Ecological Evaluation of Urban Heat Island in Chicago City, USA

Tables index

Veiw figure View Table

Weather records corresponding to the same time the LANDSAT image was acquired are needed for the algorithms used to derive the LST. These are relative humidity, dew points, and atmospheric temperatures. The weather data were acquired from The Weather Channel Interactive Inc. website via https://www.wunderground.com/. Additionally, water vapor content in the air was among the required parameters for the LST modeling. The water vapor content was calculated as a function of Relative Humidity (RH) and near surface temperature (T^o) ^[15].

2.3. Methodology

Several algorithms are available for retrieving the LST, a key parameter for the UHI estimation. These are mono-window algorithm ^{[20, 21]}; the split-window algorithm; temperature-emissivity separation algorithm ^[21] and the single-channel method ^[23]. While the split-window and temperature-emissivity separation algorithms are primarily developed for the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) data, both the mono-window and single-channel algorithms are applicable for thermal LANDSAT data, of which the mono-window algorithm is relatively simple and highly effective for retrieving the LST for the analysis of the UHI ^[18]. Three fundamental activities were conducted to quantitatively evaluate the ecological effects of the UHI in Chicago. These were a) deriving a Normalized Difference Vegetation Index (NDVI) and a Normalized Difference Built-up Index (NDBI) images; b) retrieving the LST using LANDSAT TM6 image of 2010 and analyzing the city’s UHI phenomenon; and c) calculating the Urban Thermal Field Variance Index (UTFVI) and interpreting the ecological valuation of the UHI impacts.

2.3.1. The Normalized Difference Vegetation Index (NDVI)

The Normalized Difference Vegetation Index (NDVI) is calculated from the visible red and near infrared bands. The rationale of the index is that healthy vegetation has a high reflectance in the near infrared (NIR) and a low reflectance in the red, thereby enhancing the interpretation of vegetation cover while suppressing subtle noise from other land cover types. The NDVI can be calculated as (1).

(1)

Where:

NDVI = Normalized Vegetation Index;

R = Red band of the LANDSAT Image; and

IR = Infrared band of the LANDSAT Image

In this study, the NDVI is calculated for two fundamental reasons, i.e., understanding the city’s vegetation pattern and extracting emissivity values. First, the NDVI values indicate whether the land cover type is vegetation, which has been shown to have a cooling effect on surfaces temperatures, hence adopted as the UHI mitigation strategies. Secondly, the NDVI was also used to derive emissivity values from the LANDSAT data based on their mutual direct relationship. The emissivity values are critical parameters needed for modeling the LST from the LANDSAT TM6 image.

2.3.2. The Normalized Difference Built-up Index

The Normalized Difference Built-up Index (NDBI) has been an effective technique to map built-up areas with accuracy of 92% ^[15]. The index counts on the reflectivity of urban buildings, which is higher in the fifth band and lower in the fourth band and hence is calculated as:

(2)

Where:

NDBI = Normalized Difference Built-up Index;

MIR = Middle Infrared Band of the LANDSAT Image; and

NIR = Near Infrared of the LANDSAT Image

The NDVI and NDBI were used to establish a correlation matrix with the LST. Based on their theoretical relation, the three variables were used as a second way of empirically validating the values of the LST modeled by the mono-window algorithm ^[19].

2.3.3. Retrieving the Land Surface Temperature (LST)

The LANDSAT TM6 is a thermal-infrared (TIR) band (10.40 to 12.50 μm) that has been commonly used for the LST mapping. For this study, the mono-window algorithm ^[19] was used to retrieve the LST. The mono-window algorithm ^[19] is based on a thermal radiance transfer equation and requires three parameters: emissivity, transmittance and effective mean atmospheric temperature, to retrieve the LST from the LANDSAT TM6. The algorithm can be written as:

(3)

Where:

Ts = is the Land surface temperature (^oK);

a = −67.355351

C = εi × τi ; and D = (1 − τi) [1 + (1 − εi) × τi] such that

εi = is the emissivity,

τi = is the atmospheric transmittance,

b = 0.458606

T_i= is the brightness temperature of TM6 image (K^o).

T_a = represents the effective mean atmospheric temperature (K^o).

Therefore, according to equation (3), the remainder of this section discusses how variables of mono-window algorithm ^[19] were obtained. These are brightness temperature of TM6 image (Ti) in degree Kelvin; atmospheric transmittance (τi); effective mean atmospheric temperature (Ta); and land surface emissivity (εi).

The brightness temperature of TM6 image (Ti), also known as at-sensor (radiance) temperature, is calculated in a two-step process. First, the digital number is converted into spectral radiance. The LANDSAT TM sensors acquire temperature data in TM6, and store the information as digital numbers whose values range between 0 and 255. The numbers are converted into spectral radiances by the following equation:

(4)

Where:

L = Spectral radiance

LMIN = 1.238 (Spectral radiance of DN value 1)

LMAX = 15.600 (Spectral radiance of DN value 255)

DN = Digital Number (image’s brightness value)

Secondly, the spectral radiances are converted into degree Kelvin by the help of Planck’s function, shown as:

(5)

Where:

L = Spectral radiance

K₁= Constant 1 = 60.776 MW*cm^-2*sr^-1 * µm^-1

K₂ = Constant 2= 1260.56⁰K

T_i = is the brightness temperature of TM6 image i.e., at sensor Temperature in ^oK

However, there is a difference between values of at-sensor and ground temperatures, because of the attenuation caused by the atmosphere. Hence, the atmospheric transmittance is calculated to account for this attenuation. The atmospheric transmittance was calculated using a two-step process. First, the atmosphere’s water vapor content was derived from the relative humidity (RH) and the near-surface temperature. The equation for deriving the water vapor content is given by:

(6)

Where:

W_i = is the water vapor content (g/cm2);

T_o = is the near-surface air temperature in ^oK and

RH = represents the relative humidity.

Secondly, once the water vapor content was calculated, the atmospheric transmittance was estimated with the following equations ^{[19, 20]} in Table 2.

(7)

Where:

τ_{i =}is the atmospheric transmittance of LANDSAT TM 6 and

Wi = represents the water vapor content

Table 2. Equations of calculating Atmospheric Transmittance as a function of Water Vapor Content of the Air column (profile) [19,20]

Download as

Tables index

Veiw figure View Table

Finally, the mean atmospheric temperature (Ta) was estimated from near surface air temperature based on regional conversion formula proposed ^{[19, 20]} (See Table 2). The near surface temperature in degree Kelvin is Chicago air temperature recorded at the time of satellite overpass. Accordingly the mean aerial temperature of Chicago was given by:

(8)

Where:

T_a = Effective mean aerial temperature of Chicago (^oK)

T₀ = near surface air temperature (^oK)

The LST is also a factor of the land surface emissivity, which is the efficiency of transmitting thermal energy across the surface into the atmosphere. The land surface emissivity (εi) is estimated by using the NDVI. For the NDVIs between 0.157 and 0.727, the land surface emissivity is calculated with equation (9) ^[23]. Table 4 contains equations for complete land surface emissivity estimation with the NDVI data.

(9)

Where:

εi = Land surface emissivity

NDVI = Normalized Difference Vegetation Index

Table 3. Table containing a formula for emissivity calculation from NDVI [15]

Download as

Tables index

Veiw figure View Table

2.3.4. Model Validation

The NDBI and NDVI were used for the verification of the modeled LST values. Both qualitative and quantitative validation methods are used for verification. Qualitatively, visual interpretations and comparisons of the images of the NDBI, NDVI and LST values were done. A correlation matrix was used for the quantitative evaluation. Theoretically, the LST values must have a positive relationship with the NDBI values and a negative relationship with the NDVI values.

2.3.5. The Urban Thermal Field Variance Index

Several thermal comfort indices are available for evaluating the UHI impacts on the urban quality of life. These are the Temperature Humidity Index (THI) ^[24], the Physiological Equivalent Temperature (PET) ^[25], the Wet-bulb Globe temperature (WBGT) ^[26] and the Urban Thermal Field Variance Index (UTFVI) ^[14]. In this study, the UTFVI is used for the ecological evaluation of Chicago’s urban heat island because of its prior tested application to the LANDSAT data. The UTFVI values were classified into six categories, each having corresponding interpreted ecological valuations and the UHI phenomenon ^[18]. The index, which analyzes the UHI effect on the quality of urban life, is calculated using equation (10) as follows:

(10)

Where:

UTFVI = Urban Thermal Field Variance Index

T_S= LST (^oK)

T_M= Mean LST (^oK).

3. Results and Discussion

3.1. The Accuracy Verification of LST Retrieval

The qualitative visual interpretation of the images of the NDBI, NDVI and LST values verified the theoretical relationships of these variables (Figure 2). Accordingly, areas with the high NDVI values (e.g., parks) have the low NDBI values and the low LST (LST< 32 °C). On the other hand, waterbodies while having the low NDVI (NDVI< 0), also have the low LST (i.e., < 26 °C) and the NDBI. Built-up areas, which had the high NDBI, as expected, also had the low NDVI and the high LST (i.t., > 35 °C).

Download as

Veiw figureFigures index

View next figure

Figure 2. The LST values and corresponding NDVI and NDBI values for the same areas, a) LST values from 21 to 38 °C, B) NDVI values from -1 to 1, C) NDBI values from -1 to 1

Furthermore, the quantitative evaluations of the LST, NDVI and NDBI images, through correlation matrix, also depicted similar outcomes (See Table 4.) According to the matrix, the highest relationship (i.e., r = -0.9) was observed between the NDBI and NDVI images. The high negative correlation between the NDBI and NDVI images is indicative of waning vegetation cover as the land use changes into built-up areas. On the other hand, the relationship between the LST and NDBI images is r = 0.26, while that of the LST and NDVI images is r = -0.34. Although the magnitudes of the relationships were not as strong, the direction is revealing. The positive relationship of the LST and NDBI values reveals the heating impact of built-up and impervious surface areas on surface temperature, while the negative relationship of the LST and NDVI values ascertained the cooling impact of forests, woodlands, parks, and other city green spaces. The direction of the relationship captured by the correlation matrix is significant, as correlation matrix was generated through pixel-based comparison, where confounding subtle noises in the data would obscure chances of obtaining revealing relationships.

Table 4. Correlation matrix of LST, NDVI and NDBI of Chicago city after masking the water bodies

Download as

Tables index

Veiw figure View Table

Several previous studies have shown a comparable relationship between the LST values and these indices ^{[28, 29, 30]}. For instance, a negative relationship (i.e., r = 0.6) is documented between the NDVI and LST, until the NDVI reaches a threshold value of 0.6 ^[29]. It is not quite clear why the cooling effect of the natural vegetation decimates when the percent canopy cover increases beyond this threshold value. And yet the negative correlation between the NDVI and LST indicate the effect of vegetation in reducing the UHI effect, thereby reassuring its utility for UHI mitigation. On the other hand, the positive NDBI and LST relation ^[30] indicated the logical association of the built-up areas and UHI. Both uncrowned settlement pattern, greening urban spaces and increasing the albedo (i.e., reflection coefficient) of the built-up surfaces can be adopted to lessen the impacts of the UHI in Chicago.

3.2. Analysis of Urban Heat Islands (UHI) of Chicago

The spatial distribution of the LST of Chicago is shown in Figure 3. According to the result, the mean LST is 33.6°C, the maximum and minimum LSTs are 16°C and 44°C, respectively. The higher LST dominates northwestern areas of the city, while the eastern and southern areas have the lower LST. The UHI impacts of built-up areas were found to depend on the density of settlement with medium sized building. In Chicago, neighborhoods on the South Side are less crowded and have noticeable green spaces as compared with neighborhoods to the north perhaps hinting as to why the LST in the northern areas were higher vis-à-vis the South Side. The results verify a previous US EPA UHI project ^[17], which reported a consistent high UHI intensity in northwestern Chicagoland as compared to the east and downtown area. The presence of Lake Michigan in the east and southeast are believed to have moderated the local microclimate. Additionally, it could be because of a low Sky View Factor (SVF), i.e., a relatively cooler air temperature from the shedding effect of densely populated skyscrapers around downtown Chicago.

In general, the largest portion of Chicago (i.e., 41.3%) experiences the LSTs around 32 to 35°C. This was found mainly in areas close to Lake Michigan and the wider residential areas. The second largest LST class is 35-38°C, and constitutes about 28.9% of the city, mainly in the north and northwestern areas. The hottest LST (i.e., 38°C and above) is experienced in relatively small pocket areas (i.e., 4.2%) of Chicago, mainly where the large industrial buildings and train stations are located. The high UHI detected in areas of large industrial complexes and train stations, maybe indicating the possible air heating impacts of steams and smokes coming out of the industrial plants and the concentration of steels and concretes from which railroad stations are made. On the other hand, the lowest LST (less than 26°C) recorded in only 2% of the city mainly on the lands covered by water (i.e., lakes and wetlands). Areas having the LSTs between (26 - 29°C) constitute 5.8% of the landscape, and these are lands in close proximity to the Lake Michigan, forest areas and parks. The remaining 17% of Chicago has the LSTs between 29°C and 32°C, mainly on lands that are occupied by forest vegetation, woodlands and green spaces.

Download as

Veiw figureFigures index

View next figure

Figure 3. Land Surface Temperature (LST) distribution in Chicago city on May, 2010 in degree Celsius, derived from LANDSAT images using mono-window algorithm. Divided into seven classes, each class is one standard deviation far from the mean

3.3. The Ecological Evaluation of Chicago Urban Heat Islands

The quantitative ecological evaluation of the UHI effects in Chicago is shown in Figure 4. According to the Urban Thermal Field Variance Index (UTFVI) ^[14], which measures urban ecological quality of life in terms of the degree of thermal comfort in relation to the existence of the UHI phenomenon, varying impacts of the UHI were detected in Chicago. The city has the two extremes: areas of heat stresses (i.e., UTFVI > 0.02) and areas optimal microclimate (i.e., UTFVI <0) (Figure 4 and Table 5). The largest portion of Chicago (i.e., 31%) experiences optimal thermal condition (i.e., UTFVI <0) for living. These are downtown areas and locations in close proximity to the Lake Michigan, wetlands, forest and woodlands, parks and green spaces. On the other hand, the areas that are hit by the worse UHI effects (i.e., UTFVI > 0.01) are relatively small pocket areas (i.e., 9%) of Chicago; while those experiencing bad thermal comfort accounts for 16%. In general, according to the ecological evaluation of the UHI effects, the urban thermal field variance index did not detect thermal discomfort on 75% of the city. It was only in the remaining 25% of the city that varying degrees of thermal discomfort and heat stresses were detected.

Download as

Veiw figureFigures index

Figure 4. Ecological evaluation index of Chicago city in May 2010, according to thermal field variance index (UTFVI)

In general, regional analysis of Chicago’s UHI and UTFVI conforms to studies that analyzed spatiotemporal patterns of the UHI in metropolitan Shanghai ^[31] and Western Territories of Hong Kong ^[27]. According to the study in Shanghai, urban parks were, on average, 4.5°C cooler than the surroundings city areas ^[31]; whereas in the Western Territories of Hong Kong, similar strong association between the LST and LULC types is reported. In the Western Territories of Hong Kong, urban parks and woodlands had similar air temperature as undeveloped rural vegetated areas ^[27]. Moreover, this study corroborates a report of Chicago’s UHI Policy implications, post-1995 heat wave ^[6]. According to the report, even if there is a reduction in the UHI intensity, because of the city’s UHI mitigation strategies, northwestern and pockets of areas in southern and southeastern Chicago are still experiencing the UHI stresses. Therefore, urban planning that pursues spacious settlement pattern with green spaces and parks, and wetlands preservation is important to strengthen Chicago’s UHI mitigation strategies and thereby to maintain urban quality of life.

Table 5. The interpretation of the index quantitative evaluating of the ecological effects of the Chicago city Urban Heat Island (UHI) [18]

Download as

Tables index

Veiw figure View Table