6 - China

Introduction

Urban Heat Island (UHI) refers to the phenomenon where urban areas experience higher temperatures compared to their rural surroundings, significantly influenced by human activities. This temperature differential impacts energy balance, potentially affecting the atmosphere-land-water cycle, biodiversity, and even climate change. It leads to discomfort and health risks for urban residents, increasing the incidence of heatstroke, cardiovascular, and respiratory diseases, particularly among vulnerable populations like children and the elderly. The phenomenon also increases electricity expenditure due to higher cooling demands, posing challenges for environmental protection.

The spatial distribution of heat within urban environments is inhomogeneous due to differences in albedo, surface roughness, energy consumption, emissions, and sky-view factors influenced by high-rise buildings. These urban characteristics alter natural surface energy and radiation balances compared to rural areas. High thermal mass and insulation in buildings exacerbate the UHI effect, while global warming further elevates average global temperatures, intensifying heat in cooling-intensive urban areas.

Research on urban heat islands is categorized into:

Canopy Urban Heat Island (CUHI): Small-scale studies using measurement stations for temperature data to calculate the UHI effect, exploring detailed underlying surfaces with statistical methods.
Surface Urban Heat Island (SUHI): Utilizes remote sensing for urban heat island research, providing a wide detection range and comprehensive spatial information.

Studies using large remote sensing datasets have revealed disparities in Land Surface Temperature (LST) not only between urban and natural categories but also within urban categories. Factors like urban green space (UGI) and average building height (ABH) significantly impact SUHI intensity, with UGI's contribution decreasing as building density and height increase, while ABH shows the opposite trend.

Data and Methods for Urban Extent Differentiation

Differentiation between urban and background extent is achieved through:

City social-economic data: Defining population agglomerations via Metropolitan Statistical Areas (MSAs).
City spatial data: Generating Built-up Intensity (BI) maps using land use data.

Three Issues in Existing UHI Research

Urban heat island patterns analysis: Current methods like geographically weighted regression, Geodetector, and random forests have limitations in visualization and reflecting changes in urban heat island patterns across multiple time periods.
UHI or Land Surface Temperature (LST) studies: Relying on the mean value of continuous time series data may weaken the representation of the urban heat island phenomenon. Geophysical and climatic phenomena often lack specificity when based solely on results from a singular point in space or time.
Boundary of urban and rural areas: The boundary is critical for UHI results, and many studies do not account for buffer zones between urban and rural areas.

Study Objectives

This paper aims to address these issues by:

Quantifying changes and patterns in the surface urban heat island in order to provide accurate quantitative methods for city planners and governments.
Analyzing the heat island distribution changes in Beijing and Zhengzhou over the past 20 years.
Revealing the differences in heat island patterns brought about by urban-rural changes.
Proposing methods to address the weaknesses in representation of the UHII when relying solely on mean values or data from singular points in space or time.
Proposing an extended solution aimed at addressing a oversight in traditional urban heat island effect calculations: the insufficient consideration of buffer zones between urban and rural areas.

Study Area

Beijing: A mega-city with rapid growth, its population increased from 13.636 million in 2000 to 21.843 million by the end of 2022. The urbanization rate rose from 54.9% in 1978 to 85.9% in 2018.
Zhengzhou: A megacity located in central China, it is experiencing rapid urbanization and economic growth, benefiting from China’s policy of prioritizing the development of provincial capitals.

Beijing and Zhengzhou are representative of first-tier and second-tier cities in China, respectively. They have been selected for comparative analysis to explore the SUHII (Surface Urban Heat Island Intensity) response under varying urban expansion rates.

Data Sources and Pre-processing

Globe Land Collection 30 (GLC30): 30 m resolution land cover data is used to extract built-up areas. It has a spatial accuracy of 82.4 % and it is developed by the National Geomatics Center of China. It is created using multispectral imagery from the United States’ Landsat satellites (TM5, ETM+, OLI) and China’s HJ-1 environmental disaster mitigation satellite. GLC30 is distributed through the website (http://globeland30.org).
MOD11A2: Provides an 8-day combination of MODIS land surface temperature products with a 500 m resolution.
MOD13A2: Provides Normalized Differential Vegetation Index (NDVI) values per pixel at 1 km spatial resolution.

MOD11A2 and MOD13A2 are distributed by the Level-1 and Atmosphere Archive and Distribution System (LAADS) and are available through the website (https://ladsweb.modaps.eosdis.nasa.gov).

Addressing Water Body Interference

Water bodies have a higher specific heat capacity, causing significant temperature differences with the surrounding surface, which affects the accuracy of UHI calculation. In this study, LST pixels in the water body area were removed to improve the accuracy of SUHI evaluation.

Reanalysis Dataset

ERA5-Land: A reanalysis dataset offering a coherent perspective on meteorological parameter changes over multiple decades. It provides temperature data at a finer resolution than ERA5. It provides air temperature 2 m above the surface of land, sea or inland water, within urban areas inhabited by humans, offers a more accurate reflection of the effects of urban heat island on human life.

Urban and Rural Extent Delineation

30 m resolution land cover data (GLC30) of 2000, 2010, and 2020 are divided into city urban (U), peri‑urban (P), and rural (R) areas by using:

City Clustering Algorithm (CCA)
Boundary Generation Algorithm (BGA)

The imaging cycle of GLC30 products is divided into three time periods:

P2000 (2000.3 to 2005.2)
P2010 (2008.3 to 2013.2)
P2020 (2018.3 to 2023.2)

Temperature of cluster (TC), temperature of background (TB), UHII and Temperature Difference (DT, means temperature of the cluster minus the temperature of the background) are calculated in three periods: P2000, P2010 and P2020.

Fourier Transform Models for UHI Patterns

Third order Fourier series transformation model used to build the UHI patterns to disclose the UHII change response to urban expansion due to rapid urbanization.

The formula is shown as Eq. (1):

$F(t) = a0 +∑3 _{n=1} (ancos \frac{2πnt}{ P} + bnsin \frac{2πnt}{ P} )$

where $F(t)$ means either $TB$ or $DT$ , $P$ represents 12 months, $t$ the times t; t = 1, 2…60, $a0$ , $an$ , $bn$ are the Fourier coefficients, $a0$ is especially referred to as the mean of $F(t)$ . After the coefficients are obtained, fitted values of $TBi$ and $DTi$ are used to build the fitting curve.

The UHII is calculated by the following formula:

$UHII(p) = T(p) − \frac{∑n _{i=1}T(i) p + ∑m _{j=1}T(j) r }{n + m}$

where $UHII(p)$ is the UHII of p-th pixel, $T(p)$ means the temperature of p-th pixel, n and m refer to total pixels of peri‑urban and urban. $T(i) p , T(j) r$ is the temperature value of the i th, j-th pixel in the peri‑urban, rural region, respectively.

Results

The results from the experiments are discussed in the following sections.

Three Stages Rural–Urban Dynamic Transition

The average area of Zhengzhou increased by 2.20 times from 2000 to 2010, and 2.05 times from 2010 to 2020. The average area of Beijing increased by 1.42 times from 2000 to 2010, and 1.53 times from 2010 to 2020. The cluster area of Zhengzhou in 2020 is roughly the same as that of Beijing in 2000.

Urban, peri‑urban, and rural areas of Zhengzhou have the highest mean LST values in June, increasing every year (36.32 ◦C, 37.17 ◦C, 37.67 ◦C), while the highest temperature in Beijing in each study period was distributed in different months. In P2000 and P2010, the highest temperature of Beijing was in July (37.69 ◦C, 36.54 ◦C), but in P2020, it was in June(36.66 ◦C). The majority of the high-temperature period (most are above 25 ◦C) is concentrated between April and September in Zhengzhou and Beijing, while the moderate-temperature period (most are between 15 ◦C and 25 ◦C) is concentrated in March and October and the low-temperature period (most are below 15 ◦C) is concentrated in January, February, November and December.

For all the points, the average surface temperature ratio (P2010/P2000) for the first 10 years and the average surface temperature ratio (P2020/P2010) for the next 10 years in Zhengzhou are 1.055 and 1.053, respectively. In Beijing, they are 1.006 and 1.023, respectively. It can be observed that both cities are experiencing warming trends internally. Similarly, the surface temperature ratios for the P region and R region in the last 10 years are higher than those for the previous 10 years, and this trend is also observed in Beijing.

In all three time periods, NDVI values in all areas of Beijing increased over time, while in Zhengzhou’s U region, they initially decreased and then increased. Beijing’s NDVI experienced rapid growth from P2000 to P2010. At any given time, the NDVI distribution in the U area is always lower than in the P area, which in turn is consistently lower than in the R area.

UHII Spatial and Temporal Variation

The annual mean values of UHII in Zhengzhou were 1.474, 1.107, and 1.046, respectively, and experienced a continuous decline, whereas the annual mean values in Beijing declined and then increased, and were very close to each other at P2000 and P2010, reaching 1.126, 1.114, and 1.529 in the three time periods, respectively. the mean values of UHII in spring in Zhengzhou were higher than those in autumn, whereas the opposite was the case in Beijing.

The overall UHII distribution in Beijing is relatively stable, while Zhengzhou shows a large variation in all seasons. The UHII mean values for all years in the inner-city summer in Zhengzhou were 2.483 and 2.867 for Beijing, respectively, indicating that the heat island intensity in Beijing was higher than that in Zhengzhou during this summer. The mean value of UHII in winter in Zhengzhou is − 0.142, while the mean value of UHII in Beijing cluster area is − 0.005, which indicates that the urban cold island phenomenon in winter is more prominent in Zhengzhou than in Beijing. It is worth noting that the UHII value within the city of Zhengzhou in spring is 1.658, which is significantly higher than 0.858 in Beijing.

Intra-annual UHI Patterns

The Fourier fitting curve of Zhengzhou is spiral in 2000 and 2010 periods and oblate in 2020 period, while Beijing is oblate in all periods. This transformation of Zhengzhou in the 2020 period can be explained by the higher UHII observed after September as compared to the period between January and April, which led to a counterclockwise curve. The general fitting curves of Beijing during the time periods of 2000, 2010, and 2020 show an upward trend, indicating the rise of the UHI effect in tandem with urban expansion.

In Zhengzhou, the DN-DT curves show a bimodal shape and the peak values of them are 2.08 ◦C (May) and 2.69 ◦C (July), 2.24 ◦C (May) and 2.85 ◦C(August), 1.15 ◦C (May) and 2.44 ◦C (July) in 2000, 2010 and 2020 periods. This does not apply to Beijing, while there was a unimodal shape in Beijing, which was 2.61 ◦C(August), 2.61 ◦C (July), 2.97 ◦C(August) in the same times. Similarly, we observe that Zhengzhou has a bimodal DN curve with peak values of 0.167 (March) and 0.260 (July), 0.184 (April) and 0.257 (July), 0.112 (April) and 0.168 (July), while Beijing still corresponds to a unimodal DN curve with peak values of 0.219 (August), 0.209 (July), and 0.191 (August).

Profiles of UHII and NDVI

The expansion of a city does not necessarily lead to NDVI decrease in the internal and external of city. The NDVI of the urban area of Beijing showed an increasing trend in these three time periods, but there was no specific trend in the suburbs. This may be related to the urban greening measures implemented in Beijing. However, this cannot be used to explain Zhengzhou, which had the lowest NDVI in 2010 and the highest NDVI in 2020 period in the horizontal direction of urban center.

Both Beijing and Zhengzhou exhibit a north-south temperature gradient. Vertically, Beijing shows higher temperatures in the central region and lower temperatures on both sides, with the eastern side being higher than the western side. In Zhengzhou, the temperature remains higher in the central region horizontally, but there is no consistent rule in the temperature variations of the suburban areas on the eastern and western sides.

The UHII difference during Beijing’s summer has been increasing, while the UHII range showed a reduction in 2010 period during the winter. Beijing exhibits a noticeable UHII distribution in the horizontal profile, with higher UHII in the central region and lower UHII on both sides. In the suburban areas, there is a consistent trend of lower UHII in the western region compared to the eastern region throughout the year, which might be attributed to the presence of mountainous areas in the western part of Beijing.

LST and Air Temperature Comparison

The surface temperature is usually higher than the air temperature, especially on the ground that receives direct solar radiation, while the air absorbs the long-wave radiation emitted by the ground. The atmosphere exchanges energy with the surface through processes such as convection, evaporation, transpiration, and heat conduction. The coefficient of determination for Zhengzhou in urban, peri‑urban, and rural areas is 0.8676, 0.8509, and 0.8537, respectively, while for Beijing, it is 0.9046, 0.8903, and 0.8833. It can be seen that both Zhengzhou and Beijing have better fitting effect and linear relationship.

The coefficient of determination for both Zhengzhou and Beijing are lower than that for the overall. In Zhengzhou, the four seasons of the coefficient of determination were 0.3345, 0.4565, 0.2724, 0.5677, in Beijing, 0.6828, 0.4357, 0.5088, 0.7588. This suggests that the relationship in air temperature and urban land surface temperature varies greatly across seasons and may not exhibit a significant linear relationship between different cities, while the overall trend indicates a significant linear correlation between the two variables. Hence, it is crucial to perform analyses of urban heat islands across different seasons in cities.

Discussion

Analysis of Surface Temperature Changes and Vegetation Coverage Trends in the Urbanization Process of Zhengzhou and Beijing

Areas with denser vegetation cover have a lower heat island index compared to areas with less vegetation. The changes in the NDVI index of the two cities also reflect this.

Beijing’s NDVI value has increased over time, indicating positive efforts towards greening and ecological restoration. In contrast, the NDVI in the urban area of Zhengzhou first declined and then increased, possibly reflecting the impact of early urbanization on green spaces and subsequent restoration measures.

The Pattern of Urban Heat Island Effect in Zhengzhou and Beijing

The two cities exhibited different patterns, with Zhengzhou’s being more complex. Zhengzhou’s curve transformed from spiral to oblate, while Beijing’s remained oblate. The DN-DT curve, show that the UHII distribution and trend of change in both cities were complex and showed seasonal differences. Zhengzhou’s DN-DT curve showed a bimodal shape, while Beijing’s was unimodal.

Exploring Environmental Sustainability

Urban expansion not necessarily leads to a decline in NDVI within and around cities. During these three periods, the NDVI of Beijing’s urban area showed an upward trend, while the suburbs did not show a specific trend. In contrast, Zhengzhou had its lowest NDVI during the 2010 period and its highest in the 2020 period. Zhengzhou’s performance may be related to the change in its economic development policy, as in the 2020 period, Zhengzhou increased its efforts to build an environmentally friendly city and publicized the Ecological Protection and Construction Plan of Zhengzhou Metropolitan Area (2020–2035).

Research contributions and limitations

The study reveals two major contributions.

First, it shows that the response of Urban Heat Island (UHI) change to urbanization varies in the three phases.
Second, introduced clustering of the profile results, and the results suggest that crop cultivation has a significant impact on the intra-annual structure of the UHI.

Limitations:

The MODIS LST product in use is of low resolution and does not have the ability to capture the internal composition of the three urban areas
A quantitative explanation of the various factors responsible for the marked contrasts between the two patterns is lacking

Conclusion

This investigation accentuates the distinct growth models of the two cities, incorporating peri‑urban areas into the scope of urban heat island research which fills a gap in this area of study. Comparing urban spatial changes across three stages, this study reveals the differences in urban heat island patterns of cities in China with two speeds of development over the past twenty years. This study provides support for urban thermal environment planning.