The application of a high-density street-level air temperature observation network (HiSAN): The relationship between air temperature, urban development, and geographic features

Science of the Total Environment 685 (2019) 710–722

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

The application of a high-density street-level air temperature observation network (HiSAN): The relationship between air temperature, urban development, and geographic features Yu-Cheng Chen a, Yu-Jie Liao a, Chun-Kuei Yao a, Tsuyoshi Honjo b, Chi-Kuei Wang c, Tzu-Ping Lin a,⁎ a b c

Department of Architecture, National Cheng Kung University, 1 University Rd., East Dist., Tainan 701, Taiwan Department of Landscape Architecture and Environmental Science, Chiba University, 648, Matsudo, Matsudo, Chiba 271-8510, Japan Department of Geomatics, National Cheng Kung University, 1 University Rd., East Dist., Tainan 701, Taiwan

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

• The HiSAN with buffer analysis clearly indicate the characteristics of UHI in Tainan. • Urban features have different effects on the thermal condition at different times. • Water ratio in daytime and vegetation in nighttime have positive cooling effect. • The built-up area is favored to located near the coastal for better thermal comfort. • The Ta prediction model can be built by the urban development and geographic factors.

a r t i c l e

i n f o

Article history: Received 21 February 2019 Received in revised form 4 June 2019 Accepted 4 June 2019 Available online 06 June 2019 Edited: SCOTT SHERIDAN Keywords: Urban Heat Island Observations network Thermal condition Tainan Urban development factor

⁎ Corresponding author. E-mail address: [email protected] (T.-P. Lin).

https://doi.org/10.1016/j.scitotenv.2019.06.066 0048-9697/© 2019 Elsevier B.V. All rights reserved.

a b s t r a c t The urban heat island effect in cities has become an important problem in relation to not only urban climate but also public health and urban planning. Tainan, which located in Southern Taiwan, is a compact city with intense development. Therefore, this study investigated the urban thermal condition by employing a high-density streetlevel air temperature observation network (HiSAN). A total of 100 measurement points were set in various urban development areas. The geographic factors in Tainan can be used for indicating the relationship between thermal conditions and urban built environments to comprehensively compare the approaches, such as conducting traverse measurement and utilizing only a single datum or a small amount of weather station data. Buffer zone analysis was used in this study for zones of different sizes, and it was determined that a 300-m scale is optimal to illustrate the effects of land features on microclimate. The results revealed that the thermal condition in Tainan is influenced by urban development factors, such as the floor area and land cover area, and by geographic factors, such as the distance to the sea. A better cooling effect can be obtained from a vegetation area during the night time and from a water body during the daytime. Moreover, different cooling effects are observed based on the distance to the sea. Through these results, a model for predicting the thermal condition for different periods

Y.-C. Chen et al. / Science of the Total Environment 685 (2019) 710–722

711

can be established using a multiple regression model. Urban planners and architects can proffer design and planning suggestions for different areas based on the findings of this study to reduce thermal stress in urban areas. © 2019 Elsevier B.V. All rights reserved.

1. Introduction Due to the heat storage of impermeable materials and the anthropogenic heat caused by growth in urbanization (Allen et al., 2011; Feng et al., 2012; Lin et al., 2017a, 2017b), the thermal load in urban areas has become higher than that in suburban and rural areas. This effect is known as the urban heat island (UHI) (Oke, 1973; Kotharkar and Surawar, 2015; Chen et al., 2016a, 2016b; Lin et al., 2017a, 2017b; Chen et al., 2019). In previous studies, the UHI has been observed using various approaches, such as mobile measurement, remote sensing data, and weather station data (Saaroni et al., 2000; Heusinkveld et al., 2010; Nakayoshi et al., 2015; Kaloustian and Bechtel, 2016; Lin et al., 2017a, 2017b; Liu et al., 2017; Chen et al., 2016a, 2016b; Son et al., 2017; Mikami et al., 2011; Chen et al., 2018; Pigliautile and Pisello, 2018). In terms of mobile measurement, the meteorological measurement instruments and a track recorder equipped with a global positioning system can be mounted on vehicles or other platform including wearable monitoring system to measure the micro climate variation in the land typology of the city. The advantages of mobile measurement are as follows: high flexibility in receiving a wide range of data, high flexibility in modifying the observation period, and ability of directly measuring the meteorological information at the pedestrian level. However, the accuracy is affected by the speed of travel and the heat emission rate of vehicles. Therefore, for data correction, it is necessary to use data from a weather station that is located at a fixed point (Saaroni et al., 2000; Heusinkveld et al., 2010; Lin et al., 2017a, 2017b; Chen et al., 2019). In the case of this issue, the wearable monitoring system that produces low heat emission, can maintain a stable speed and be able to detect very site-specific environmental conditions can be promising to investigate UHI (Nakayoshi et al., 2015; Pigliautile and Pisello, 2018). For remote sensing, satellite and airborne images are widely used for determining the intensity of UHI because of the possibility of obtaining a wide range of thermal data at certain time without conducting a measurement survey. However, the observation period is limited, and the satellite poses some limitations in terms of the image resolution (Lo et al., 1997; Li et al., 2011; Liang et al., 2012; Chen et al., 2016a, 2016b; Kaplan et al., 2018; Naserikia et al., 2019). As only land surface temperature can be estimated using this method, it is more difficult to use this method for explaining the influence of different built environment factors on urban climate. For determining the UHI intensity by using the traditional weather station that is fixed at a point, the method can be used to precisely present the climate condition of a small area. However, for analyzing the UHI thermal conditions of a wide area and of built environments in complex urban areas, a single or few weather stations in the same land area are insufficient for representing the meteorological changes or for reflecting the effect of development on the microclimate in different regions (Chang et al., 2010; Merbitz et al., 2012; Honjo et al., 2015; Chen et al., 2018). A detailed information on microclimate conditions in each district must be obtained appropriately. Therefore, a high-density street-level air temperature observation network (HiSAN), consisting of 100 measurement points, was established to investigate the urban microclimate conditions, such as air temperature and humidity, of land used for various purposes and with different land covers in the entire study area. The network was used to comprehend the climate problems, including high UHI intensity, and the relationship between a built environment and climate conditions (Chen et al., 2018; Lin et al., 2019).

Due to rapid urbanization, a large amount of the population continuously moves to urban areas, which increases the building density and development intensity (Koellner and Scholz, 2007; Chen et al., 2016a, 2016b; Lin et al., 2017a, 2017b). The construction of impermeable pavements and similar structures modify the original land pattern by altering components such as greenery, soil, and water. Therefore, the resulting microclimate can barely be adjusted by the natural environment (Baker et al., 2002; Chen et al., 2018). Various researchers indicated that the different built environment parameters, such as the building area, total floor area, impermeable pavement ratio, and green area ratio, can influence microclimate (Xiao et al., 2007; Alcoforado et al., 2009; Dawod et al., 2011; Weng et al., 2011; Lu et al., 2012). However, few studies have proved the influence of area size on thermal conditions by using various urban development parameters, although this factor is very important for quantifying the value of each parameter in a certain area (Eliasson and Svensson, 2003; Dimitriou and Zacharias, 2010; Acero and Herranz-Pascual, 2015). Therefore, this study used buffer zone analysis with scales of several sizes to determine the optimum buffer zone size to determine the relationship between built environment and the thermal condition more precisely. In a previous study on urban climate, the thermal condition was mainly evaluated based on the land use and type of land cover (Jittawikul et al., 2004; Matzarakis et al., 2008; Smith et al., 2009; Ren et al., 2010; Chun and Guhathakurta, 2017). However, the study area of this study is a city located beside the coastal area, and the climate of this area is significantly influenced by the sea (Chen et al., 2016a, 2016b; Chen et al., 2018). Previous studies proposed that the climate characteristics of coastal cities were significantly influenced by the sea (Roth et al., 1989; Saaroni et al., 2000; Honjo, 2012; Hjort et al., 2016). In Tel-Aviv, Israel, the UHI intensity diminishes and enhances based on the different speeds of the sea breeze and is strongly affected by changes in the duration of sea and land breeze. This phenomenon indicates the particular climate conditions of coastal areas have effects on urban thermal conditions (Saaroni et al., 2000). In Tokyo, Japan, minimal wind from the sea may be a characteristic of urban waterfront areas. In these areas, temperature increase persists throughout the day and even during the night because of the heat accumulated on the ground surface and the artificial exhaust heat. The differences in sea and land characteristic cause dynamic movement of the UHI centric points from the east to the west during the night time and from the west to the east during the daytime (Honjo, 2012). In Turku, Finland, the ocean parameters are often more important than the urban development factors for controlling the spatial variability in the extreme air temperatures, especially for temperatures in coastal cities (Hjort et al., 2016). Therefore, in this study, to investigate UHI and improve the mitigation approach in the urban planning regulations, a HiSAN was used for long-term measurement. By combining the HiSAN measurement data and urban development parameters, including the ratios of various types of land use and geographic characteristics such as distance to the coast at each measurement point, the influence of each parameter on the thermal condition at various periods was quantified. By presenting the influence of each parameter, the thermal conditions can be estimated. However, most of the thermal condition prediction models were established based on a single period due to the lack of meteorological information (Smith et al., 2009; Ren et al., 2012; Ren et al., 2013; Chen et al., 2019). In this study, the HiSAN continuously

712

Y.-C. Chen et al. / Science of the Total Environment 685 (2019) 710–722

recorded data every hour from 2016 to the present. This amount of meteorological information is sufficient for analysis and for developing a prediction model for estimating air temperature during various periods, including daytime and night time and four seasons. By using this method, the UHI features can be detected more precisely, and the strategy of decreasing the intensity of UHI can also be based on the results of different areas and different periods. The HiSAN is a platform for visualizing the thermal conditions and serves as a platform to connect architects and urban planners who do not have a meteorological background. The network can be used to understand the effect of a built environment on the climate so as to make appropriate decisions on urban planning and designs to alleviate the growing problem of thermal stress and UHI (Chen et al., 2018; Lin et al., 2019).

2. Methodology 2.1. Study area In this study, Tainan city was selected as the main research area (22° 59′N, 120° 11′E). The total area of the city is 23 × 28 km2. The terrain on the Jianan plain is flat without any towering terrain and has an elevation from 0 to 15 m in the study area. Therefore, the elevation factor can be removed from the factors influencing the thermal conditions in this area (Fig. 1). However, elevation should be considered for other cities with complex topographies. Tainan is located in Southern Taiwan, where the climate zone is tropical. The background thermal condition in Tainan is warm throughout the year. According to the observation data (Central Weather Bureau of Taiwan, 2017), the annual average temperature is 24.6 °C, the highest monthly average temperature is in July and is 30 0.4 °C, and the lowest monthly average temperature is in January and is 17.6 °C. The region receives annual sunshine equal to 2380 h with only 237.8 h of rainfall.

Tainan is a city with a long history of development. Various types of buildings are distributed in this area, thus causing the geometric pattern to have a complex composition and low homogeneity. As the city is located on the waterfront, the climatic condition is affected not only by the different built environments but also by the sea. Therefore, this study considers land use and land cover information and includes the information of the distance to the sea (Chen et al., 2018). 2.2. Research structure The primary framework of this study combines the meteorological data obtained from the HiSAN with the urban development data and geometry data to obtain their relationship results. The size of the buffer zone that provides the best performance for determining the thermal condition was confirmed. In the beginning, the urban development parameters were collected from various sources, including satellites, drones, and the government. Second, the meteorological data was classified into several periods, including daytime and night time and four seasons. The urban development factors were also selected in this step to construct an air temperature prediction model based on the correlation coefficient with the air temperature (Fig. 2). 2.3. HiSAN The HiSAN is an observation network that comprises hundred temperature (Ta) and relative humidity (RH) measuring points that are located on parcels of land used for various purposes and with land cover environments from 2016 to the present in Tainan (Fig. 3). To record long-term meteorological data, LOGPRO TR-32 (Tecpel Co. Ltd.), which has the advantages of high energy efficiency and large storage, was selected as the research instrument. LOGPRO TR-32 has a resolution of 0.1 °C and an accuracy of 0.5 °C for estimating Ta and has a resolution of 1% and accuracy of 5% for estimating RH. For protection, radiation covers were used to prevent the influence of radiation and damage by rain and wind. All the instruments

Fig. 1. Location of study area.

Y.-C. Chen et al. / Science of the Total Environment 685 (2019) 710–722

713

Fig. 2. Research structure.

were installed on utility poles at a height of 2.5 m to ensure good ventilation, pedestrian safety, and transportation safety. All points were arranged based on different types of urban environments, including different land uses and land covers, to represent the real shading and radiation absorption conditions in a complex urban area. Three principles were considered while selecting the points of the HiSAN. First, the measurement area was equally distributed into areas of 3 × 3 km2. Moreover, the measurement points were scattered evenly, which implies that there is at least one point in each area. In the second principle, regional features were diversified to obtain meteorological information from different regional features, including land use and land cover, and to prevent too many points in similar urban patterns. The points were assigned to various types of conditions. The third principle includes determining the homogeneity of green areas such as parks, forests, and campus to compare their ability under changing thermal conditions (Chen et al., 2018).

by the Central Weather Bureau, were used to determine the weather conditions of the day. A day was designated to be sunny when the solar radiation rate was N80% on that day and no rain fell. The data of sunny days were used for subsequent analysis. We also used the following criteria. The first criterion is season—summer (June to August), fall (September to November), winter (December to February), and spring (March to May). The second criterion is day and night—daytime (6:00–18:00) and night time (18:00–06:00).

2.4. Period classification

2.5.1. Building information The location and building area data were obtained from “Open Street Map” and “Taiwan e-Map” (National Land Surveying and Mapping Center, 2018). The building height information was acquired from the digital surface model that has a resolution of 1 m. The building information could be obtained by combining these data (Fig. 4). By assuming that one floor of a building is 3-m high, the number of floors of the building were calculated. Moreover, the

The measurement data obtained from the HiSAN from June 1, 2016, to May 31, 2017, was used in this study. This period is the most complete data record period to construct the UHI prediction model and evaluate the UHI characteristic in different periods. Due to the long-term and huge amounts of data obtained from the HiSAN and the complex background climate conditions, we first conducted weather screening. The meteorological data of the Tainan Climate Station, which is owned

2.5. Collection of urban development factors and geographic features This study examined the relationship between air temperature and environmental factors recorded at the HiSAN measurement points, as described in the following sections.

Fig. 3. The instrument and points distribution of HiSAN, the red points mean the surrounding area of instrument is urban area, and green points mean the surrounding area of instrument is greenery. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

714

Y.-C. Chen et al. / Science of the Total Environment 685 (2019) 710–722

Fig. 4. Land cover with height information and location.

distribution, floor area, and total floor area were calculated simultaneously.

2.5.2. Land use and land cover Land cover information was also important for investigating the land patterns in previous studies (Arnold Jr and Gibbons, 1996; Liang et al., 2012; Findell et al., 2017). The land use data are important for quantifying the thermal development condition and thermal intensity of a certain area. A “map of investigated land use results” made using Spot6, which has a resolution of 6 m, was used for classification by using the supervised approach to obtain the real conditions and land use information. The floor area, vegetation area, and road area ratios were calculated using this land use information and used to estimate the influence of each factor on the urban thermal condition (Fig. 5). The area and quantity of each land cover type was easily acquired for describing the development characteristics in different areas.

2.5.3. Distance to the sea In a previous study of the HiSAN, Chen et al. (2018) found that the spatial and temporal distribution of temperature in Tainan changed dynamically. The change was mainly in the east–west direction or toward the orientation of the land and sea. To include this effect in the multiple regression analysis, the distance to the sea was obtained from the HiSAN stations and was used as one of the geographic characteristics. 2.6. Establishment of the buffer zone To investigate the relationship between the air temperature measured using the HiSAN and the surrounding environment, a buffer zone analysis was conducted. This type of buffer zone is a circular area with a center and a radius; buffer zone analysis is an effective approach to quantify the urban development factors in an area (Kota et al., 2017; Kotharkar and Bagade, 2018). The 100 points of the HiSAN stations were considered as the centers of buffer zones and overlaid spatially with different urban development

Y.-C. Chen et al. / Science of the Total Environment 685 (2019) 710–722

715

Fig. 5. Land cover information in Tainan including (a) vegetation (the distribution of greenery including park, farmland and lawn), (b) building (the distribution of artificial obstacle including any kind of structure), (c) road (the distribution of artificial pavement including parking lot and plaza) and (d) water (the distribution of sea, rivers and fishfarm).

factors on the quantum geographic information system. The area ratio of each urban development factor recorded by the HiSAN stations within the buffer zone was quantified and used to estimate the influence of the factors on temperature. Various buffer zone radius ranging from 100 to 500 m were tested to obtain the most appropriate range for describing the thermal conditions. The urban development factors of a radius of 300 m was selected because the factors at this value were the most interpretable for evaluating the relationships between the different ratios of each development factor to Ta in the study area. The measurement data of the HiSAN stations and the development factor of the 100 buffer zones were analyzed using regression. The correlation coefficient of each factor for the four seasons and for the night time and the daytime was obtained. To indicate the relationships between Ta and the development factors, we classified the urban development factors including high water

area ratio, high building area ratio, high vegetation area ratio, and high road area ratio, and geographic features including “far from the sea” and “near the sea” into different groups (Fig. 6) to illustrate the warming or cooling effect of each buffer zone area as shown in Table 1. 2.7. Calculation of the UHI intensity The correlation coefficients between the urban development factors and temperature data exhibit different changes for different periods; the relationships between different factors exhibit different intensities of effects due to seasonal and daytime–night-time changes. However, the analysis results can only show the correlation coefficients between temperature and urban development factors. Therefore, this study selected the representative stations of various urban development factors and geographic features, including ‘near the sea’ and ‘far

716

Y.-C. Chen et al. / Science of the Total Environment 685 (2019) 710–722

Fig. 6. The classify of urban development factors and geography feature.

from the sea’ distances, high floor area ratio, high road area ratio, high vegetation area ratio, and high water area ratio, to calculate the mean Ta of each group. Moreover, the obtained Ta value was used to subtract the average temperature of all stations for obtaining the UHI intensity by analyzing Ta deviation in different periods to determine the changes in the seasons and days and nights:

selected for establishing the model. The urban development factors and geographic features were selected to calibrate with a standard weather station through the regression model for different periods and were applied to estimate the thermal condition at various times to ensure the best predicting ability of the model. The temperature of each location could be predicted using the standard temperature of the Tainan Station:

Ta deviation ¼ average value of Ta in a group−average value of Ta of all stations

ð1Þ

△Ta ¼ f ðurban development factor; geographic characteristic factorÞ ð3Þ

2.8. Prediction model of air temperature The situation and ratio of land pattern in each buffer zone was obtained to predict the temperature in the region. The temperature was estimated using the floor area ratio, road area ratio, vegetation area ratio, water area ratio, and distance to the sea, as mentioned in Section 2.5. We divided the data based on four seasons and daytime and night time. Thus, eight temperature estimation models were obtained. As the temperature predicted by a standard reference station is required as a reference value for subsequent predictions, the Tainan weather station was used as the reference station. The value predicted by the model is the temperature difference (△Ta) between the Tainan weather station and the HiSAN: △Ta ¼ temperature data from HiSAN −temperature data from the Tainan Weather Station

ð2Þ

Regarding the urban development factors in the prediction model, the top three factors that have the highest influence on △Ta were Table 1 The definition of urban develop factors and geography feature. Parameters

Class

Definition

Points

Distance to the Sea

Far from the sea Near the sea High ratio of water area High ratio of building area High ratio of vegetation area High ratio of road area

b2KM N12KM~20KM b51–83% b51–66% 51–92% 20–28%

16 points 16 points 7 points 9 points 17 points 22 points

Water area Building area Vegetation area Road area

3. Results and discussion 3.1. UHI intensity in different areas In the daytime, the mean temperature of the representative station of each factor is subtracted from the average temperature of all the stations. The temperature difference is obtained for different seasons and during the daytime (Fig. 7a). In Fig. 7, the values on the x-axis signify the deviation in the average Ta of each representative station group and the average Ta of all the HiSAN stations in the daytime during all four seasons. Zero implies the average Ta of all the HiSAN stations at the corresponding time. The positive value implies that the average Ta of a certain group is higher than the average Ta of all the HiSAN stations, and vice versa. The results revealed that the stations in the ‘near the sea’ group have lower Ta deviation, and stations in the ‘far from the sea’ group have higher Ta deviation. The difference in Ta deviations between these two groups was nearly 1 °C for each season. As the specific heat of soil was five times lower than that of the water, the aforementioned phenomenon can be due to the temperature adjustment function of the sea because of its huge heat storage ability. In terms of the urban development factors, although water and vegetation are both permeable, the effect of these factors on Ta in the daytime is very different. The evaporation of water during the daytime by insolation causes the ambient temperature to decrease; therefore, the cooling effect in the water area is more significant than that in the vegetation areas. In all seasons, the Ta deviations were nearly −1 °C in the high water area ratio group.

Y.-C. Chen et al. / Science of the Total Environment 685 (2019) 710–722

717

Fig. 7. Air temperature deviation of different group in (a) daytime (b) night time.

The groups of high floor area and road area ratios indicated the warming effect of the impermeable materials. During the daytime, the pavements and buildings absorb solar radiation, and the Ta deviations were approximately 0.5 °C. During the night time, some factors revealed the opposite effect compared with those during the daytime, including near or far distance to the sea, high vegetation area ratio, and high water area ratio. The distance factor showed opposite results because of the heat storage of the sea. The sea cools slower than the land, and the Ta deviation near the sea becomes higher than that far from the sea (Fig. 7b). In all seasons, the high vegetation area ratio during the night time presents a low Ta deviation. Especially, in summer, the ratio reached nearly −1 °C. This result reveals that the evaporation of the moisture of vegetation present in the air can effectively to reduce the temperature in the vicinity. However, during the daytime, this phenomenon is diluted by the heat from the strong solar radiation and reflection of the land surface. 3.2. Hourly variation in the correlation coefficient between air temperature and different factors 3.2.1. Distance to the sea As the study area is a coastal city, the distance to the sea is one of the factors that significantly affect Ta. The correlation coefficient for each

month presented in Fig. 8a shows a significant decrease in Ta in the area far from the sea during the night time and shows a significant negative correlation with Ta during the night time from 00:00 to 06:00. When the sun rises after 6:00, the land surface and the sea begin to heat up. The specific heat of the sea becomes much warmer, thus making the air in the coastal areas that are closer to the sea relative cooler than the air in the areas far from the sea. After 18:00, the area far from the sea begins to cool. The more inland an area is, the lower that area's Ta is. The difference between the distance to the sea for the seasons is mainly due to the higher correlation between March during the night time and June during the daytime.

3.2.2. Floor area The floor area ratio reveals a positive correlation with the building area and Ta. As all the land properties on the entire area during the day receive direct sunlight, the area can easily present a high thermal condition. Therefore, the correlations during the daytime in all the seasons are low (Fig. 8b). However, due to the large heat absorption by the buildings during the night time, the buildings begin to radiate short-wave radiation when there is no direct solar radiation at night. This effect causes a slow decrease in the Ta of the area with a high building area ratio at

718

Y.-C. Chen et al. / Science of the Total Environment 685 (2019) 710–722

Fig. 8. The hourly variation of the correlation coefficient between air temperature and (a) distance to the sea, (b) floor area, (c) road area, (d) vegetation area, and (e) water area.

night and causes the urban Ta to be higher than that in other areas. Therefore, all the seasons have a high correlation at night. 3.2.3. Road area The road system is an artificial, impervious surface pavement in the city, and its distribution range is more comprehensive than that of buildings. As the result, the road area ratio has a higher correlation with Ta in this study. The results reveal that the larger the area is, the higher the Ta is. This trend is similar to that of the building area ratio. When there is direct solar radiation during the daytime, the correlation with the Ta is low, for the same reason as for the floor area: all the long-wave and short-wave radiations heat the Ta to a high thermal

condition, thereby reducing the ability of the road area to increase Ta. Even though the artificial pavement absorbs heat during the daytime, the heat is emitted after 18:00 at night time, thus increasing the correlation coefficient (Fig. 8c). 3.2.4. Vegetation area It is well known that areas covered in vegetation can cool a city. However, the line presented in Fig. 8d indicates a high negative correlation after 14:00. This result indicates that the cooling effect of the vegetation area begins in the afternoon and continues till 06:00 in the morning, thus achieving the maximum cooling effect of the vegetation area.

Y.-C. Chen et al. / Science of the Total Environment 685 (2019) 710–722

719

Table 2 The coefficient of prediction model in different period. Spring

Distance to the sea (km) Floor area Road area Vegetation area Water area Constant R2

Summer

Fall

Day time

Night time

Day time

Night time

0.0422

−0.054

0.0365

−0.063

1.79

3.60 −0.97

−1.05 −0.022 0.47

−0.395 0.72

−2.35 −0.96 0.455 0.40

The highest solar radiation intensity occurs in the summer; therefore, a district with a high vegetation area ratio has higher radiation reflectivity than a district with a high artificial pavement ratio. Therefore, the heat absorption in an area with a high vegetation area ratio is lower than that in an area with a high artificial pavement ratio. Based on this effect, a significant cooling effect occurs at night in June owing to the high vegetation area ratio. 3.2.5. Water area The water area is also considered as one of the factors that improve the urban thermal condition because it has a significant cooling effect during the day. The trends in each season are close owing to the fact that the specific heat of water is relatively greater than that of land. Therefore, during daytime hours with the same amount of radiation, the temperature of the water rises slowly while the temperature of the land rises faster. During the night time, the water emits the same amount of heat as emitted by the land. However, the Ta of water declines slowly because the water releases more heat, thus causing the temperature to drop by 1 °C. In Fig. 8e, after 08:00, the land and water bodies undergo a large amount of radiation to increase Ta because the sun rises. However, compared to the land, the same heat can increase the Ta of water relatively little owing to the great specific heat of the water. Therefore, the districts with higher water area ratio have low Ta, and the maximum cooling effect is achieved with the strongest radiation at 12:00 in the noon. 3.3. Prediction model Based on the results of the relationship between Ta data and the urban development and geographic factors, we developed a multiple regression model to predict ΔTa in any location of the study area for eight periods, as aforementioned in Section 2.8. The coefficients of the models are shown in Table 2. To simplify the equation, three important factors were used as the parameters at each period. The squares filled in gray in the table are the factors used in the period for predicting ΔTa. The positive and negative values of the square represent the predicted trend of the factor affecting the Ta. For example, the model of summer and night time is expressed as follows: the distance to the sea and the vegetation area have negative effects on ΔTa, whereas the road area has positive effects on ΔTa: ΔTasummer night time ¼ −0:063  Distance to the sea þ 2:27  Road area−1:78  Vegetation area þ 0:275 ð4Þ

3.4. Comparison between the prediction and measurement results It is difficult to make the HiSAN constantly measure data because of the difficulty of maintaining the instrument and the time-consuming data collection process. By applying the model presented in Table 2, it

Day time 0.035 0.79

2.27 −1.78 0.275 0.74

−0.97 −0.086 0.38

Winter Night time

Day time

Night time

−0.034

0.0294

−0.036

2.92 −2.86

1.14

3.25 −1.49

−0.433 0.60

−1.40 −0.004 0.30

−0.573 0.46

is possible to estimate the on ΔTa and also Ta by determining the temperature of the Tainan Climate Station. The prediction model developed for summer and night time in 2016 was used to estimate the Ta during the summer and at night time in 2018 to validate the prediction and compare the measured and predicted temperature (Fig. 9). The results revealed that the prediction model can estimate the Ta distribution trend and hotspot locations well due to the 100 measurement points of the HiSAN, which are located in different urban environments, that are considered during the development of the prediction model. The data of at points was corrected by conducting data validation by using the standard weather station data. The correlation between the measured and predicted temperature was 0.67 (Fig. 10). The model has the potential to evaluate the thermal condition to prevent disasters, including heat stress, agricultural problems due to an increase in Ta in coastal areas, and diseases due to thermal conditions, not only in the present but also in the future. 4. Conclusion In this study, a high-density Ta observation network known as the HiSAN was applied to determine the relationships between the Ta and the urban development and geographic factors. The factors include high ratio of buildings, vegetation, road and water areas, and the distance of each measurement station in Tainan from the sea. In terms of the thermal condition, owing to the adequate information of the HiSAN, the Ta distribution condition and the correlation with various factors can be determined for different periods, including the daytime and night time for the four seasons. This study also utilized buffer zones to quantify the urban development and geographic factors inside the 300-m radius area. The three main findings of this study are addressed below: the importance of geographic features, different effects of the factors in different seasons, and the application of visualizing the urban thermal condition. First, the distance to the sea strongly affects the temperature distribution during the day and night. As the study area is a coastal city, the specific heat of the sea is much higher than that of the land. From the multiperiod correlation analysis, the geographic feature of the distance to the sea can be found using the high correlation coefficient in the multiple regression analysis of all seasons. This implies that “the distance to the sea” is one of the main factors that affect Ta variation. Second, different seasons have different dominant factors that affect temperature. In this study, various factors were tested based on different seasons and times. The factors affecting the temperature variation during the daytime and night time differ for each season. For example, the permeable pavement of both the vegetation and water areas has a different environmental cooling effect during different periods. Finally, regarding the application of the approach for developing future urban planning regulations, as this study discussed temperature variation caused by urban development factors and geographic features, the results reflect the influence of land cover and land use on the increase and decrease in temperature. For land use, consider a daytime

720

Y.-C. Chen et al. / Science of the Total Environment 685 (2019) 710–722

Fig. 9. The comparison of the predicted and measured temperature. (a) predicted temperature distribution in summer night time in 2018, (b) measured temperature distribution in summer night time in 2018.

Y.-C. Chen et al. / Science of the Total Environment 685 (2019) 710–722

Fig. 10. The relation between predicted temperature and measured temperature.

scenario as an example. In this case, the built-up area includes industrial districts, commercial districts, and tourist attractions. This type of land use is mainly concentrated on daytime use. Based on the results of this study, the districts that have high land use intensity are suggested to be located near the coastal area to achieve a comfortable, lowtemperature background environment during the daytime. For the land cover, water can be considered as a permeable factor within a radius of 300 m to provide better thermal conditions in urban areas. Therefore, this study not only can present the thermal condition characteristics in urban areas but also can provide meteorological knowledge to urban planners and architects who should consider the urban climate. In the future, wearable observations system or small meteorological instruments can also be applied to further explore the details and features of microclimate conditions in blocks and alleys, which will be helpful to comprehend the issues of mitigating the severe climatic problems, including high UHIs and climate changes, when conducting urban planning and building designs. Acknowledgements The authors would like to thank the Ministry of Science and Technology of Taiwan, for financially supporting this research under Contract No. 107-2221-E-006-048-MY3. References Acero, J.A., Herranz-Pascual, K., 2015. A comparison of thermal comfort conditions in four urban spaces by means of measurements and modelling techniques. Build. Environ. 93, 245. Alcoforado, M.J., Andrade, H., Lopes, A., 2009. Application of climatic guidelines to urban planning: the example of Lisbon (Portugal). Landsc. Urban Plan. 90, 56–65. Allen, L., Lingberg, F., Grimmond, C.S.B., 2011. Global to city scale urban anthropogenic heat flux: model and variabilityInt. J. Climatol. 31, 1990–2005. Arnold Jr., C.L., Gibbons, C.J., 1996. Impervious surface coverage: the emergence of a key environmental indicator. J. Am. Plan. Assoc. 62 (2), 243–258. Baker, L.A., Brazel, A.J., Selover, N., Martin, C., McIntyre, N., Steiner, F.R., Musacchio, L., 2002. Urbanization and warming of phoenix (Arizona, USA): impacts, feedbacks and mitigation. Urban Ecosyst 6 (3), 183–203. Central Weather Bureau of Taiwan, 2017. Monthly report on climate system. http://www. cwb.gov.tw/V7/forecast/long/long_season.htm. Chang, B., Wang, H.Y., Peng, T.Y., Hsu, Y.S., 2010. Development and evaluation of a city-wide wireless weather sensor network. J. Educ. Technol. Soc. 13 (3), 270–280. Chen, Y.C., Chen, C.Y., Matzarakis, A., Liu, J.K., Lin, T.P., 2016a. Modeling of mean radiant temperature based on comparison of airborne remote sensing data with surface measured data. Atmos. Res. 174, 151–159. Chen, Y.C., Lin, T.P., Lin, C.T., 2016b. A simple approach for the development of urban climatic maps based on the urban characteristics in Tainan, Taiwan. Int. J. Biometeorol. 61 (6), 1029–1041.

721

Chen, Y.C., Yao, C.K., Honjo, T., Lin, T.P., 2018. The application of a high-density street-level air temperature observation network (HiSAN): dynamic variation characteristics of urban heat island in Tainan, Taiwan. Sci. Total Environ. 626, 555–566. Chen, Y.C., Lo, T.W., Shih, W.Y., Lin, T.P., 2019. Interpreting air temperature generated from urban climatic map by urban morphology in Taipei. Theor. Appl. Climatol. https://doi.org/10.1007/s00704-018-02764-x. Chun, B., Guhathakurta, S., 2017. Daytime and nighttime urban heat islands statistical models for Atlanta. Environ. Plan. B Urban Anal. City Sci. 44 (2), 308–327. Dawod, G., Mirza, M., Al-Ghamdi, K., 2011. GIS-based spatial mapping of flash flood hazards in Makkah city, Saudi Arabia. J. Geogr. Inf. Syst. 3 (3), 217–223. Dimitriou, E., Zacharias, I., 2010. Identifying microclimatic, hydrologic and land use impacts on a protected wetland area by using statistical models and GIS techniques. Math. Comput. Model. 51, 200–205. Eliasson, I., Svensson, M., 2003. Spatial air temperature variations and urban land use-a statistical approach. Meteorol. Appl. 10 (02), 135, 135–149. Feng, J.M., Wang, Y.L., Ma, Z.G., Liu, Y.H., 2012. Simulating the regional impacts of urbanization and anthropogenic heat release on climate across China. J. Clim. 25 (20), 7187–7203. Findell, K.L., Berg, A., Gentine, P., Krasting, J.P., Lintner, B.R., Malyshev, S., Santanello Jr., J.A., Shevliakova, A., 2017. The impact of anthropogenic land use and land cover change on regional climate extremes. Nat. Commun. 8, 989. Heusinkveld, B.G., Hove, L.W.A. van, Jacobs, C.M.J., Steeneveld, G.J., Elbers, J.A., Moors, E.J., Holtslag, A.A.M., 2010. Use of a mobile platform for assessing urban heat stress in Rotterdam. Proc. 7th Conf. On Biometeorology, Freiburg, Germany. 433–438. Hjort, J., Suomi, J., Käyhkö, J., 2016. Extreme urban–rural temperatures in the coastal city of Turku, Finland: Quantification and visualization based on a generalized additive model. Sci. Total Environ. 569-570, 507–517. Honjo, T., 2012. Daily movement of heat island in Kanto area. 6th Japanese-German Meeting on Urban Climatology. Hiroshima Institute of Technology, Japan. Honjo, T., Yamato, H., Mikami, T., Grimmond, C.S.B., 2015. Network optimization for enhanced resilience of urban heat island measurements. Sustain. Cities Soc. 19, 319–330. Jittawikul, A., Saito, I., Ishihara, O., 2004. Climatic maps for passive cooling methods utilization in Thailand. J Asian Archit Build 3 (1), 109–114. Kaloustian, N., Bechtel, B., 2016. Local climatic zoning and urban Heat Island in Beirut. Procedia Eng 169, 216–223. Kaplan, G., Avdan, U., Avdan, Z.Y., 2018. Urban heat island analysis using the Landsat 8 satellite data: a case study in Skopje. Macedonia. Proceedings. 2 (7), 58. Koellner, T., Scholz, R.W., 2007. Assessment of land use impacts on the natural environment. Int. J. Life Cycle Assess. 12 (1), 16, 16–23. Kota, N., Shankaraiah, K., Pamu, G., Suresh, M., 2017. Land use/land cover characteristics in and around Suryapet Town, Nalgonda District for urban planning. J Telecommun Syst Manage 6 (1), 148. Kotharkar, R., Bagade, A., 2018. Evaluating urban heat island in the critical local climate zones of an Indian city. Landsc Urban Plan Landscape and Urban Planning 169, 92–104. Kotharkar, R., Surawar, M., 2015. Land use, land cover, and population density impact on the formation of canopy urban heat islands through traverse survey in the Nagpur urban area. India. J Urban Plann Dev. 142 (1). Li, J., Song, C., Cao, L., Zhu, F., Meng, X., Wu, J., 2011. Impacts of landscape structure on surface urban heat islands: a case study of Shanghai, China. Remote Sens. Environ. 115 (12), 3249–3263. Liang, B.P., Li, Y., Chen, K.Z., 2012. A research on land features and correlation between NDVI and land surface temperature in Guilin City. Remote Sens. Tech. Appl. 27, 429–435. Lin, T.P., Chen, Y.C., Matzarakis, A., 2017a. Urban thermal stress climatic mapping: combination of long-term climate data and thermal stress risk evaluation. Sustain. Cities Soc. 34, 12–21. Lin, T.P., Lin, F.Y., Wu, P.R., Hämmerle, M., Höfle, B., Bechtold, S., Hwang, R.L., Chen, Y.C., 2017b. Multiscale analysis and reduction measures of urban carbon dioxide budget based on building energy consumption. Energy Build 153, 356–367. Lin, F.Y., Huang, K.T., Lin, T.P., Hwang, R.L., 2019. Generating hourly local weather data with high spatially resolution and the applications in bioclimatic performance. Sci. Total Environ. 653, 1262–1271. Liu, L., Lin, Y., Liu, J., Wang, L., Wang, D., Shui, T., Chen, X., Wu, Q., 2017. Analysis of local-scale urban heat island characteristics using an integrated method of mobile measurement and GIS-based spatial interpolation. Build. Environ. 117, 191–207. Lo, C.P., Quattrochi, D.A., Luvall, J.C., 1997. Application of high resolution thermal infrared remote sensing and GIS to assess the urban heat island effect. Int. J. Remote Sens. 18, 287–304. Lu, J., Li, C., Yu, C., Jin, M., Dong, S., 2012. Regression analysis of the relationship between urban heat island effect and urban canopy characteristics in a mountainous city, Chongqing. Indoor and Built Environ 21 (6), 821–836. Matzarakis, A., Röckle, R., Richter, C.J., Höfl, H.C., Steinicke, W., Streifeneder, M., Mayer, H., 2008. Planungsrelevante Bewertung des Stadtklimas—Am Beispiel von Freiburg im Breisgau. Gefahrstoffe - Reinhalt. Luft. 68, 334–340. Merbitz, H., Buttstädt, M., Michael, S., Dott, W., Schneider, C., 2012. Gis based identification of spatial variables enhancing heat and poor air quality in urban areas. Appl. Geogr. 33, 94–106. Mikami, T., Yamato, H., Extended, M.R.G., 2011. High-resolution temperature observations using extended-METROS in the Tokyo metropolitan area and their urban climatological significance. J. Geogr. 120 (2), 317–324. Nakayoshi, M., Kanda, M., Shi, R., de Dear, R., 2015. Outdoor thermal physiology along human pathways: a study using a wearable measurement system. Int. J. Biometeorol. 59, 503–515.

722

Y.-C. Chen et al. / Science of the Total Environment 685 (2019) 710–722

Naserikia, M., Asadi Shamsabadi, E., Rafieian, M., Leal Filho, W., 2019. The urban Heat Island in an urban context: a case study of Mashhad, Iran. Int. J. Environ. Res. Public Health 16, 313. National Land Surveying and Mapping Center, 2018. Taiwan e-Map https://www.nlsc. gov.tw/EN. Oke, T.R., 1973. City size and the urban heat island. Atmos. Environ. 7, 769–779. Pigliautile, I., Pisello, A.L., 2018. A new wearable monitoring system for investigating pedestrians' environmental conditions: development of the experimental tool and start-up findings. Sci. Total Environ. 630, 690–706. Ren, C., Ng, E., Katzschner, L., 2010. Urban climatic map studies: a review. Int. J. Climatol. 31 (15), 2213–2233. Ren, C., Spit, T., Lenzholzer, S., Yim, H.L.S., Heusinkveld, B., van Hove, B., Chen, L., Kupski, S., 2012. Burghardt., Katzschner, L. Urban climatic map system for dutch spatial planning. Int J Appl Earth Obs Geoinf. 18, 207–221. Ren, C., Lau, K.L., Yiu, K.P., Ng, E., 2013. The application of urban climatic mapping to the urban planning of high-density cities: the case of Kaohsiung, Taiwan. Cities 31 (1–16). Roth, M., Oke, T.R., Emery, W.J., 1989. Satellite-derived urban heat islands from three coastal cities and the utilization of such data in urban climatology. Int. J. Rem. Sens. 10, 1699–1720.

Saaroni, H., Ben-Dor, Eyal., Bitan, A., Potchter, Oded., 2000. Spatial distribution and microscale characteristics of the urban heat island in Tel-Aviv, Israel. Landsc. Urban Plan. 48, 1–8. Smith, C., Lindley, S., Levermore, G., 2009. Estimating spatial and temporal patterns of urban anthropogenic heat fluxes for UK cities: the case of Manchester. Theor. Appl. Climatol. 98, 19–35. Son, N.T., Chen, C.F., Chen, C.R., Thanh, B.X., Vuong, T.H., 2017. Assessment of urbanization and urban heat islands in Ho Chi MinhCity, Vietnam using Landsat data. Sustain. Cities Soc. 30, 150–161. Weng, Q., Rajasekar, U., Hu, X., 2011. Modeling urban heat islands and their relationship with impervious surface and vegetation abundance by using aster images. Geosci. Remote Sens. 49 (10), 4080–4089. Xiao, R.B., Ouyang, Z., Zheng, H., Li, W.F., Erich, S., Wang, X.K., 2007. Spatial pattern of impervious surface and their impacts on land surface temperature in Beijing, China. J. Environ. Sci. 19 (2), 250–256.