1000/1000
Hot
Most Recent
Urban Heat Island (UHI ) studies have been conducted for over 200 years, since the first conceptualization by Luke Howard in 1818. Generally, an urban heat island (UHI) is an urban area or metropolitan area that is significantly warmer than its surrounding rural areas because of human activities. The temperature difference is usually greater at night than during the day and is most apparent when winds are weak.
Urbanization is known to have substantial impacts on landscapes and ecosystems [1][2][3][4], and urban inhabitants are expected to reach 70% of the world population by 2050 [5]. Moreover, the nature of urban development has been changing from a single city model to a group of cities (urban agglomeration) worldwide. Urban heat island (UHI), urbanization, and climate change are increasingly interconnected, resulting in several environmental consequences (such as heat stress, biodiversity loss, fire risk, warming water due to run off, and diminished air quality) at both local and regional levels [2][6][7][8][9]. Such UHI related impacts are also called UHI regional impacts (UHIRIP). Generally, UHI research includes data from two major sources: Air temperature data that are observed by weather or climate stations and remotely sensed data to observe UHI through land surface temperature. Before the availability of remotely sensed data, UHI was widely observed in the field, with the first scientific observation of UHI in 1833 [10]. Field observations of UHI continue to be a critical source of training and validation data [11][12]. These observations, along with modeling studies, continue to help unravel the factors that are responsible for UHI development, and are providing a basis for the development and application of sustainable adaptation strategies. Communicating scientific knowledge quickly and effectively of UHI and UHIRIP to architects, engineers, scientists, and planners could help inform urban design and decision making. Remotely sensed data have been used to observe UHI and UHIRIP on environments, ecosystems, human health, and economics in urban and non-urban areas for decades. Remote sensing offers the benefits of long data archives, repeated observations, efficiency, and multiple temporal and spatial resolutions. UHI studies using remotely sensed data have been published for hundreds of cities worldwide [6][7][13][14][15][16][17][18][19]. Remotely sensed data provide highly efficient, long-term, and broad-scale information for assessing UHIRIP. However, studies integrating high spatial resolution imagery (e.g., Landsat at 30 × 30 m and ECOSTRESS at 70 × 70 m) from multiple sensors to evaluate UHI and UHIRIP across a time series have been uncommon. Challenges to such studies include image frequency and calibration, cloud contamination, and the need for large storage and high-performance computing capabilities [20][21]. Early generations of broad-scale UHI assessment using remote sensing often poorly represented the spatial and temporal variance in UHI, especially at the urban and non-urban interface. As the resolution of algorithms and satellite imagery improved and interest in UHIRIP grew, researchers sought better representations of UHI. Initially, this took the form of modifications based on surface physical characteristics such as roughness length, albedo, thermal conductivity, and thermal diffusivity [22][23]. Many studies have been conducted to understand the urban thermal climate or the potential for heat island mitigation using this framework of simplified algorithms [24][25][26]. In more recent efforts, researchers have incorporated more sophisticated parameterization schemes that have included distributions of demography, policies, and behavior of government; ecological variables and ecosystem services; land use and land cover change (LULCC) patterns; and social and economic factors to represent the complicated impacts of UHI [27][28][29][30][31][32][33][34][35][36].
Historically, the study of UHI using remote sensing data, often Landsat data, was mainly based on comparing images at two different times using the bitemporal approach [37][38][39]. Although the bitemporal approach is mathematically simple and does not need large amounts of data, it is less useful than a time series approach that is able to provide a more comprehensive understanding of the complexity of UHI. Most early research [17][40][41][42] in UHI focused on cities or urban areas, and often ignored the urban and non-urban interface at regional scales. In recent decades, the cost of data storage has dramatically decreased, and we have witnessed an overwhelming increase in computing power and open source software that provide the foundations for time series analysis using higher resolution thermal data from satellite archives. Some studies used Landsat time series to detect historical changes [20][43][44][45][46], but few have focused on UHI and its interaction with land use and land cover (LULC) dynamics. A research team at the USGS Earth Resources Observation and Science (EROS) Center recently developed the Land Change Monitoring, Assessment, and Projection (LCMAP) project [47], which is produced with Landsat Analysis Ready Data (ARD) [48] and land surface temperature (LST) data. LCMAP data provide the potential to use Landsat LST data to analyze UHI in urban agglomerations, as well as the urban and non-urban interface at local, regional, and global scales.
The large amount of heat generated from urban structures and pavements, as they absorb and re-radiate solar radiation, as well as the heat from other anthropogenic sources, are the main causes of UHI. These heat sources increase the temperatures of an urban area compared with its surroundings, which is known as UHI intensity (UHII). Traditionally, regardless of the methodology employed, whether it refers to (1) differences between two fixed observatories, one urban and another peripheral or non-urban; (2) mobile urban transects; or (3) remote sensing analysis, UHII provides a value of thermal differences between contrasted points, sectors, or areas, one urban and another that could be termed non-urban. Thus, the intensity of the UHI is seen in the temperature difference expressed at a given time between the hottest sector (areas) of the city and the surrounding non-urban space. The intensity of the heat island is the simplest and most quantitative indicator of the thermal modification imposed by the city upon the territory in which it is situated and of its relative warming in relation to the surrounding rural environment. The intensity could be defined for various time scales and geographical locations [49][50].
Applying theories of landscape ecology [51], UHI studies focus on moving from static spatial structures of urban thermal patterns to the change dynamic of spatial patterns and processes of urban thermal characteristics. The spatial structure of UHI patterns determines the processes of UHI impacts. Li et al. [52] simulated the urban climate of various generated cities under the same weather conditions. By studying various city shapes, they generalized and proposed a reduced form to estimate UHI intensities based only on the structure of urban sites, as well as their relative distances. They concluded that in addition to the size, the UHI intensity of a city is directly related to the density and the amplifying effect that urban sites have on each other. Their approach can serve as a UHI rule of thumb for the comparison of urban development scenarios. Ramírez-Aguilar and Lucas Souza [53] present a study based on the relationship between UHI and population size (p) by considering the population density (PD) and the urban form parameters of different neighborhoods in the city of Bogotá, Colombia. They concluded that urban form, expressed by land cover and urban morphology changes caused by population density, has a great effect on temperature differences within a city. Advances in computing technology have fostered the development of new and powerful deep learning techniques that have demonstrated promising results in a wide range of applications. In particular, deep learning methods have been successfully used to classify remotely sensed data collected by Earth observation instruments [54]. Deep learning algorithms, which learn the representative and discriminative features in a hierarchical manner from the data, have recently become a hotspot in the machine-learning area, and have been introduced into the geoscience and remote sensing community for remotely sensed big data analysis [55]. With climate change, the simulation and projection of UHI and its regional impact by using computer technology (deep learning) and remotely sensed data are becoming more important for urban planning and policy makers.
UHI is a result of continued urbanization, urban agglomeration, and associated increases in paved areas and buildings. Mitigation strategies have been developed to increase vegetation and water surface areas within urban areas to reduce the magnitude of the temperature. One measure of UHI’s ecological footprint is estimated by calculating the increase of the cooling demand caused by the heat island over the urban area, and then translating the increased energy use to environmental cost [56][57][58]. Some research shows that the UHI effect has become more prominent in areas of rapid urbanization and in urban agglomerations [59][60]. The spatial distribution of UHI has changed from a mixed pattern, where bare land, semi-bare land, and land under development were warmer than other LULC types, to extensive UHI, as contiguous urbanized blocks grew larger [38][61]. Some analyses showed that the higher temperature in the UHI had a scattered pattern and was related to certain LULC types [62]. In order to analyze the relationship between UHI and LULC changes, some studies attempted to employ a quantitative approach for exploring the relationship between surface temperature and several indices, including the Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), Normalized Difference Bareness Index (NDBaI), and Normalized Difference Build-up Index (NDBI). It was found that correlations between NDVI, NDWI, NDBI, and temperature are negative when NDVI and NDWI are limited in range, but there is a positive correlation between NDBI and temperature [63][64][65][66].
Method | Sensor | Period | Example |
---|---|---|---|
Calculate LST | All thermal bands | 1970s–current | Avdan and Jovanovska [72], and Peng et al. [73] |
Determine the UHIE | Landsat | 2009 | Tang et al. [74] |
Determine the UHII | MODIS | 2001, 2003 | Tran et. al. [75] |
Compare multi-temporal LST images | The normalization of the temperature based on the mean and standard deviation in high and low temperature areas. | Streutker [39] | |
Common normalization of temperture based on min and max LST of the same image in the same way as for NDVI. A normalized ratio scale technique. | Chen et al. [38] | ||
Statistical analyses of UHI | The relationship between LST, NDVI, ground vegetation (GV), and impervious surface area (ISA). Multiple linear regression. Geographically weighted regression. | Weng et al. [76], Tran et al. [75], Schwarz et al. [77], Szymanowski and Kryza [78], and Firozjaei et al. [79] | |
A support vector machine regression (SVR) mode. LST | 2012 (daily) | Lai et al. [80] | |
Data fusion | Landsat, MODIS | 1988–2013, | Shen et al. [81], Wengand Fu [17], and Schmitt and Zhu [69] |
Gap filling | Landsat | 2020 | Yan and Roy [82], Zhou et al. [83], Fu et al. [84], and Zhou et al. [85] |
Time-series analysis | Landsat | 1984–2015 | Huang et al. [86], Peres et al. [87], Fu and Weng [88], and Xian et al. [62] |
Uncertainty and accuracy assessment | MODIS, Landsat | Shen et al. [81], Lee et al. [89], Yuan and Bauer [90], and Chen et al. [91] |