Listen

Evapotranspiration (ET) is an important process of the regional hydrothermal cycle. However, it is unclear how the mitigation of urban ET in urban thermal environments occurs in a spatial context. Landsat 8 satellite images from 2014 to 2018 of Xuzhou and corresponding meteorological observations were selected, and the improved mono-window algorithm (IMW) and urban RS-PM model were applied to invert land surface temperature (LST) and ET, respectively. In addition, spatial analysis methods (a profile analysis, standard deviation ellipse (SDE) analysis, and bivariate Moran's I) were employed to quantify and simulate the spatial characteristics of the ET effect on LST. The results indicated the following: (1) There was a significant linear negative correlation between ET and LST, which confirms that ET has a negative effect on LST; (2) the SDE overlap ratios between patches with higher ET and lower LST imply that higher ET patches have a significant impact on the spatial distribution of LST; and (3) bivariate Moran's I between ET and LST and their linear mixture spectral analysis (LISA) maps reveal a significant negative spatial correlation between ET and LST. In addition, the landscape pattern of higher ET parches is also an important factor affecting the environmental temperature.

Listen

  • A significant negative spatial correlation between ET and LST was found.

  • Areas with various ET intensities have different regulatory effects on LST.

  • Higher ET intensity is not an absolute factor in mitigating the urban thermal environment.

  • Landscape pattern of the patches with higher ET also has a significant impact on LST regulation.

evapotranspiration, geospatial correlation, Landsat 8, land surface temperature, urban
Listen

During the process of urbanization, a large amount of natural ground surface is replaced by impervious surfaces, greatly altering the surface radiation, thermal characteristics, and humidity of urban areas (Battista et al. 2023). Additionally, the low ventilation capacity of urban canyons formed by high-rise buildings and the release of heat from human energy consumption (Zhang et al. 2022) ultimately lead to the formation of the urban heat island (UHI) effect. UHI not only brings a series of urban environmental and human health problems, such as the deterioration of the urban atmospheric environment, regional extreme climates, a dramatic increase in energy consumption, and deterioration of resident health but also is an important factor leading to global warming (Kim & Brown 2021). Therefore, the study of UHI has significant implications for urban planning, public safety, and emergency response.

Previous studies have indicated that vegetation indices, such as the normalized difference vegetation index (NDVI), fractional vegetation coverage (FVC), and leaf area index (LAI), as well as the landscape patterns of urban green spaces, such as their proportion, shape, and aggregation, are strongly linked to land surface temperature (LST) (Zhang et al. 2015; Estoque et al. 2017; Sun et al. 2020), which can lead to vegetation being 5–10 °C cooler than surrounding impervious surfaces (Soydan 2020). The cooling effect of vegetation is mainly achieved through two pathways: evapotranspiration (ET) and shading (Fung & Jim 2019; Wang et al. 2020; Zhang et al. 2021). ET is the process by which water is evaporated from the vegetation and soil into the atmosphere, which involves the phase change from liquid to gas and thus absorbs heat, resulting in a decrease in ambient temperature (Wang et al. 2020). Previous studies have revealed the temporal variations of the cooling effect of urban ET (Chen et al. 2019; Hayat et al. 2020; Qiu et al. 2020; Iqbal et al. 2021), and have also found that the impact of ET on environmental temperature is influenced by the type and morphological characteristics of vegetation (Rahman et al. 2020). A consensus is that vegetation with higher ET rates has a stronger humidifying and cooling effect on the UHI (Jiao et al. 2017). However, these studies mostly rely on single-point measurement methods such as gas analyzers, porometers, lysimeters (Rahman et al. 2017), and liquid flow measurements (Hayat et al. 2022), which provide relatively accurate results but cannot reflect the spatial characteristics of the impact of ET on urban LST.

Remote sensing technology can provide the spatial distribution of ET at the regional scale. However, the commonly used remote sensing models for ET inversion can only be applied to natural or agricultural surfaces with simple land cover types (Hadadi et al. 2022; Yang et al. 2022), and cannot estimate the ET of urban underlying surfaces with high spatial heterogeneity (Wang et al. 2020). This is mainly because the distribution of vegetation and soil in urban areas is relatively scattered and fragmented, making it difficult to extract relatively accurate surface parameters from low and medium resolution satellite images. Therefore, there have been relatively few studies that quantify the spatial impact of urban ET on the thermal environment based on remote sensing technology. Although some remote sensing-based urban ET estimation models have been developed to investigate the relationship between ET and LST, such as the Energy Balance Model for Urban areas (SEBU) (Abunnasr et al. 2022) and the surface energy balance algorithm for land-urban (uSEBAL) (Danda et al. 2023), which have revealed a significant negative correlation between ET and LST, these models do not consider the extraction of sub-pixel evapotranspiration parameters from remote sensing images, which can lead to certain errors in the results. The remote sensing Penman–Monteith (Urban RS-PM) model (Zhang et al. 2018a; Wang et al. 2020) not only solves the problem of extracting surface parameters of vegetation and soil components from mixed pixels in low-resolution remote sensing images in urban areas, but also optimizes the parameterization scheme for ET estimation, providing an effective tool for studying the impact of ET on the thermal environment.

Currently, most studies on the impact of evapotranspiration on the urban thermal environment only investigate the statistical correlation between evapotranspiration and LST (Rocha et al. 2022; Shen et al. 2022), and few studies explore the spatial relationship between ET and LST, which can not only reveal the pattern of UHI mitigation but also be beneficial for urban planning and climate improvement. For example, spatial overlay analysis of areas with strong evapotranspiration and high temperature in cities can reveal the characteristics and contradictions of the spatial development of urban green spaces and urban heat islands. In addition, the spatial correlation distribution map between ET and LST can reveal the differences in ET cooling effects in different regions of the city. All these can provide valuable references for urban environmental regulation. Therefore, the content and innovation points of this study include the following aspects: (1) applying the urban RS-PM model to invert urban ET with higher accuracy, thereby improving the statistical accuracy of the correlation between ET and LST; (2) applying spatial overlay analysis and standard deviation ellipse analysis to reveal the spatial correlation between regions with different ET intensities and areas with different levels of UHI; (3) applying bivariate spatial autocorrelation analysis to reveal the spatial distribution differences in the cooling intensity of different ET values in the study area.

Listen

Study area

Listen
Located southeast of the North China Plain (117°01′–117°25′E, 34°06′–34°22′N), the city of Xuzhou was selected as the study area (Figure 1) and mostly consists of built-up land, with some suburban natural and agricultural surfaces. It has temperate monsoon climate with four distinct seasons but does not experience extreme hot or cold events in summer and winter. The average annual temperature, number of annual sunshine hours, and sunshine ratio were 14 °C; 2,284–2,495 h; and 52–57%, respectively. The vegetation coverage fraction of the entire city and its center area was 30.3 and 43.3%, respectively. The high vegetation coverage of the study area was conducive to studying the relationship between LST and ET.
Figure 1
Geographical map of the study area: (a) location of the study area in China; (b) location of the study area in Xuzhou; and (c) satellite image of the study area.
View largeDownload slide

Geographical map of the study area: (a) location of the study area in China; (b) location of the study area in Xuzhou; and (c) satellite image of the study area.

Figure 1
Geographical map of the study area: (a) location of the study area in China; (b) location of the study area in Xuzhou; and (c) satellite image of the study area.
View largeDownload slide

Geographical map of the study area: (a) location of the study area in China; (b) location of the study area in Xuzhou; and (c) satellite image of the study area.

Close modal

Data source and preprocessing

Listen

Three Landsat 8 OLI (Operational Land Imager) and TIRS (Thermal Infrared Sensor) images from 2014 to 2018 were selected for use. The OLI data were mainly used to extract land cover information and conduct a linear spectral analysis, while the TIRS data were used to invert LST. Meteorological data obtained to match the same acquisition time of the images (air temperature, air relative humidity, atmospheric pressure, and wind speed) were selected to invert ET and LST. In addition, flux data obtained simultaneously from a 30-m-high observation tower equipped with an open-path eddy covariance (EC) system were used to verify the accuracy of the predicted ET values (Table 1). All the meteorological and flux tower data were collected from the collaborative observation site of the China University of Mining and Technology, located in the study area. Meteorological and flux data were recorded every 30 min.

Table 1

Landsat 8, meteorological, and flux datasets

DateLandsat 8 scene IDAir temperature (K)Wind speed (m s–1)Atmospheric pressure (kPa)Relative humidity (%)Latent heat flux (ET) (W m–2)
2014-05-01 LC81210362014121LGN00 297.42 2.66 101.12 55.12 128.25 
2017-05-16 LC81220362017136LGN00 296.33 1.69 101.19 39.76 114.19 
2018-05-03 LC81220362018123LGN00 294.96 4.77 101.69 48.00 178.00 
DateLandsat 8 scene IDAir temperature (K)Wind speed (m s–1)Atmospheric pressure (kPa)Relative humidity (%)Latent heat flux (ET) (W m–2)
2014-05-01 LC81210362014121LGN00 297.42 2.66 101.12 55.12 128.25 
2017-05-16 LC81220362017136LGN00 296.33 1.69 101.19 39.76 114.19 
2018-05-03 LC81220362018123LGN00 294.96 4.77 101.69 48.00 178.00 
View Large

ET inversion by the urban RS-PM model

Listen
The traditional RS-PM model (Cleugh et al. 2007; Mu et al. 2011) is based on a dual-source parallel model (Norman et al. 1995) that assumes that ET of each pixel in the image includes vegetation transpiration and soil moisture evaporation, whose areal sum amounts to 100% in a pixel. Therefore, the proportion of vegetation and soil was calculated as FVC and 1-FVC, respectively. However, not only are there many impervious surfaces in urban areas, but also multiple land cover types are mixed in a single pixel of medium-to-low-resolution images. Zhang et al. (2018a) developed the RS-PM model and further proposed an urban RS-PM model. The developed urban model is capable of extracting the component proportion of each endmember in the mixed pixels based on a linear spectral analysis, thus improving the parameterization scheme of ET inversion according to the surface radiation characteristics of the urban area, as expressed by Equations (1) and (2):
(1)
(2)
where ETv and ETs are the evapotranspiration of vegetation and soil, respectively; fv and fs are the endmember fractions of vegetation and soil in a mixed pixel, respectively; △ = d(es)/d(T) is the curve slope of the saturated vapor pressure (es); R*n,v and R*n,s are the net radiations of pure vegetation and soil pixels, respectively; G*s is the soil heat flux of a pure soil pixel; ρ is the air density; Cp is the air specific heat capacity; ea is the actual vapor pressure; rah,v and rah,s are the aerodynamic resistances of vegetation and soil, respectively; γ is the psychrometer constant; rs,v is the surface resistance of vegetation canopy transpiration; rtot is the sum of aerodynamic resistance and surface resistance for vapor transport; and RH is relative humidity. The parameters of fv, fs, R*n,v, R*n,s, rah,v, rah,s, rs,v, and rtot were derived from the remote sensing data, and the other parameters were obtained from the ground observations.

Fully constrained linear spectral mixture analysis

Listen
Most pixels in urban medium-to-low-resolution images contain a mixture of vegetation, soil, and high-to-low albedo impervious surfaces. Water bodies generally exist independently and were directly masked out in previous studies. Linear mixture spectral analysis (LISA) was used to extract vegetation and soil endmembers in urban mixture pixels so that the ET values of the mixed pixels were calculated. Wu (2004) proposed a fully constrained linear mixture spectral analysis (FCLS) based on LISA, which can limit the sum of the component-specific fractions of all endmembers to 100%, as follows:
(3)
(4)
where fi is the fraction of endmember, i; N = 4 is the number of all endmembers; is the normalized reflectance of endmember i in band b; is the actual standardization reflectance of the mixed pixel; and eb is the error term.

Net radiation calculation

Listen
The traditional calculation methods of net radiation are only applicable for use with natural or agricultural surfaces (Cleugh et al. 2007; Mu et al. 2007). The improved calculation method of net radiation considers the radiation characteristics of urban surfaces as follows:
(5)
(6)
(7)
(8)
where av and as are the albedo values of pure vegetation and soil pixels, respectively, and can be calculated according to Liang et al. (2003); Sd is incoming solar radiation, which can be calculated using the following parameters: the solar constant Isc = 1,367 W m−2, the relative Earth–Sun distance dm, the solar zenith angle θ, the latitude of the study area φ, the solar declination δ, the solar hour angle ω, and the atmospheric broadband transmissivity τb (Spencer 1971; Zhang et al. 2018a); εair is atmosphere effective emissivity (Brutsaert 1975); εv and εs are the typical emissivity of pure vegetation and soil pixels, respectively; Tair is air temperature; and Tv and Ts are the component temperatures of vegetation and soil in the mixed pixels, respectively, which can be calculated according to Mao et al. (2005).

Aerodynamic resistance calculation

Listen
The aerodynamic resistance of vegetation and soil (rah,v and rah,s) in mixed pixels can be calculated as follows (Kustas & Norman 1999):
(9)
(10)
where Z is the observation height from the ground of wind speed and air temperature; k = 0.41 is the von-Karman constant; uz is the observed wind speed at Z height above the ground; do,v, Zom,v, and Zoh,v are the zero-plane displacement, momentum roughness length, and heat roughness length of vegetation, respectively, which can be estimated according to Brutsaert (1982) and Kustas et al. (1989); do,s, Zom,s, and Zoh,s are the zero-plane displacement, momentum roughness length, and heat roughness length of soil, respectively, which can be estimated according to Liu et al. (2007) and Brutsaert (1982); and L is the Monin–Obukhov length.

Surface resistance calculation

Listen
The surface resistance of the vegetation component, rs,v, was calculated using the stomatal conductance of vegetation canopy, as shown in Equations (11)–(13):
(11)
(12)
(13)
where CL = 0.0013 is the empirical coefficient of potential stomatal conductance per unit leaf area index (LAI) (Leuning et al. 2008); and m(Tmin) and m(VPD) are the constraints for minimum air temperature and vapor pressure deficit, respectively, which can be estimated according to Mu et al. (2007).
The surface resistance of the soil component was obtained by correcting for the air temperature (Tair) and pressure (P) for the total surface aerodynamic resistance under standard conditions.
(14)
(15)
where rcorr is the correction term; and rtotc = 107 m·s−1 is a constant based on the observation data.

Soil heat flux calculation

Listen
A relationship between soil heat flux, G*s, net radiation of soil, R*n,s, and solar zenith angle, θ, was derived from site observations, using Equation (16) (Zhang et al. 2018a):
(16)

ET validation by footprint model

Listen
EC flux data represent area-based values instead of point-based ones. The footprint model considers the contribution of the turbulent flux around the EC tower. Thus, the model was used to simulate the contribution of the source area corresponding to EC observations in the ET inversion image (Figure 2). The footprint model assigns a normalized weight value to each pixel in the source area, thus resulting in a weighted average inverse ET of the source area that spatially matched EC observations as follows (Jia et al. 2012):
(17)
(18)
where x is the downwind distance; zm is the EC observation height; Dy(x, y) is the Gaussian crosswind distribution function of the lateral dispersion; fy(x, zm) is the crosswind integrated footprint; ETi is the inverse ET of pixel i in the source area; xi is the contribution weighted value of pixel I; and ETF is the weighted average inverse ET of the source area.
Figure 2
(a) ET inversion map (gray part) overlaid with footprint source area (colored part) of EC observation (using May 1, 2014 data as an example) and (b) photo of flux tower with EC in China University of Mining and Technology.
View largeDownload slide

(a) ET inversion map (gray part) overlaid with footprint source area (colored part) of EC observation (using May 1, 2014 data as an example) and (b) photo of flux tower with EC in China University of Mining and Technology.

Figure 2
(a) ET inversion map (gray part) overlaid with footprint source area (colored part) of EC observation (using May 1, 2014 data as an example) and (b) photo of flux tower with EC in China University of Mining and Technology.
View largeDownload slide

(a) ET inversion map (gray part) overlaid with footprint source area (colored part) of EC observation (using May 1, 2014 data as an example) and (b) photo of flux tower with EC in China University of Mining and Technology.

Close modal

LST inversion by the improved mono-window algorithm

Listen
The two thermal infrared bands (bands 10 and 11) of Landsat 8 can be used to inverse LST. Due to the uncertainty associated with band 11 information, the United States Geological Survey recommends the use of band 10 for single-band calculations but not the use of the split-window algorithm. Based on the mono-window algorithm (Qin et al. 2001), an improved mono-window algorithm (IMW) was proposed to inverse LST, based on band 10 (Wang et al. 2015). The inversion accuracy of IMW was proven to be higher than that of the single-channel algorithm (Wang et al. 2015). In addition to the thermal infrared data, IMW requires only the three parameters of surface emissivity (ε), atmospheric transmittance (τ), and effective average atmospheric temperature (Tair_e) as follows:
(19)
(20)
(21)
(22)
where a = −70.1775 and b = 0.4581 are the linear regression coefficients of Planck's black body radiation function of Band 10 (0–70 °C); and T10 is the brightness temperature based on band 10. T10 and ε can be calculated according to Equations (23) and (24), respectively:
(23)
(24)
where K1 = 774.89 and K2 = 1,321.08 are the preset constants of Landsat 8 band 10, respectively, which can be found in the header file of the image; L10 is the radiance value converted from the initial DN value of band 10; Pv is the FVC index, which can be calculated using normalized difference vegetation index (NDVI); Rv is the vegetation radiation ratio of a mixed pixel; Rx is the radiation ratio of soil (Rs) or impervious surface (Rimp) (Mao et al. 2005); εv = 0.973 is the surface emissivity of a pure vegetation pixel; εx is the surface emissivity of a pure soil (εs = 0.966) or an impervious surface (εimp = 0.962) pixel; and dε is the interaction value of the surface emissivity of each component of the pixel (Qin et al. 2001).

The Moderate Resolution Atmospheric Transmission (MODTRAN 4) program was used to simulate the linear relationship between atmospheric transmittance and atmospheric water content during mid-latitude summers, as shown in Table 2 (Wang et al. 2015). The atmospheric water content (w) was calculated from water vapor pressure data (ea) (Zhang et al. 2017).

Table 2

Linear relationship between atmospheric water content and atmospheric transmittance in mid-latitude summer

w (g·cm−2)Τ
0.2–1.6 0.9184–0.0725 w 
1.6–4.4 1.0163–0.1330 w 
4.4–5.4 0.7029–0.0620 w 
w (g·cm−2)Τ
0.2–1.6 0.9184–0.0725 w 
1.6–4.4 1.0163–0.1330 w 
4.4–5.4 0.7029–0.0620 w 
View Large

Statistical correlation analysis

Listen
As shown in Figure 1(c), four profile lines were established in the study area, and approximately 400 sample points were generated every 240 m on each line. The sample points located on the water have been removed to avoid its interference. To demonstrate the correlation between ET and LST, which were extracted from the sample points, a linear regression and correlation analysis was performed. This analysis aimed to determine the strength and direction of the relationship between the two variables. The equations are as follows:
(25)
(26)
(27)
where a and b are the intercept and slope of the regression equation, respectively; r is the correlation coefficient; n is the number of data samples; xi and yi are the the values of the two variables for the data point i; and and are the mean values of the two variables.

Spatial correlation analysis

Listen

Standard deviational ellipse analysis

Listen
The standard deviational ellipse (SDE) method quantitatively describes the spatial distribution characteristics of geographical elements, using an ellipse created by the basic parameters of central coordinate, major axis, minor axis, and azimuth (Mamuse et al. 2009). The central coordinate represents the relative position of the geographical elements in two-dimensional space. The azimuth reflects the main trend direction of the distribution of the geographical elements (clockwise rotational angle from due north to the major axis of the ellipse), while the major axis represents the dispersion degree of the geographical elements in the main trend direction. The parameters are expressed as follows:
(28)
(29)
(30)
(31)
(32)
where (Xi, Yi) is the spatial location of the geographical elements:; Wi is the weighted value; (⁠⁠, ⁠) is the weighted average center; β is the azimuth of the ellipse; (⁠⁠, ⁠) is the coordinate deviation from the location of each geographical element to the average center; and σX and σY are the standard deviations along the x and y axes, respectively.

Bivariate spatial autocorrelation analysis

Listen
Global and local bivariate Moran's I were employed to reflect the spatial correlation between ET and LST (spatial aggregation or dispersion). The global bivariate Moran's I indicates whether or not there is a spatial correlation between ET and LST as well as the degree of spatial correlation, whereas the local bivariate Moran's I reveals the spatial correlation within different spatial units (Zhang et al. 2018b). The equations used in Moran's I are as follows:
(33)
(34)
where Ieu and Íeu are the global and local bivariate Moran's I, respectively; N is the number of spatial units; Wij is the spatial weight matrix, established by the Queen adjacency mode, and reflects the adjacency relation of spatial units i and j; Zie is the normalized ET value of a spatial unit I; Zju is the normalized LST value of spatial unit j; and Ieu/Íeu is an index range of −1 to 1. A positive value indicated that ET and LST showed a positive spatial correlation; in other words, that spatial patterns with high ET and LST were clustered together. In contrast, a negative value indicated that ET and LST exhibited a negative spatial correlation, where spatial patterns with high ET and low LST were clustered (Zhang et al. 2018b).
Listen

ET and LST inversion

Listen
The remote sensing-based inversion maps of ET and LST are shown in Figure 3, where the spatial distributions of ET and LST differed significantly. The accuracy of the ET inversion results (Figure 4) showed the prediction errors of 9.7, 26.5, and 21.0% between the weighted average ET values of the source area and the EC observations, which are applicable to subsequent studies. Based on the mean standard deviation method, the ET and LST values were normalized and divided into the following five levels: high ET (LST), sub-high ET (LST), medium ET (LST), sub-low ET (LST), and low ET (LST). The area proportions for each level are listed in Table 3. The area proportions of the medium ET and LST patches were largest (both above 32%), and these patches mostly comprised of highly fragmented impervious surfaces, vegetation, and soil patches. The area proportions of the high ET and low LST patches were within 13–19%, and these patches were mostly comprised of large urban parks, urban forests, and suburban farmlands. The proportions of the low ET and high LST areas were within 16–20%, with these patches being mainly covered by impervious surfaces.
Table 3

Proportional areas of ET and LST patches according to their levels

Level2014-05-01
2017-05-16
2018-05-03
ET area (%)LST area (%)ET area (%)LST area (%)ET area (%)LST area (%)
High 14.69 17.32 18.81 18.40 18.38 16.31 
Sub-high 14.14 13.17 14.92 12.60 12.80 12.14 
Medium 38.83 38.05 33.61 38.17 32.67 39.76 
Sub-low 14.32 16.59 12.93 17.38 17.43 17.17 
Low 18.02 14.87 19.73 13.45 18.73 14.62 
Level2014-05-01
2017-05-16
2018-05-03
ET area (%)LST area (%)ET area (%)LST area (%)ET area (%)LST area (%)
High 14.69 17.32 18.81 18.40 18.38 16.31 
Sub-high 14.14 13.17 14.92 12.60 12.80 12.14 
Medium 38.83 38.05 33.61 38.17 32.67 39.76 
Sub-low 14.32 16.59 12.93 17.38 17.43 17.17 
Low 18.02 14.87 19.73 13.45 18.73 14.62 
View Large
Figure 3
Remote sensing inversion results of ET and LST from 2014 to 2018: (a–c) LST maps and (d–f) ET maps.
View largeDownload slide

Remote sensing inversion results of ET and LST from 2014 to 2018: (a–c) LST maps and (d–f) ET maps.

Figure 3
Remote sensing inversion results of ET and LST from 2014 to 2018: (a–c) LST maps and (d–f) ET maps.
View largeDownload slide

Remote sensing inversion results of ET and LST from 2014 to 2018: (a–c) LST maps and (d–f) ET maps.

Close modal
Figure 4
Validation of ET inversion results.
View largeDownload slide

Validation of ET inversion results.

Figure 4
Validation of ET inversion results.
View largeDownload slide

Validation of ET inversion results.

Close modal

Statistical correlation between ET and LST on profile lines

Listen
The normalized ET (ETN) and LST (LSTN) values on four profile lines were extracted from the corresponding sample points, and profile diagrams were constructed to analyze the variation trends and correlations between them, as shown in Figure 5. The ETN and LSTN profiles showed opposite trends in each period. A stronger ET corresponded to a lower LST, indicating that urban ET exerted a negative impact on the thermal environment. The linear relationship between the ETN and LSTN profile data (Figure 6) showed a significant linear negative correlation (p < 0.001), and the correlation coefficient for negative correlation r ranged from −0.61 to −0.69, while the fitting goodness R2 ranged from 0.37 to 0.48, pointing to the significant mitigative effect of ET on LST.
Figure 5
Profile analysis of ETN and LSTN in the study area from 2014 to 2018.
View largeDownload slide

Profile analysis of ETN and LSTN in the study area from 2014 to 2018.

Figure 5
Profile analysis of ETN and LSTN in the study area from 2014 to 2018.
View largeDownload slide

Profile analysis of ETN and LSTN in the study area from 2014 to 2018.

Close modal
Figure 6
Linear relationship between ETN and LSTN profile data.
View largeDownload slide

Linear relationship between ETN and LSTN profile data.

Figure 6
Linear relationship between ETN and LSTN profile data.
View largeDownload slide

Linear relationship between ETN and LSTN profile data.

Close modal

Spatial correlation between ET and LST

Listen

SDE analysis results

Listen
The SDE patterns were generated for each ET and LST level. As shown in Figure 7, the five SDE groups were established as follows: high ET and low LST (H-L), sub-high ET and sub-low LST (SH-SL), medium ET and LST (M-M), sub-low ET and sub-high LST (SL-SH), and low ET and high LST (L-H). The spatial characteristics of H-L, SH-SL, SL-SH, and L-H SDE were highly consistent with respect to their azimuth, shape, and overlapping area; however, the spatial distributions of M-M SDE differed significantly. The statistical results for the overlapping area of each SDE group are shown in Figure 8. The ratio of the overlapping areas of H-L, SH-SL, SL-SH, and L-H to the corresponding SDE area of the ET levels ranged from 91 to 99%. This suggested that the areas with the higher ET levels had the lower LST, whereas the areas with the lower ET levels had the higher LST. In other words, ET exerted a significant impact on the spatial distribution of LST. However, the maximum ratio of the SDE overlapping area of M-M to the corresponding SDE area of the medium ET level only ranged from 72 to 81%, with the medium ET level area less strongly affecting the spatial distribution of LST.
Figure 7
SDE distributions of all ET level and LST level areas (taking data of 2014-05-01 as an example): (a) high ET and low LST areas (H-L); (b) sub-high ET and sub-low LST areas (SH-SL); (c) medium ET and medium LST areas (M-M); (d) sub-low ET and sub-high LST areas (SL-SH); and (e) low ET and high LST areas (L-H).
View largeDownload slide

SDE distributions of all ET level and LST level areas (taking data of 2014-05-01 as an example): (a) high ET and low LST areas (H-L); (b) sub-high ET and sub-low LST areas (SH-SL); (c) medium ET and medium LST areas (M-M); (d) sub-low ET and sub-high LST areas (SL-SH); and (e) low ET and high LST areas (L-H).

Figure 7
SDE distributions of all ET level and LST level areas (taking data of 2014-05-01 as an example): (a) high ET and low LST areas (H-L); (b) sub-high ET and sub-low LST areas (SH-SL); (c) medium ET and medium LST areas (M-M); (d) sub-low ET and sub-high LST areas (SL-SH); and (e) low ET and high LST areas (L-H).
View largeDownload slide

SDE distributions of all ET level and LST level areas (taking data of 2014-05-01 as an example): (a) high ET and low LST areas (H-L); (b) sub-high ET and sub-low LST areas (SH-SL); (c) medium ET and medium LST areas (M-M); (d) sub-low ET and sub-high LST areas (SL-SH); and (e) low ET and high LST areas (L-H).

Close modal
Figure 8
Statistics of overlap area of each SDE group from 2014 to 2018.
View largeDownload slide

Statistics of overlap area of each SDE group from 2014 to 2018.

Figure 8
Statistics of overlap area of each SDE group from 2014 to 2018.
View largeDownload slide

Statistics of overlap area of each SDE group from 2014 to 2018.

Close modal

Bivariate global and local Moran's I between ET and LST

Listen
The results of the bivariate global Moran's I of the three phases were −0.6388, −0.6729, and −0.5782 (p < 0.001) (Table 4), indicating a significant negative spatial correlation between ET and LST. In other words, a significantly low LST aggregation occurred near the high ET area, whereas a pronounced cluster of high LST occurred near the low ET area. This result further supported the spatial mitigation effect of ET on the thermal environment. Local indicators of spatial association (LISA) show the three specific spatial distribution characteristics of the impact of ET on LST (Figure 9). First, the areas with the high ET and low LST concentrations were mainly concentrated in large and continuous patches of urban landscape parks, forests, and unused land, where ET exerted the most significant mitigation effect on the thermal environment. Second, the areas with the low ET and high LST were mainly concentrated in patches of urban residential areas, commercial areas, and industrial areas, mainly covered by impervious surfaces. Overall, these areas had lower proportions of vegetation and soil as well as higher anthropogenic heat efflux, thus resulting in lower ET values that cannot effectively regulate temperature. Finally, the areas with no significant spatial correlation between ET and LST mainly comprised highly fragmented urban green and impervious patches. This indicated that the ET-driven regulation of the thermal environment was not completely controlled by ET intensity, and the spatial characteristics of the ET patches, such as areal and spatial distributions, were also important.
Table 4

Bivariate Moran's I between ET and LST

DateMoran's IZ-value
2014-05-01 −0.6388*** −1,157.92 
2017-05-16 −0.6729*** −1,132.39 
2018-05-03 −0.5782*** −1,032.08 
DateMoran's IZ-value
2014-05-01 −0.6388*** −1,157.92 
2017-05-16 −0.6729*** −1,132.39 
2018-05-03 −0.5782*** −1,032.08 

***Statistically significant p < 0.001.

View Large
Figure 9
LISA cluster maps between ET and LST (H-H: high ET and high LST; L-L: low ET and low LST; L-H: low ET and high LST; and H-L: high ET and low LST. Water body was masked).
View largeDownload slide

LISA cluster maps between ET and LST (H-H: high ET and high LST; L-L: low ET and low LST; L-H: low ET and high LST; and H-L: high ET and low LST. Water body was masked).

Figure 9
LISA cluster maps between ET and LST (H-H: high ET and high LST; L-L: low ET and low LST; L-H: low ET and high LST; and H-L: high ET and low LST. Water body was masked).
View largeDownload slide

LISA cluster maps between ET and LST (H-H: high ET and high LST; L-L: low ET and low LST; L-H: low ET and high LST; and H-L: high ET and low LST. Water body was masked).

Close modal
Listen

Zhang et al. (2018a) reported that the estimation accuracy of the Urban RS-PM model was R2 = 0.8965 with an error rate of 23.7%. The error rate we obtained when applying this model was 19.1%. This accuracy is higher than the ET precision of R2 = 0.7259 obtained by Danda et al. (2023) using the uSEBAL algorithm. Du et al. (2017) also pointed out that the uSEBAL algorithm generally underestimates ET. This is mainly because the Urban RS-PM model considers the ET in the pixels mixed with vegetation, soil, and impervious surface, which is ignored by most models when calculating ET. Therefore, applying the Urban RS-PM model to estimate ET can more accurately reflect its impact on urban thermal environment.

Elliot et al. (2020) and Shukla & Jain (2021) discovered that variations in urban land cover types play a crucial role in the urban heat environment, and Sun et al. (2020) also identified natural surfaces and water bodies as the primary regulators of the urban heat environment. In our study, ET and LST were normalized and divided into five levels across three time periods. The results showed that the mean LST values for low, sub-low, medium, sub-high, and high ET areas were 308.6, 307.6, 306.0, 303.7, and 301.5 K, respectively, indicating a gradual decrease in LST with increasing ET intensity. The relationship between the profile data of ETN and LSTN shows a significant negative correlation. As high ET areas overlap highly with densely vegetated urban areas, this further confirms that ET and shading are the main driving factors for regulating the environmental temperature in urban vegetation, which is consistent with previous studies (Fung & Jim 2019; Qiu et al. 2021; Shukla & Jain 2021; Hayat et al. 2022).

The study by Qiu et al. (2015) showed that the daily average ET of an oasis in a natural environment is about 1.4 mm higher than that of a desert, while the surface temperature of an oasis is 8 K lower than that of a desert. According to the statistical results of Xiong et al. (2016), there is a linear negative correlation between ET and LST (R2 = 0.83). Cui et al. (2019) also demonstrated a linear negative correlation between ET and LST in urban areas based on data from eight cities in Oklahoma, including Tulsa, Norman, Stillwater, Tahlequah, McAlester, Chickasha, and Pauls Valley, with R2 values ranging from 0.29 to 0.44. Our results also demonstrate a significant strong linear negative correlation between ET and LST, with an average correlation coefficient r = −0.66 and 0.37 < R2 < 0.48, which is consistent with these previous research works.

Wang et al. (2020) indicated that land patches with ET intensity ranking in the top 20% of the region significantly increase LST, with a decrease of 0.56 K in LST for every 10 W m−2 increase in ET. Our study also found a similar phenomenon, where 91–98% of the high and sub-high ET level SDE groups (corresponding to the top 29–34% of ET intensity in the study area) were consistent with the low and sub-low LST level SDE groups, indicating that high and sub-high ET levels can effectively mitigate LST. In addition, the three Landsat 8 images selected for our study were all acquired in May, which is in the warm season, and significant strong correlations were observed between ET and LST in each period. This finding also confirms the conclusion of Wang et al. (2020) that the effect of ET on LST is stronger in warm seasons and weaker in cold seasons.

Recently, there have been few studies reporting on the bivariate spatial autocorrelation analysis between ET and LST, aiming to reveal whether the geographical variation of ET leads to changes in the neighboring LST. Through buffer analysis, Wang et al. (2020) have shown that regions with higher ET are surrounded by regions with lower LST. In addition, some studies (Fan & Wang 2020; Kowe et al. 2022) have also revealed a negative spatial correlation between vegetation index and LST, where areas with higher vegetation index tend to aggregate around areas with lower LST. These findings are consistent with our spatial analysis results, which show a negative bivariate global Moran's I between ET and LST (p < 0.001). Our LISA cluster maps also show that areas dominated by impervious surfaces, such as built-up areas, have lower ET and higher LST, while suburban areas show the opposite pattern. Danda et al. (2023) also reported this phenomenon, where the central business district (CBD) area has lower ET, while the green areas have higher ET. This is mainly due to the higher ET rate of water bodies and green areas in suburban regions, which results in lower LST compared to built-up areas.

The bivariate Moran's I analysis also showed areas (referred to as ‘NS areas’ hereafter) where there was no significant spatial correlation between ET and LST (p > 0.05). The proportion of the five ET levels in NS areas is shown in Figure 10. The proportion of moderate ET levels is the highest (40.3–52.7%), reflecting the high uncertainty of the influence of moderate ET intensity on environmental temperature. However, in the NS areas, all five ET levels were observed, and the proportions of high and sub-high ET levels were 5.5–8.5% and 14.4–22.0%, respectively. This indicates that higher ET intensity is not the only factor that effectively mitigates heat in the area.
Figure 10
Area proportion of each ET level in NS area.
View largeDownload slide

Area proportion of each ET level in NS area.

Figure 10
Area proportion of each ET level in NS area.
View largeDownload slide

Area proportion of each ET level in NS area.

Close modal
One possible factor was that LST was also regulated by the landscape pattern of each ET level patch, such as the degrees of fragmentation, aggregation, and connectivity and other landscape indicators have already been proven to regulate environmental temperatures via vegetation (Yao et al. 2020). Because the areas of the high and sub-high ET levels significantly mitigated the thermal environment, the two levels were merged into the higher ET patches in order to explore their landscape patterns. In the LISA maps (Figure 9), the high ET patches were extracted from the significant areas (SA, p > 0.05) and non-significant areas (NSA, p < 0.05). The four landscape metrics of number of patches (NP), mean area of patches (AREA_MN), aggregation index (AI), and patch cohesion index (COHESION) were selected to represent the fragmentation degree, size, aggregation degree, and natural connectivity of the higher ET patches, respectively. The metrics were calculated using Fragstats 4.2, and the results are shown in Figure 11. The NP values of the higher ET patches were all higher in NSA than in SA, while the values of AREA_MN, AI, and COHESION of the higher ET patches were all lower in NSA than in SA. This in turn indicated that the fragmentation degree of the higher ET patches was higher, whereas the size, aggregation degree, and natural connectivity were lower in NSA. These landscape pattern characteristics were the important reasons for the unstable regulatory effect of the high ET patches on LST in NSA, which was consistent with the effect of vegetation pattern characteristics on LST (Yao et al. 2020; Zhou & Cao 2020).
Figure 11
Landscape metrics of higher ET patches in SA and NSA of LISA maps.
View largeDownload slide

Landscape metrics of higher ET patches in SA and NSA of LISA maps.

Figure 11
Landscape metrics of higher ET patches in SA and NSA of LISA maps.
View largeDownload slide

Landscape metrics of higher ET patches in SA and NSA of LISA maps.

Close modal

Previous studies have proved that the density and spatial pattern of green space are effective approaches to mitigate UHI (Estoque et al. 2017; Gage & Cooper 2017; Yao et al. 2020). Our findings further reveal that ET of vegetation and soil in green space are important factors for thermal environment regulation, especially the appropriate spatial distribution of the areas with high ET can enhance the urban cooling effect. Owing to the high–low spatial distribution between ET and LST, our findings also suggest that in urban planning, the proportion of evergreen broad-leaf forest with high ET should be increased in commercial, industrial and residential areas with high LST, which can be an effective way to regulate regional climate and improve the living conditions of urban residents. In addition, increasing the number, aggregation degree, and natural connectivity degree of high ET patches can also strengthen the urban LST mitigation effect.

Listen

This study explored the spatial characteristics of the impact of urban ET on the thermal environment. Based on the remote sensing inversion and spatial analysis of the three periods from 2014 to 2018, the results were three-fold. First, ET and LST showed the opposite variation trends with a significant negative correlation, indicating that ET significantly regulated LST. Second, once the normalized ET and LST values were divided into the five levels from high to low, the SDE groups of the high, sub-high, sub-low, and low ET highly overlapped with those of the low, sub-low, sub-high, and high LST, respectively. This indicated that the higher ET value significantly influenced the spatial distribution characteristics of LST. Finally, given bivariate Moran's I, ET and LST showed a significant negative correlation with the spatial distribution of the high–low and low–high aggregations. In addition, there were areas with non-significant spatial correlation between ET and LST. These results indicated that the regulatory effect of the medium ET intensity on the environmental temperature was uncertain. The landscape pattern of the higher ET parches also played an important role in affecting the environmental temperature, which warrants future studies about the specific spatial effect of the landscape patterns of the higher ET patches on LST. These findings provide important insights into optimizing urban planning, regulating regional climate, and improving the living conditions of urban residents.

Listen

This study was funded by the National Natural Science Foundation of China (Grant No. 42101256), Higher School in Jiangsu Province College Students' Practice Innovation Training Programs (Grant No. 202110320032Z), and Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD) (The Fourth Phase). The comments and suggestions of the editor and the anonymous reviewers are gratefully acknowledged.

Listen

All relevant data are included in the paper or its Supplementary Information.

Listen

The authors declare there is no conflict.

Abunnasr
Y.
,
Mhawej
M.
 &
Chrysoulakis
N.
2022
SEBU: A novel fully automated Google Earth Engine surface energy balance model for urban areas
.
Urban Clim.
44
,
101187
.
https://doi.org/10.1016/j.uclim.2022.101187
.
Google Scholar
Crossref
Search ADS
 
Battista
G.
,
de Lieto Vollaro
E.
,
Ocłoń
P.
 &
de Lieto Vollaro
R.
2023
Effects of urban heat island mitigation strategies in an urban square: a numerical modelling and experimental investigation
.
Energy Build.
282
,
112809
.
https://doi.org/10.1016/j.enbuild.2023.112809
.
Google Scholar
Crossref
Search ADS
 
Brutsaert
W.
1975
On a derivable formula for long-wave radiation from clear skies
.
Water Resour. Res.
11
,
742
–
744
.
https://doi.org/10.1029/WR011i005p00742
.
Google Scholar
Crossref
Search ADS
 
Brutsaert
W.
1982
Evaporation into the Atmosphere: Theory, History and Applications
.
D. Riedel Publishing Company, Springer Netherlands
,
Berlin
,
Netherlands
.
https://doi.org/10.1007/978-94-017-1497-6
.
Google Scholar
Crossref
Search ADS
 
Chen
X.
,
Zhao
P.
,
Hu
Y.
,
Ouyang
L.
,
Zhu
L.
 &
Ni
G.
2019
Canopy transpiration and its cooling effect of three urban tree species in a subtropical city - Guangzhou, China
.
Urban For. Urban Greening
43
,
126368
.
https://doi.org/10.1016/j.ufug.2019.126368
.
Google Scholar
Crossref
Search ADS
 
Cleugh
H. A.
,
Leuning
R.
,
Mu
Q.
 &
Running
S. W.
2007
Regional evaporation estimates from flux tower and MODIS satellite data
.
Remote Sens. Environ.
106
,
285
–
304
.
https://doi.org/10.1016/j.rse.2006.07.007
.
Google Scholar
Crossref
Search ADS
 
Cui
Y.
,
Xiao
X.
,
Doughty
R. B.
,
Qin
Y.
,
Liu
S.
,
Li
N.
,
Zhao
G.
 &
Dong
J.
2019
The relationships between urban-rural temperature difference and vegetation in eight cities of the Great Plains
.
Front. Earth Sci.
13
,
290
–
302
.
https://doi.org/10.1007/s11707-018-0729-5
.
Google Scholar
Crossref
Search ADS
 
Danda
T. J.
,
Kusangaya
S.
,
Mushore
T. D.
 &
Mujere
N.
2023
Assessing the role of evapotranspiration in reducing surface temperatures in Harare using the SEBAL algorithm
.
Urban Clim.
49
,
101442
.
https://doi.org/10.1016/j.uclim.2023.101442
.
Google Scholar
Crossref
Search ADS
 
Du
H.
,
Cai
W.
,
Xu
Y.
,
Wang
Z.
,
Wang
Y.
 &
Cai
Y.
2017
Quantifying the cool island effects of urban green spaces using remote sensing data
.
Urban For. Urban Greening
27
,
24
–
31
.
https://doi.org/10.1016/j.ufug.2017.06.008
.
Google Scholar
Crossref
Search ADS
 
Elliot
T.
,
Babí Almenar
J.
 &
Rugani
B.
2020
Modelling the relationships between urban land cover change and local climate regulation to estimate urban heat island effect
.
Urban For. Urban Greening
50
,
126650
.
https://doi.org/10.1016/j.ufug.2020.126650
.
Google Scholar
Crossref
Search ADS
 
Estoque
R. C.
,
Murayama
Y.
 &
Myint
S. W.
2017
Effects of landscape composition and pattern on land surface temperature: an urban heat island study in the megacities of Southeast Asia
.
Sci. Total Environ.
577
,
349
–
359
.
https://doi.org/10.1016/j.scitotenv.2016.10.195
.
Google Scholar
Crossref
Search ADS
PubMed
 
Fung
C. K. W.
 &
Jim
C. Y.
2019
Microclimatic resilience of subtropical woodlands and urban-forest benefits
.
Urban For. Urban Greening
42
,
100
–
112
.
https://doi.org/10.1016/j.ufug.2019.05.014
.
Google Scholar
Crossref
Search ADS
 
Gage
E. A.
 &
Cooper
D. J.
2017
Relationships between landscape pattern metrics, vertical structure and surface urban heat island formation in a Colorado suburb
.
Urban Ecosyst.
20
,
1229
–
1238
.
https://doi.org/10.1007/s11252-017-0675-0
.
Google Scholar
Crossref
Search ADS
 
Hadadi
F.
,
Moazenzadeh
R.
 &
Mohammadi
B.
2022
Estimation of actual evapotranspiration: a novel hybrid method based on remote sensing and artificial intelligence
.
J. Hydrol.
609
,
127774
.
https://doi.org/10.1016/j.jhydrol.2022.127774
.
Google Scholar
Crossref
Search ADS
 
Hayat
M.
,
Zha
T.
,
Jia
X.
,
Iqbal
S.
,
Qian
D.
,
Bourque
C. P.-A.
,
Khan
A.
,
Tian
Y.
,
Bai
Y.
,
Liu
P.
 &
Yang
R.
2020
A multiple-temporal scale analysis of biophysical control of sap flow in Salix psammophila growing in a semiarid shrubland ecosystem of northwest China
.
Agric. For. Meteorol.
288–289
,
107985
.
https://doi.org/10.1016/j.agrformet.2020.107985
.
Google Scholar
 
Hayat
M.
,
Xiang
J.
,
Yan
C.
,
Xiong
B.
,
Wang
B.
,
Qin
L.
,
Saeed
S.
,
Hussain
A.
,
Zou
Z.
 &
Qiu
G. Y.
2022
Environmental control on transpiration and its cooling effect of Ficus concinna in a subtropical city Shenzhen, southern China
.
Agric. For. Meteorol.
312
,
108715
.
https://doi.org/10.1016/j.agrformet.2021.108715
.
Google Scholar
Crossref
Search ADS
 
Iqbal
S.
,
Zha
T.
,
Jia
X.
,
Hayat
M.
,
Qian
D.
,
Bourque
C. P.-A.
,
Tian
Y.
,
Bai
Y.
,
Liu
P.
,
Yang
R.
 &
Khan
A.
2021
Interannual variation in sap flow response in three xeric shrub species to periodic drought
.
Agric. For. Meteorol.
297
,
108276
.
https://doi.org/10.1016/j.agrformet.2020.108276
.
Google Scholar
Crossref
Search ADS
 
Jia
Z.
,
Liu
S.
,
Xu
Z.
,
Chen
Y.
 &
Zhu
M.
2012
Validation of remotely sensed evapotranspiration over the Hai River Basin, China
.
J. Geophys. Res. Atmos.
117
,
D13113
.
https://doi.org/10.1029/2011JD017037
.
Google Scholar
Crossref
Search ADS
 
Jiao
M.
,
Zhou
W.
,
Zheng
Z.
,
Wang
J.
 &
Qian
Y.
2017
Patch size of trees affects its cooling effectiveness: a perspective from shading and transpiration processes
.
Agric. For. Meteorol.
247
,
293
–
299
.
https://doi.org/10.1016/j.agrformet.2017.08.013
.
Google Scholar
Crossref
Search ADS
 
Kim
S. W.
 &
Brown
R. D.
2021
Urban heat island (UHI) intensity and magnitude estimations: a systematic literature review
.
Sci. Total Environ.
779
,
146389
.
https://doi.org/10.1016/j.scitotenv.2021.146389
.
Google Scholar
Crossref
Search ADS
PubMed
 
Kowe
P.
,
Dube
T.
,
Mushore
T. D.
,
Ncube
A.
,
Nyenda
T.
,
Mutowo
G.
,
Chinembiri
T. S.
,
Traore
M.
 &
Kizilirmak
G.
2022
Impacts of the spatial configuration of built-up areas and urban vegetation on land surface temperature using spectral and local spatial autocorrelation indices
.
Remote Sens. Lett.
13
,
1222
–
1235
.
https://doi.org/10.1080/2150704X.2022.2142073
.
Google Scholar
Crossref
Search ADS
 
Kustas
W. P.
 &
Norman
J. M.
1999
Evaluation of soil and vegetation heat flux predictions using a simple two-source model with radiometric temperatures for partial canopy cover
.
Agric. For. Meteorol.
94
,
13
–
29
.
https://doi.org/10.1016/S0168-1923(99)00005-2
.
Google Scholar
Crossref
Search ADS
 
Kustas
W. P.
,
Choudhury
B. J.
,
Moran
M. S.
,
Reginato
R. J.
,
Jackson
R. D.
,
Gay
L. W.
 &
Weaver
H. L.
1989
Determination of sensible heat flux over sparse canopy using thermal infrared data
.
Agric. For. Meteorol.
44
,
197
–
216
.
https://doi.org/10.1016/0168-1923(89)90017-8
.
Google Scholar
Crossref
Search ADS
 
Leuning
R.
,
Zhang
Y. Q.
,
Rajaud
A.
,
Cleugh
H.
 &
Tu
K.
2008
A simple surface conductance model to estimate regional evaporation using MODIS leaf area index and the Penman-Monteith equation
.
Water Resour. Res.
44
,
W10419
.
https://doi.org/10.1029/2007WR006562
.
Google Scholar
Crossref
Search ADS
 
Liang
S.
,
Shuey
C. J.
,
Russ
A. L.
,
Fang
H.
,
Chen
M.
,
Walthall
C. L.
,
Daughtry
C. S. T.
 &
Hunt
R.
2003
Narrowband to broadband conversions of land surface albedo: II. Validation
.
Remote Sens. Environ.
84
,
25
–
41
.
https://doi.org/10.1016/S0034-4257(02)00068-8
.
Google Scholar
Crossref
Search ADS
 
Liu
S.
,
Lu
L.
,
Mao
D.
 &
Jia
L.
2007
Evaluating parameterizations of aerodynamic resistance to heat transfer using field measurements
.
Hydrol. Earth Syst. Sci.
11
,
769
–
783
.
https://doi.org/10.5194/hess-11-769-2007
.
Google Scholar
Crossref
Search ADS
 
Mamuse
A.
,
Porwal
A.
,
Kreuzer
O.
 &
Beresford
S.
2009
A new method for spatial centrographic analysis of mineral deposit clusters
.
Ore Geol. Rev.
36
,
293
–
305
.
https://doi.org/10.1016/j.oregeorev.2009.06.001
.
Google Scholar
Crossref
Search ADS
 
Mao
K.
,
Qin
Z.
,
Shi
J.
 &
Gong
P.
2005
A practical split-window algorithm for retrieving land-surface temperature from MODIS data
.
Int. J. Remote Sens.
26
,
3181
–
3204
.
https://doi.org/10.1080/01431160500044713
.
Google Scholar
Crossref
Search ADS
 
Mu
Q.
,
Heinsch
F. A.
,
Zhao
M.
 &
Running
S. W.
2007
Development of a global evapotranspiration algorithm based on MODIS and global meteorology data
.
Remote Sens. Environ.
111
,
519
–
536
.
https://doi.org/10.1016/j.rse.2007.04.015
.
Google Scholar
Crossref
Search ADS
 
Mu
Q.
,
Zhao
M.
 &
Running
S. W.
2011
Improvements to a MODIS global terrestrial evapotranspiration algorithm
.
Remote Sens. Environ.
115
,
1781
–
1800
.
https://doi.org/10.1016/j.rse.2011.02.019
.
Google Scholar
Crossref
Search ADS
 
Norman
J. M. M.
,
Kustas
W. P. P.
 &
Humes
K. S. S.
1995
Source approach for estimating soil and vegetation energy fluxes in observations of directional radiometric surface temperature
.
Agric. For. Meteorol.
77
,
263
–
293
.
https://doi.org/10.1016/0168-1923(95)02265-Y
.
Google Scholar
Crossref
Search ADS
 
Qin
Z.
,
Karnieli
A.
 &
Berliner
P.
2001
A mono-window algorithm for retrieving land surface temperature from Landsat TM data and its application to the Israel-Egypt border region
.
Int. J. Remote Sens.
22
,
3719
–
3746
.
https://doi.org/10.1080/01431160010006971
.
Google Scholar
Crossref
Search ADS
 
Qiu
G. Y.
,
Li
C.
 &
Yan
C.
2015
Characteristics of soil evaporation, plant transpiration and water budget of Nitraria dune in the arid Northwest China
.
Agric. For. Meteorol.
203
,
107
–
117
.
https://doi.org/10.1016/j.agrformet.2015.01.006
.
Google Scholar
Crossref
Search ADS
 
Qiu
G. Y.
,
Wang
B.
,
Li
T.
,
Zhang
X.
,
Zou
Z.
 &
Yan
C.
2021
Estimation of the transpiration of urban shrubs using the modified three-dimensional three-temperature model and infrared remote sensing
.
J. Hydrol.
594
,
125940
.
https://doi.org/10.1016/j.jhydrol.2020.125940
.
Google Scholar
Crossref
Search ADS
 
Rahman
M. A.
,
Moser
A.
,
Rötzer
T.
 &
Pauleit
S.
2017
Microclimatic differences and their influence on transpirational cooling of Tilia cordata in two contrasting street canyons in Munich, Germany
.
Agric. For. Meteorol.
232
,
443
–
456
.
https://doi.org/10.1016/j.agrformet.2016.10.006
.
Google Scholar
Crossref
Search ADS
 
Rahman
M. A.
,
Stratopoulos
L. M. F.
,
Moser-Reischl
A.
,
Zölch
T.
,
Häberle
K.-H.
,
Rötzer
T.
,
Pretzsch
H.
 &
Pauleit
S.
2020
Traits of trees for cooling urban heat islands: a meta-analysis
.
Build. Environ.
170
,
106606
.
https://doi.org/10.1016/j.buildenv.2019.106606
.
Google Scholar
Crossref
Search ADS
 
Rocha
A. D.
,
Vulova
S.
,
Meier
F.
,
Förster
M.
 &
Kleinschmit
B.
2022
Mapping evapotranspirative and radiative cooling services in an urban environment
.
Sustain. Cities Soc.
85
,
104051
.
https://doi.org/10.1016/j.scs.2022.104051
.
Google Scholar
Crossref
Search ADS
 
Shen
X.
,
Liu
Y.
,
Wu
L.
,
Ma
R.
,
Wang
Y.
,
Zhang
J.
,
Wang
L.
,
Liu
B.
,
Lu
X.
 &
Jiang
M.
2022
Grassland greening impacts on global land surface temperature
.
Sci. Total Environ.
838
,
155851
.
https://doi.org/10.1016/j.scitotenv.2022.155851
.
Google Scholar
Crossref
Search ADS
PubMed
 
Shukla
A.
 &
Jain
K.
2021
Analyzing the impact of changing landscape pattern and dynamics on land surface temperature in Lucknow city, India
.
Urban For. Urban Greening
58
,
126877
.
https://doi.org/10.1016/j.ufug.2020.126877
.
Google Scholar
Crossref
Search ADS
 
Spencer
J. W.
1971
Fourier series representation of the position of the sun
.
Search
2
,
172
.
Google Scholar
 
Sun
X.
,
Tan
X.
,
Chen
K.
,
Song
S.
,
Zhu
X.
 &
Hou
D.
2020
Quantifying landscape-metrics impacts on urban green-spaces and water-bodies cooling effect: the study of Nanjing, China
.
Urban For. Urban Greening
55
,
126838
.
https://doi.org/10.1016/j.ufug.2020.126838
.
Google Scholar
Crossref
Search ADS
 
Wang
F.
,
Qin
Z.
,
Song
C.
,
Tu
L.
,
Karnieli
A.
 &
Zhao
S.
2015
An improved mono-window algorithm for land surface temperature retrieval from Landsat 8 thermal infrared sensor data
.
Remote Sens.
7
,
4268
–
4289
.
https://doi.org/10.3390/rs70404268
.
Google Scholar
Crossref
Search ADS
 
Wu
C.
2004
Normalized spectral mixture analysis for monitoring urban composition using ETM+ imagery
.
Remote Sens. Environ.
93
,
480
–
492
.
https://doi.org/10.1016/j.rse.2004.08.003
.
Google Scholar
Crossref
Search ADS
 
Xiong
Y.
,
Zhao
S.
,
Yin
J.
,
Li
C.
 &
Qiu
G.
2016
Effects of evapotranspiration on regional land surface temperature in an arid oasis based on thermal remote sensing
.
IEEE Geosci. Remote Sens. Lett.
13
,
1885
–
1889
.
https://doi.org/10.1109/LGRS.2016.2616409
.
Google Scholar
Crossref
Search ADS
 
Yang
Y.
,
Anderson
M.
,
Gao
F.
,
Xue
J.
,
Knipper
K.
 &
Hain
C.
2022
Improved daily evapotranspiration estimation using remotely sensed data in a data fusion system
.
Remote Sens.
14
,
1772
.
https://doi.org/10.3390/rs14081772
.
Google Scholar
Crossref
Search ADS
 
Yao
L.
,
Li
T.
,
Xu
M.
 &
Xu
Y.
2020
How the landscape features of urban green space impact seasonal land surface temperatures at a city-block-scale: an urban heat island study in Beijing, China
.
Urban For. Urban Greening
52
,
126704
.
https://doi.org/https://doi.org/10.1016/j.ufug.2020.126704
.
Google Scholar
Crossref
Search ADS
 
Zhang
Y.
,
Chen
L. L.
,
Wang
Y.
,
Chen
L. L.
,
Yao
F.
,
Wu
P.
,
Wang
B.
,
Li
Y.
,
Zhou
T.
 &
Zhang
T.
2015
Research on the contribution of urban land surface moisture to the alleviation effect of urban land surface heat based on Landsat 8 data
.
Remote Sens.
7
,
10737
–
10762
.
https://doi.org/10.3390/rs70810737
.
Google Scholar
Crossref
Search ADS
 
Zhang
Y.
,
Li
L.
,
Chen
L.
,
Liao
Z.
,
Wang
Y.
,
Wang
B.
 &
Yang
X.
2017
A modified multi-source parallel model for estimating urban surface evapotranspiration based on ASTER thermal infrared data
.
Remote Sens.
9
,
1029
.
https://doi.org/10.3390/rs9101029
.
Google Scholar
Crossref
Search ADS
 
Zhang
Y.
,
Li
L.
,
Qin
K.
,
Wang
Y.
,
Chen
L.
 &
Yang
X.
2018a
Remote sensing estimation of urban surface evapotranspiration based on a modified Penman–Monteith model
.
J. Appl. Remote Sens.
12
,
1
.
https://doi.org/10.1117/1.JRS.12.046006
.
Google Scholar
 
Zhang
Y.
,
Liu
Y.
,
Zhang
Y.
,
Liu
Y.
,
Zhang
G.
 &
Chen
Y.
2018b
On the spatial relationship between ecosystem services and urbanization: a case study in Wuhan, China
.
Sci. Total Environ.
637–638
,
780
–
790
.
https://doi.org/10.1016/j.scitotenv.2018.04.396
.
Google Scholar
PubMed
 
Zhang
J.
,
Shen
X.
,
Wang
Y.
,
Jiang
M.
 &
Lu
X.
2021
Effects of forest changes on summer surface temperature in Changbai Mountain, China
.
Forests
12
,
1551
.
https://doi.org/10.3390/f12111551
.
Google Scholar
Crossref
Search ADS
 
Zhang
Y.
,
Wang
Y.
,
Ding
N.
 &
Yang
X.
2022
Spatial pattern impact of impervious surface density on urban heat island effect: a case study in Xuzhou, China
.
Land
11
,
2135
.
https://doi.org/10.3390/land11122135
.
Google Scholar
Crossref
Search ADS
 
Zhou
W.
 &
Cao
F.
2020
Effects of changing spatial extent on the relationship between urban forest patterns and land surface temperature
.
Ecol. Indic.
109
,
105778
.
https://doi.org/10.1016/j.ecolind.2019.105778
.
Google Scholar
Crossref
Search ADS
 
© 2023 The Authors
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/).