2.7.4. Impact on Biodiversity

Besides UHI, urban development causes wildlife habitat loss and fragmentation, threatens wildlife populations, increases fire risk, and reduces biodiversity [2,148]. These problems are of particular concern in the wildland urban interface (WUI), where homes and associated structures are built among forests, shrublands, or grasslands [1,148,149]. The WUI has received considerable attention because of recent increases in both the number of structures destroyed and the area burned annually by wildland fire. Ceplov ˇ á et al. [150] studied three habitats with different disturbance regimes in 45 central European settlements of three different sizes. Their results highlight the importance of urban size as a factor shaping the biodiversity of native and alien plant communities in individual urban habitats, and the important role of habitat mosaic for maintaining high species richness in city floras. The study of Coluzzi et al. [151] represented a first step to improve the description of relevant processes to protect natural habitats and quality agriculture, therefore combating land degradation and detrimental climate change effects. Kaiser et al. [152] monitored temperature and relative air humidity in wooded sites characterized by different levels of urbanization in the surroundings, and investigated the effect of urbanization at the local and landscape scale on two key traits of biological fitness in two closely related butterfly species that differ in thermal sensitivity.

### **3. Remotely Sensed Thermal Datasets**

Remote sensing derived LST is effective for UHI and UHIRIP studies. Satellites can quickly obtain continuous information over a large geographic area that can be maintained in long-term archives. LST for large geographic areas can be derived from surface radiation of heat measured by satellite sensors. This is particularly attractive when investigating the surface UHI in multiple cities or urban agglomerations at various spatial extents. Along with the extensive spatial coverage, many satellites record multiple wavelengths of electromagnetic energy that can be used to decipher a wealth of information, in addition to thermal information (Figure 2). Consequently, multispectral imagery allows for a comparative analysis between LST and other variables, such as land cover and vegetation indexes [50,153], specifically the interaction between UHI and LULCC [154]. Remote sensing can also be used to track the patterns of change in UHI over time through various time periods from a day, to years, and even a time series of decades [38,155–158]. Because information is desired at a high spatial resolution and dense temporal frequency, data from multiple sensors can make more accurate and reliable quantitative assessments of UHIRIP studies [60]. Table 3 includes a list of the main remotely sensed datasets that have been recently used to derive LST and analyze UHI and UHIRIP.

**Figure 2.** Schematic diagram for using remotely sensed data to evaluate UHI and UHIRIP. Bold outlines indicate high importance. USH—US historical weather data; USC—US climate data; UHI—urban heat island; UHII—UHI intensity; UHIE—UHI effects; IR—infrared band.

The rapid development of remote sensing technology offers more potential for accurate and reliable quantitative assessments of UHI (Table 3 and Figure 3). Many researchers (Table 2) have used remotely sensed LST to assess UHI over various geographic areas. However, for all of these studies, the 1 × 1 km spatial resolution of coarse datasets was found to be suitable only for broad-scale urban temperature mapping (Table 4). The higher resolution of Landsat time series is suitable for UHIRIP at various scales (Table 4).

**Figure 3.** Timeline of satellite data availability. Data availability to 2020 indicates ongoing availability.

**Table 4.** Proportion of reviewed UHI, UHII, and UHIE studies using various remotely sensed data.


Voogt and Oke [6], and others [156,178] pointed out that improved spatial and spectral resolution of sensors and advances in digital image processing techniques increase the usefulness of remote sensing for UHI and UHIRIP studies. Forster [179] also stated that satellite, radar, and airborne sensors can provide spatially continuous information pertaining to numerous variables in urban environments that complement field observations. An increasing number of studies directly relate remotely sensed data to in situ field data [180,181], and applications of remote sensing technology will expand UHI studies to various geographic extents. An exciting recent trend in UHI and UHIRIP research involves

coupling remotely sensed data with ancillary and social economic datasets from multiple sources (Table 2). The typical examples include (1) fractional vegetation cover derived from satellite data to improve model simulations of UHI [182], (2) incorporation of remotely sensed data into a model that partitioned various fluxes in the surface energy balance [183], (3) integrating high-resolution multispectral data with property tax records to investigate the contribution of residential land use to UHI formation [184], (4) studying the potential application of change in urban green space as an indicator of urban environmental quality change [185], (5) using parameters from thermal satellite data and three-dimensional virtual reality models to better understand the factors controlling urban environmental quality (UEQ) [186], (6) further advancing the use of remotely sensed imagery to evaluate UEQ by estimating ground-level particulate matter (PM) concentrations using satellite-based data [187], and (7) estimating the value of U.S. urban tree cover for reducing heat-related health impacts and electricity consumption [188]. In addition, NASA's Ecosystem Spaceborne Thermal Radiometer Experiment on the International Space Station (ECOSTRESS) was launched in June 2018, and is able to image fine-scale temperatures in cities at a 70 × 70 m resolution throughout different times of the day, every 3–5 days on average, over most of the globe [146]. With new algorithm development, ECOSTRESS can accurately monitor UHI trends over time in vulnerable areas such as the urban and non-urban interface. With more available remotely sensed data (Figure 3), innovative studies like these hint at the potential for remote sensing to play an even more prominent role in research of urban climate, urban environment, urban ecological service, and urban planning in the future.

### **4. Algorithms for UHI and UHIRIP in Urban and Non-Urban Interface Studies Based on Remotely Sensed Data**

Generally, the methods for evaluating UHI and UHIRIP can be summarized into four basic types: (1) historical weather station data, (2) field observation, (3) computer simulation, and (4) remote sensing technology. The limitations of the first three methods have been well documented [53,57,105,180]. In this paper, we only focus on the methods that use remote sensing technology. A number of algorithms (or methods) have been developed to estimate UHI and UHIRIP from remotely sensed data (Table 5), including simple empirical approaches to complex methods based on remotely sensed data assimilation using various models. The structure of the UHIRIP pattern centroid in three dimensions indicates the overall variation of the intensity and distribution of the UHI in space and time. The simplified relationship of thermal data and UHI has been applied from a local spatial scale using airborne very high-resolution images to a broad scale with AVHRR, MODIS, ASTER, and Landsat data at regional and continental levels. Assimilation procedures of UHI often require remotely sensed data over different spectral domains to retrieve input parameters that characterize surface properties such as thermal properties, albedo, NDVI, and other indices. A brief review of these approaches is presented in Table 5, with a discussion about the main physical bases and assumptions of various models.

Detailed knowledge of UHI and UHIRIP, especially latent and sensible heat flux components, is important for monitoring the climate change of the land surface. The main methods classically used to measure UHI are appropriate to field observations [24–26,189], but do not allow for an estimation of UHI at large spatial scales. For operational applications to ecological conservation and city planning, managers and engineers need accurate estimates of land surface temperature and UHI at broad spatial scales. New algorithms based on remotely sensed data have been developed to use the imagery of various spatial resolutions and temporal frequency to evaluate UHI [190–192]. It is often difficult to classify these methods because their complexity depends on the balance between the empiricaland physical-based modules used. Nevertheless, we summarize some algorithm (model) categories in the following subsection.


**Table 5.** Methods used to measure UHI and UHIRIP using remotely sensed data.

### *4.1. LST and UHI Intensity Calculation*

LST calculation, including empirical direct methods where remotely sensed data are introduced directly in semi-empirical models to estimate LST, is the simplified relationship between thermal infrared remotely sensed and meteorological data [14]. This method allows for the characterization of UHI intensity both at the local scale, using ground measurements, and over large areas, using satellite data, by calculating a cumulative temperature difference [55,92]. Most current operational models [60] use remote sensing directly to estimate the input parameters and LST.

Seasonal information captures the annual profile of LST and its trend over long time periods, and is essential to the study of UHI [207]. Therefore, remote sensing has been used to accurately monitor and compare the LST difference in the same season in different years and trends over long time periods. In the last 10–15 years, thermal sensor technology has been rapidly developing (Figure 3). Three types of methods have been developed to estimate LST with remotely sensed data: the single infrared channel method; the split window method; and a new day–night MODIS LST method, which is designed to take advantage of the unique capability of the MODIS instrument [55]. Recently, Peng et al. [194] proposed a wavelet coherence approach to exploring spatial heterogeneity and the scale-dependence of the relationship between LST and multiple influencing factors. The advantages, disadvantages, and applicability of these three types of algorithms are summarized in Table 6.


**Table 6.** Advantages, disadvantages, and applicability of commonly used algorithms for calculating LST.

### *4.2. Comparing the Difference between Core Urban and Non-Urban Area*

Many studies have documented the use of LST data to observe meso-scale temperature differences between urban and rural areas in cities worldwide [156,208,209]. The land surface temperature (LST) of core urban areas is generally higher than the surrounding rural areas, and has a strong correlation with land cover [153]. UHII analysis is the most common method to compute the magnitude and extent of UHI by evaluating the LST difference between urban and surrounding non-urban areas [162]. These analyses are often supported with auxiliary land surface information, such as land cover and impervious surface area (ISA). Deterministic models generally are generally based on more complex models that compute the intensity of UHIRIP in space and time. Remotely sensed data are used at different modeling levels, either as the input parameters to characterize the different surface covers, or in assimilation procedures, which aim to retrieve adequate parameters for the LST computation. Some examples of these studies are shown in Table 5. UHI intensity was typically quantified in two steps in these studies [60]. First, urban and non-urban areas were defined and delineated from land cover or ISA maps. Urban areas are usually defined as land with a relatively higher proportion of ISA [38,95], whereas nonurban areas have various definitions in different studies, but generally include non-urban land cover classes. Different sized rural and suburban zones have been used as reference areas. Other land covers, such as water bodies, cropland, forest, and low-intensity ISA, have also been used as references in the studies [101]. Second, the area-weighted mean urban-reference LST differences were calculated to reflect the UHI intensity [69,210] or magnitude. Some studies identified "hotspots" based on positive UHI intensity in certain time periods [211,212]. A positive value of UHI intensity indicated an urban heating effect, while a negative value represented a cooling effect. A few studies also quantified the UHI intensity using small numbers of representative pixels in urban and reference areas instead of the area-weighted mean value for the purpose of surface-air UHI comparison [122–124] or UHI attribution analysis [125,126]. The urban-reference difference method facilitates a comparative analysis of UHIs among cities and urban agglomerations, regions, and across the globe, but the validity of such comparisons can be limited by the large uncertainties associated with urban and reference definitions [68]. Recent research [97] performed a comprehensive and consistent analysis of surface UHI and UHIRIP using Landsat LST

ARD time series and dynamic land cover datasets in the Sioux Falls, SD, area. It shows that the use of time series of LST and land change dynamic data provided a consistent and quantitative analysis for the distribution and change of UHI intensity and UHIRIP (Figure 4). We further discuss limitations in Section 5.

**Figure 4.** A general workflow chart of the use of time series of LST and land change dynamic data that provides a consistent and quantitative analysis for the distribution and change of UHI intensity and UHIRIP in Sioux Falls, SD.

### *4.3. UHI and UHIRIP Analysis by Using Urban Ecological Indices*

Many studies have compared UHIRIP to ecological indices [103,149,213–215], vegetation fraction, and percent ISA, finding strong correlations with mean LST. Landscape metrics indicate that urban landscape configuration also influences the surface UHIRIP [216]. The latest vegetation index methods and inference methods use remote sensing to compute a reduction factor (such as Kc or Priestley Taylor-alpha parameters) for the estimation of the actual UHI [203]. Different papers deal with these approaches in the various journals, and these approaches use land cover [217,218], LST pattern [219,220], and a combination of land cover and LST pattern [221–223] as monitoring indicators of UHI.

Urban ecological status is closely related to the quality of human life and the development of urban economies. A timely and objective understanding of urban ecological status, particularly in urban and non-urban interface areas [224], has become an increasing important. Scientists have been developing a remote sensing-based ecological index for the measure of urban ecological status under UHI [213,225]. This urban ecological status index (UESI) aims to integrate four important ecological indicators that are frequently used in evaluating urban ecology. The four indicators include greenness, wetness, dryness, and heat, and can be represented by four remote sensing indices or components: NDVI, normalized difference built-up and soil index (NDBSI), wetness component of the tasseled cap transformation (Wet), and LST, respectively. Instead of a simple or weighted addition of the four indicators, a principal component analysis (PCA) can be utilized to compress the four indicators into one index in order to assess the overall urban ecological status under UHI. The calculation of the UESI can be fully automated, avoiding the need to assign

threshold values or weights during the computing procedure. Therefore, the UESI can be used to easily and objectively assess urban ecological status. Combined with change detection, UESI can also be used to monitor the change of the ecological status of the core urban and surrounding non-urban areas between different years. In practice, the index was successfully applied in a multitemporal ecological status assessment [34]. Pan [221] used the G index spatial aggregation analysis to calculate the urban heat island ratio index, and the landscape metrics to quantify the changes of the spatial pattern of the UHI from the aspects of quantity, shape, and structure. Pan found that the heat island strength had a negative linear correlation with urban vegetation coverage, and a positive logarithmic correlation with urban impervious surface coverage. Bala et al. [226] developed the Urban Heat Intensity Ratio Index (UHIRI) to quantify urban heat intensity. This work analyzed the variation in LST with land cover changes in Varanasi, India, from 1989 to 2018, using Landsat images, and concluded that the replacement of vegetation with urban land cover has a severe impact on increasing UHI intensity.
