated from the Landsat TM/ETM heat band are comparable to actual field temperatures, temperatures of the various LUCC properties must be measured from field observations [ 57 ]. Considering the values reported at LST, the daily mean air temperatures of the reported data (the daily mean air temperature on 19 June 2005 is 27.5 °C; on 10 July 2010, it is 23.03 °C; on 23 July 2015, it is 26.86 °C; and the daily mean air temperature on 11 August 2020 is 29.53 °C) are all parallel to each other (See Figure 6 ).

Spatial distribution map for change detection of ( A ) LUCC (km 2 ) and ( B ) LST (°C) during 2005–2020.

Pearson’s correlation analysis shows LST is statistically associated with populated/developed areas. Even though LST is bad for water and plants and does not have much to do with them, it is strongly linked to forested areas. In the same way, LSTs in cities have a negative and insignificant effect on water and plants. LST and urban/built-up areas have a significant and favourable relationship, as shown by the simple correlation coefficient [ 51 ]. In urban areas, a temperature rise may also be caused by the construction of new buildings, highways, businesses, and industrial regions. Negative and insignificant correlations are observed with barren land, while optimistic and negligible correlations are marked with arable and cropland. Pearson correlation analysis results are reported for all LUC variables and LST indices ( Figure 7 ).

Pearson correlation analysis of land use change and land surface temperature during 2005-2020.

3.3. Variations of LST Changes over Different LUCC

We estimated the mean LST distributions for LUCC classes over 2005–2020. During the study period, mean values of LST increased significantly in all LUCC classes, but matters of LST were substantially higher in built-up areas and bare ground. The importance of LST in the built-up area increased from 28.86 °C to 37.23 °C between 2005 and 2020, while in the empty ground area, they increased from 21.56 °C to 25.01 °C. Over the past two decades, the average LST distribution in built-up and bare-ground regions has risen by about 9 °C and 4 °C, respectively. The LST distribution in water bodies and vegetated areas have also changed. In 2000, the mean LST for vegetated areas was 21.31 °C, but it is expected to reach 25.98 °C by 2020. The LST of water bodies increased from 20 to 24.45 °C. The following figure ( Figure 8 ) briefly describes the changes in LUCC types and their relative impacts on land surface temperature.

Shows classified maps of LUC and LST for the year 2020, and predicted maps of 2020, 2035, and 2050.

3.4. Validation of Predicted LUCC and LST Scenarios

To validate the accuracy of the predicted values, we first used the CA-Markov model to estimate the LUCC and LST for 2020 ( Figure 8 ). Based on various kappa parameters, the predicted and estimated maps were compared using the land use Change Modeler in Clark Lab’s Terrset software. The average error value for all parameters during the comparison was about 12.86%, and all kappa parameters, percentage of accuracy, and total kappa values were above 0.80.

3.5. Predicted LUCC for 2035 and 2050

We could predict the scenario for 2035 and 2050 based on the classified maps for the study period. According to the predicted LUCC map, the growth of urban areas will be concentrated by 37% in the northwestern and central regions if the trend of the building continues without planned actions. Urban areas will replace the lowlands and vegetation cover. Vegetation cover has decreased by 9.62% from 12.82% in 2020. Based on the study scenario, LUCC would face a 20.51% increase in developed land, followed by a significant decrease in lowlands, vegetation cover, and water bodies of 10.87%, 9.62%, 8.32%, and 2.45%, respectively ( Figure 8 ). The category-wise land use statistics for the forecast years are shown in the following table: Ecosystem services, urban health, and thermal characteristics may be affected by decreased vegetation cover and increased urbanization. If unplanned urban expansion continues, the environmental, economic, and medical problems will increase significantly. A proper land use plan, the protection of water bodies, and the reforestation of forests are needed to make Tianjin city more environmentally sustainable.

By forecasting LST for 2035 and 2050, the simulation showed that higher temperatures will occur in the built-up areas in the northwest and central parts of the country ( Figure 8 ), ranging from 41.56 °C to 44.34 °C in 2035 and 2050, respectively. We divide the temperature zone into five classes to estimate how much area is covered by each temperature range ( Table 4 ). Based on the projections, LST has increased over the past two decades (2005–2020), with urban areas influencing the prevalence of LST. UHI effects will increase as urban areas and vegetation cover decrease. It would be possible to explain the temperature increase without urbanization by climate change, greenhouse effects, and surface features. The LST prediction highlights the real risks of the temperature rise in the trend, including higher UHI effects. A combination of energy use, greenhouse gas emissions, and air pollution contribute to the UHI effect. It threatens aquatic systems (rivers, lakes, ponds, streams, and oceans) and human health. Human health is primarily harmed by increased greenhouse gas emissions, which affect urban health and reduce the urban environment’s sustainability [ 58 ].

Land surface temperature (LST) statistics in 2035 and 2050.

3.6. Limitations of the CA-Markov Model

The prediction of LUCC and LST can be improved using the CA-Markov model if the previous LUCC and LST patterns are consistent. As a result, CA-Markov models do not provide accurate spatial predictions for raster datasets [ 59 ]. Since influential factors can be directly determined between CA-Markov and other factors, CA-Markov is based on a probability matrix [ 60 ]. Given the relative importance of the different variables in identifying the most important variables, it is essential to note that the CA-Markov model generates training patterns and automatically begins training after receiving inputs from the strata. The input parameters are not individually weighted according to established standards [ 61 ]. Since urbanization, the loss of green space and increase in surface temperatures are primarily influenced by human activities and conscious decisions at regional to metropolitan scales; it is impossible to predict them accurately. It is essential to recognize that dynamic models have some limitations. Still, they help develop hypotheses and make decisions about changes in land cover or surface temperatures in any given area, regardless of their rules. In recent years, LUCC and LST variability and predictive maps have emerged as one of the best tools for managing and mitigating vital natural resources.

4. Discussion

Tianjin’s rapid urbanization and development between the 1990s and 2020s significantly altered the LUCC landscapes caused by farmland separation and reduced total vegetation cover [ 25 , 33 ]. The city’s urban development also resulted in the establishment of new industries and residential areas. Rapid vegetation cover loss affects an area’s natural cooling effect [ 29 , 62 ]. Some factors contributing to this phenomenon are vegetation shading and transpiration. To amplify this, LST and NDVI have shown that VC, due to its cooling effect, serves as a sink in an urban heat island [ 11 , 63 ]. Rapid vegetation cover loss has several consequences for an area’s natural cooling effect. It has the potential to eventually eliminate the processes that regulate surface transpiration and evaporation [ 11 , 17 , 56 , 64 ]. Urbanization leads to distorted construction, reducing soil infiltration and increasing surface runoff. As a result, the water table and groundwater table decrease. Evapotranspiration is not adequately realized due to these two factors. Climate change leads to a deterioration of the water balance [ 49 ]. Climate variables such as daily maximum and minimum temperatures are affected by changes in land use. Surface albedo changes due to changes in land use. Therefore, land use changes disturb the balance of Earth’s radiation [ 65 ]. An important factor in reducing air temperatures is the conversion of wetlands to agricultural land with high albedo [ 66 ].

Although the impact of this phenomenon on the LST of various types of plants is less than that of urban tree cover and gardens, studies have shown that it still contributes to the overall reduction of the area’s natural cooling effect [ 32 , 67 ]. The impact of various types of urban vegetation, water bodies, and forests on the LST varies according to their proportional area [ 23 , 37 ]. In urban areas, vegetation plays a vital role in controlling or mitigating temperature. Evaporation from urban water bodies contributes to moisture accumulation in the surrounding air. According to studies, these bodies regulate the LST in residential areas. It is also known that urban areas contribute to the development of intricate heat flows within these regions [ 46 , 68 ]. Various private and public entities have worked together to revitalize large tracts of land for industrial, commercial, and residential development. Traditional wooden structures have been demolished and replaced with tall structures made of non-evaporative materials such as glass, concrete, and aluminium. These materials can directly impact heat flows in urban areas [ 8 , 10 ]. According to studies, urban areas in China are more vulnerable to severe LST than rural areas. LST has risen due to the government’s decision to convert agricultural and forest land into urban areas [ 38 , 69 ]. The government has relocated factories and businesses to the outskirts of cities to improve their efficiency. These facilities are typically found in developed areas. Before the development of urban areas, forests and vegetation were regarded as buffer zones between rural and urban areas, absorbing excess heat generated by factories and automobiles [ 3 , 29 , 40 ]. According to the scientific literature, the cooling effect of LUCC is well-matched to the expected warming effect caused by the physical interaction of the Indian region and its surroundings [ 32 ]. For example, the maximum cooling contribution from forested areas is 0.27%, while the minimum cooling effect is 0.06%. The most negligible difference between the surface temperature and the impervious surface is the primary reason why vegetation contributes the least to the cooling effect. The greatest cooling effect, on the other hand, is observed when forested areas are converted into water bodies. This is due to the fact that the contribution of land cover to cooling is negligible in various areas, such as urban areas, water bodies, and vegetation. The results of the study revealed that the built-up area in the southeastern and central port areas will continue growing. The paper discussed the various effects of the LUCC on the Tianjin city’s development. The study used the CA model and Geographic Information Systems to analyze the data. The results of the analysis helped improve the Tianjin city’s planning process. In addition, the paper discussed the use of remote sensing tools for improving the urban planning process.

5. Conclusions

The objective of this study was to analyze the influence of LUCC on land surface temperature (LST) in a large urban area of Tianjin. Data from RS were used to observe the area’s various socioeconomic and development parameters. The study also used the CA-Markov model and Pearson correlation coefficient to evaluate the contribution of landscape dynamics to temperature. A 5.94% increase in built-up area was found to increase the temperature by 1.5%. However, the increase in vegetation cover by 10% showed a negative correlation. In addition, the study concluded that LUCC has a cooling effect of about 1.40 °C in the city. The average warming effect of LUCC on the UHI is about 0.5%.

On the other hand, the cooling effect of LUCC compared to the shifts in the reverse direction is 0.11%. The positive contribution of LUCC to the UHI was higher than the negative one. Urban development and infrastructure planning should be further targeted to minimize the impacts of climate change. In addition to improving water bodies and parks, other measures, such as the establishment of green spaces and linear planting of woody plants, should also be implemented. The study found that further research is needed to analyze the impact of land use change on the climate of regions and cities. As more areas are affected by climate change, the government and private sector must work together to develop effective cooling strategies. Environmental education should be made accessible to promote the development of ecological resources. This needs effective urban planning and green policies to address the increasing thermal stress. In addition, a quantitative analysis of these parameters needs to be conducted. Although the study found that urbanization directly impacts land surface temperature, it is not yet clear how the effects of this process are related to the other factors. The practical application of the study provides essential guidance for urban landscape planning. It shows how landscape connectivity between impervious and green areas can affect LST. Future research should also address infrastructure stress and public health issues associated with rapid urbanization.

Acknowledgments

The authors wish to express his appreciation and gratitude to the anonymous reviewers and editors for their insightful comments and suggestions to improve the paper’s quality.

Funding Statement

Reconstructing the Architecture System based on the coherence mechanism of “Architecture-human-environment” in the Chinese context, Key project of National Natural Science Foundation of China, grant number 52038007, 2021-01-2025-12.

Author Contributions

Conceptualization, N.U., M.A.S., Y.Z. and Y.H.; data curation, N.U., M.A.S., M.D., S.G., S.T. and T.Z.; formal analysis, N.U., M.A.S., S.T., Z.K. and M.D.; funding acquisition, Y.H. and Y.Z.; methodology, N.U., M.A.S., S.T., I.A.K. and S.G.; project administration, Y.H. and Y.Z.; software, N.U. and M.A.S.; supervision, Y.H. and Y.Z.; visualization, N.U., M.A.S., T.Z., Z.K. and Y.H.; writing—original draft, N.U. and M.A.S.; writing—review and editing, N.U., M.A.S., I.A.K. and Y.H. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Data availability statement, conflicts of interest.

The authors declare no conflict of interest.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.