Modeling of mean radiant temperature based on

110 downloads 0 Views 2MB Size Report
weather station, microscale ground surface measurements, land surface temperature (LST) and ...... 23rd International Conference on Image and Vision Com-.
ATMOS-03595; No of Pages 9 Atmospheric Research xxx (2016) xxx–xxx

Contents lists available at ScienceDirect

Atmospheric Research journal homepage: www.elsevier.com/locate/atmosres

Modeling of mean radiant temperature based on comparison of airborne remote sensing data with surface measured data Yu-Cheng Chen a, Chih-Yu Chen a, Andreas Matzarakis b, Jin-King Liu c, Tzu-Ping Lin a,⁎ a b c

Department of Architecture, National Cheng Kung University, 1 University Rd., Tainan 701, Taiwan Research Center Human Biometeorology, Deutscher Wetterdienst, Germany LIDAR Technology Co., Ltd, Hsinchu County 30274, Taiwan

a r t i c l e

i n f o

Article history: Received 31 July 2015 Received in revised form 22 December 2015 Accepted 5 January 2016 Available online xxxx Keywords: Mean radiant temperature LIDAR Thermal comfort Urban surfaces Microscale modeling

a b s t r a c t Assessment of outdoor thermal comfort is becoming increasingly important due to the urban heat island effect, which strongly affects the urban thermal environment. The mean radiant temperature (Tmrt) quantifies the effect of the radiation environment on humans, but it can only be estimated based on influencing parameters and factors. Knowledge of Tmrt is important for quantifying the heat load on human beings, especially during heat waves. This study estimates Tmrt using several methods, which are based on climatic data from a traditional weather station, microscale ground surface measurements, land surface temperature (LST) and light detection and ranging (LIDAR) data measured using airborne devices. Analytical results reveal that the best means of estimating Tmrt combines information about LST and surface elevation information with meteorological data from the closest weather station. The application in this method can eliminate the inconvenience of executing a wide range ground surface measurement, the insufficient resolution of satellite data and the incomplete data of current urban built environments. This method can be used to map a whole city to identify hot spots, and can be contributed to understanding human biometeorological conditions quickly and accurately. © 2016 Elsevier B.V. All rights reserved.

1. Introduction 1.1. Remote sensing application on urban thermal environment Previous research on remote sensing in urban thermal environments has commonly involved the use of data captured by thermal infrared imager to determine LST and air temperature (Ta), which are utilized to analyze the severity of the urban heat island effect (Dozier, 1981; Lo et al., 1997). The detection of land use and land cover are also used to calculate such indices as the Normalized Difference Vegetation Index (NDVI) and the percentage impervious surface area (ISA), which are highly related to the thermal environment (Zhang et al., 2008; Sun & Kafatos, 2007). Matzarakis et al. (2008) used a high-resolution thermal imager to identify the highest and lowest LST in one day in Freiburg. Their data include not only surface temperature, but also information related to surface radiation; they finally incorporate point information to produce a map of thermal comfort conditions based on the thermal index of

⁎ Corresponding author at: Department of Architecture, National Cheng Kung University, 1 University Rd., East Dist., Tainan 701, Taiwan. E-mail addresses: [email protected] (Y.-C. Chen), [email protected] (C.-Y. Chen), [email protected] (A. Matzarakis), [email protected] (J.-K. Liu), [email protected] (T.-P. Lin).

Tmrt. Their result shows that, on account of building density and green area, regions that are farther from downtown have a lower Tmrt, because the lower sky view factor in downtown areas indicates great shelter, so LST drops more slowly than in rural areas during cooling. For example, in Guilin city, (Liang et al., 2012), the LST and NDVI were calculated using the Landsat Thematic Mapper (TM), with the purpose of elucidating the correlation between LST and NDVI. The highest mean NDVI was observed in the forest and the lowest NDVI was found in water, whereas the highest LST was found in construction areas and the lowest in the forest, revealing that the relationship between NDVI and LST is significantly negative. In Minnesota, Yuan and Bauer (2007) compared the ISA and NDVI indicators with LST in an urban area, using TM and enhanced thematic mapper plus (ETM +) data. They found that the ISA is strongly linearly related with LST in all four seasons, and the relationship between NDVI and LST is weaker. A comparison of these two indicators can support research into urban heat islands, based on remotely sensed data. The need for infrastructure from extreme conditions and modifications to existing conditions caused by changes in landforms have motivated the development of LIDAR as a novel remote sensing technology to gather data about the land surface. It has been widely used in observing large areas of land, recording topographical information (Bowen & Waltermire, 2002; Shan & Sampath, 2005). LIDAR has become a very important and valuable remote sensing observation technology in recent years due to its very high resolution.

http://dx.doi.org/10.1016/j.atmosres.2016.01.004 0169-8095/© 2016 Elsevier B.V. All rights reserved.

Please cite this article as: Chen, Y.-C., et al., Modeling of mean radiant temperature based on comparison of airborne remote sensing data with surface measured data, Atmos. Res. (2016), http://dx.doi.org/10.1016/j.atmosres.2016.01.004

2

Y.-C. Chen et al. / Atmospheric Research xxx (2016) xxx–xxx

This study uses LIDAR point cloud data directly to create a Digital Surface Model (DSM) map of the study area. The resolution of the DSM map that is based on data from LIDAR can reach 0.5–2 m, which is better than the resolution of 30–40 m achieved using satellites. Moreover, building heights, building areas, vegetation and planimetric features can be obtained by LIDAR. This study analyzes the impact of the sky view factor (SVF) and the shading factor on LST to elucidate an urban thermal environment to quantify the thermal comfort of humans. 1.2. Related issues and absence of remote sensing in previous studies Thermal infrared (TIR) imagers on satellites have previously been used to estimate the Ta from the LST (Price, 1983; Li et al., 2004; Aniello et al., 1995). However, this method is restricted by spatiotemporal factors. For example, the satellite passes through a fixed location, limiting the observable area. An LST photograph taken by such a thermal imager cannot be captured in various periods. Additionally, certain weather conditions, such as cloud cover, reduce the ability to obtain data. Therefore, the ability to verify thermal imaging data, and to use the method at different times, is limited. In contrast, LIDAR and a TIR imager can be used to make observations and obtain DSM and LST data at any time. Due to the high flexibility of airborne LIDAR and the TIR imager, obscuring clouds and weather conditions do not pose problems. LIDAR and TIR can generate useful data faster than a satellite (Vallet, 2008; Cook et al., 2013). The resolution of images captured by a satellite is generally not high enough to analyze the thermal environment of a small area and the LST associated with different pavement materials (Neale et al., 2009). The DSM obtained using airborne LIDAR includes the heights of objects on the ground, supporting environmental analysis (Sohn & Dowman, 2007). Information about vertical surfaces, such as those of buildings and vegetation, which may provide shade, cannot be easily retrieved from satellite images. Shade and LST affection, such as SVF and the ratio of the height of buildings to the width of streets, clearly affect the thermal environment (Chen et al., 2012). Failure to consider such factors reduces the accuracy of estimation. This study considers the following three key issues. 1. Urban environments must be represented accurately and in their current situation, because shading significantly changes the thermal environment. 2. The resolution of satellite images is too low to support analysis of human biometeorological conditions. 3. Land surface temperatures obtained from images from a thermal imager must be calibrated using not just one equation, but separately for shaded and non-shaded areas. 1.3. Research purpose Urban heat islands and global warming have made in temperature increases in urban areas an urgent issue (Lin et al., 2011). Since largescale surface measurements are difficult to make, this study used the airborne remote sensing device to estimate human biometeorological conditions. For this purpose, the Tmrt index is required, as it incorporates radiation and the shading of a location. The solar radiation obviously affects temperature. Various data are considered in the calculation of the most accurate Tmrt that can be estimated from data of airborne LIDAR and TIR image. This study addresses the issues raised in Chapter 1.2 by pursuing the following three goals. 1. To confirm that high-resolution LIDAR can accurately present the current urban built environment and capture the precise SVF, and the thermal images are consistent with surface measurements thereof. The accuracy of the calculation of SVF using the DSM is verified by comparison with the SVF calculated from fish-eye photographs. The accuracy of the thermal images is confirmed by comparison with surface measurements by thermocouple.

2. To estimate human biometeorological conditions based on Tmrt, which is highly related to radiation factors, such as shading and LST. High-resolution LIDAR and thermal image data are combined. Shaded areas are identified by simulation model (SkyHelios) using LIDAR DSM data, to analyze the relationship between Tmrt and shading. Finally, the LST data from thermal image is used to increase the accuracy of Tmrt estimation. 3. To calculate Tmrt using various combinations of methods, involving airborne and surface measurements as well as data from weather stations. The best approach to assess human biometeorological conditions is thus identified. This study adopts various methods to estimate Tmrt by comparing airborne LIDAR data with surface measurement data, in order to generate a Tmrt distribution map to visualize the urban thermal environment.

2. Method 2.1. Study area Banqiao district (25.0096703°N 121.4590989°E) located in northern Taiwan, is the major business center of New Taipei City. The population density of Banqiao is 24,027 km2 (2015.1). Most of the terrain is flat. The lowest average monthly temperature is 15.2 °C in January. The highest average monthly temperature is 28.3 °C in July. The Banqiao district is frequently the hottest area in Taiwan, so the problem of thermal stress in Banqiao district must urgently be solved. This study is based on an airborne photographic survey of an area with a length of 1.6 km and a width of 2.5 km in the Banqiao central business district. The area includes buildings with multiple uses, land uses and land coverage, such as train stations, MRT stations, the city hall, shopping centers, playgrounds, parks, and rivers. The highresolution airborne remote sensing data demand extensive information calculations in the modeling process. Accordingly, the most highly developed area of Banqiao with red frame in Fig. 1 was selected for visualization and further analysis. 2.2. Airborne measurement This study used two sets of equipment to conduct the airborne measurement survey at 08:45 AM on 2015/09/14. The first equipment was LIDAR, which is a remote sensing technology for observing surface obstacles based on the times of reflection of various lasers. The elevation and location of surface obstacles in Banqiao were clearly identified by GPS positioning. The point density of the high-resolution LIDAR Leica ALS-60 in this study is 2 points per m2; the spatial resolution is 1 m × 1 m per pixel, and the elevation precision is 0.18 m. The DSM covers an elevation range of 227.16 m to − 7.97 m in Banqiao. The laser can measure distance with high accuracy, and surface DSM can be obtained. Therefore, building heights, building areas, vegetation and surface objects can be obtained and calculated to determine the SVF and the shading simulation. The TIR imager can be used to make land surface temperature measurements rapidly over a wide area. The imager records information, including temperature and radiation inertia, from the reflection and emission of infrared light by surface objects. A thermal sensor detects the infrared radiation from objects, and generates an electronic image, based on that information, that captures temperature differences. The resolution of the TABI-1800 Thermal Imagery TIR imager is 0.5 m × 0.5 m per pixel, and the accuracy is 0.05 °C. The range of temperatures recorded in Banqiao is from 17.80 °C to 66.64 °C. The thermal image is used to retrieve the LST and calculate Tmrt from the urban environment information. These two technologies are combined to create an innovative method of assessing human bio-meteorological conditions in the urban environment based on DSM using the LIDAR, and LST using the TIR imager.

Please cite this article as: Chen, Y.-C., et al., Modeling of mean radiant temperature based on comparison of airborne remote sensing data with surface measured data, Atmos. Res. (2016), http://dx.doi.org/10.1016/j.atmosres.2016.01.004

Y.-C. Chen et al. / Atmospheric Research xxx (2016) xxx–xxx

3

Fig 1. Location of study area and demonstrated area.

2.3. Surface survey measurement

2.4. Calculation of sky view factor

As well as airborne measurements, this study performed surface measurements to calibrate the DSM and LST obtained from airborne remotely sensed data. The study area had 13 surface meteorological measurement points in different land cover. These data were collected for correction with the surface temperature that was obtained from the thermal images. The Tmrt values were estimated from the Ta, relative humidity, LST and the images obtained using a fish-eye camera. Two types of surface measurement were performed in this study. The first type of measurements, including Ta, relative humidity and globe temperature, was performed at 13 points using a Center314 sensor to calculate the real Tmrt. The second type of measurements, including the real LST, was measured using a Center310 device and a thermocouple to identify 11 pavement types as shown in Table 1, and was adopted to correct the LST obtained from the TIR images. The fish-eye camera was used to capture a real SVF image at each point to verify the accuracy of the SVF that is obtained from LIDAR. The same measuring instruments were used at all measurement points: the tripod was 1 m high, and the instrumentation was protected with aluminum. Radiation data were obtained from the Banqiao weather station close to the study area, as shown in Fig. 1, and were utilized to estimate Tmrt.

To elucidate the characteristics of the high-development area in the urban environment of the Banqiao district, the LIDAR DSM data was used to create a layer of SVF using the climatology calculating models RayMan and SkyHelios (Matzarakis et al., 2007, 2010; Matzarakis & Matuschek, 2011). The SVF value represents the extent to which the sky is shaded by trees or buildings, and specifies the area of the open sky above an area as a proportion of the overall visual area; it is used to evaluate outdoor thermal stress (Oke, 1981, 1987). In this study, the SVF was calculated using two methods. In the first method, photographs were taken by a fish-eye camera at 13 measurement points on the surface. The calculations were made using the RayMan model, and referred to the real SVF. The second method involved importing LIDAR DSM data into the SkyHelios model, to identify all obstacles and obtain relevant altitude-related information. The SVF image and values at the same 13 points were then simulated, and referred to as the LIDAR SVF. These two methods were adopted together to verify the accuracy of LIDAR data for elucidating the shading from trees or buildings. The SVF comparison method increases the availability and credibility of LIDAR in characterizing an urban built environment. Some surface measurements could be substituted in future research as observations of large urban areas become easier to obtain.

2.5. Correction of land surface temperature Table 1 Surface survey measurement point data. Measure point

Albedo (%)

SVF

LST (°C)

Ta (°C)

RH (%)

Tg (°C)

A_metal A_brick B_brick B_grass B_asphalt G_grass G_concrete G_brick F_PU F_grass F_concrete

0.4 0.3 0.3 0.26 0.05 0.26 0.1 0.3 0.3 0.26 0.1

0.485 0.485 0.477 0.313 0.512 0.332 0.467 0.386 0.680 0.747 0.715

31.8 32.7 32.5 28.7 31.9 37.4 39.9 39.7 39.9 30.5 39.2

31.3 31.3 31.2 31.2 31.2 35.2 35.2 35.2 32.3 32.3 34

65.6 65.6 65.3 65.3 65.3 55.8 55.8 55.8 63.3 63.3 57.8

32.7 31.8 32.5 28.7 31.9 37.4 39.9 39.7 39.9 30.5 39.2

As the reflected infrared energy declines across the atmosphere, the LST obtained from airborne thermal images falls below that obtained from surface measurements made using a thermocouple. Therefore, the LST values need to be corrected. Since this research concerns the effects of radiation, the shading factor becomes significant in LST calibration. In this study, DSM data based on LIDAR, local coordinates, date and time were imported into the SkyHelios model to simulate the shade state positions when the aerial photographs were taken (Matzarakis, 2012). A map was then generated to show the shaded and nonshaded areas in different colors. Since the aerial photographs were taken in the early morning, and Banqiao has many large buildings, the shaded area exceeded the non-shaded area in the map.

Please cite this article as: Chen, Y.-C., et al., Modeling of mean radiant temperature based on comparison of airborne remote sensing data with surface measured data, Atmos. Res. (2016), http://dx.doi.org/10.1016/j.atmosres.2016.01.004

4

Y.-C. Chen et al. / Atmospheric Research xxx (2016) xxx–xxx

Therefore, the map was divided into shaded and non-shaded parts based on the shading factor. The 13 measurement points were classified accordingly using the georeference function in GIS. Ten points were assigned to the shaded area, and three points were placed in the nonshaded area. Regression equations based on airborne TIR and surface measurements were used to correct the LST. The regression equations improve the accuracy of the LST data obtained from the thermal image, so that they better reflect the measurements made using the thermocouple. This correction method considers the effect of radiation on the surface. In a correction method with only one equation, extremely high values at points in the non-shaded area influence the correlation coefficient, making the measurement results inaccurate. Therefore, since Tmrt is strongly affected by the shading factor, two regression equations should be applied separately at points in the shaded and non-shaded regions. 2.6. Mean radiant temperature and thermal comfort A single physical environmental parameter, such as Ta, cannot fully reflect the thermal conditions, including the effects of long-wave and short-wave radiation. Therefore, an outdoor thermal environment should be described using an easy-to-understand index that integrates factors that relate to thermal comfort (VDI, 1998). To clarify the relationship between radiation and the thermal environment, the outdoor environment must be elucidated from the perspective of humanbiometeorology. Previous studies have adopted a variety of indicators that integrate physical environmental parameters and thermal balance on human to assess outdoor thermal comfort, such as the predicted mean vote (PMV), standard effective temperature (SET *) and physiologically equivalent temperature (PET). This study used the mean radiant temperature (Tmrt) owing to its strong positive correlation with the effects of radiation on the outdoor thermal environment. Tmrt, measured in units of °C, is the mean temperature that is felt by the human body, as determined by radiation and wind speed. The Tmrt is a thermal comfort index that incorporates all short- and long-wave radiation fluxes to which the human body is exposed, and is one of the most important meteorological parameters for human energy balance and thermal comfort (Thorsson et al., 2007). The Tmrt is defined as the uniform temperature of an imaginary enclosure in which the radiant heat transfer from the human body equals the radiant heat transfer in the actual non-uniform enclosure (ASHRAE, 2001). This index allows comparison between the experiences of a normal person outdoors and indoors, and can be easily applied to capture the state of an outdoor complex environment. The index enables urban planners who are not familiar with biometeorology to better understand its implications. Therefore, this study adopted Tmrt to assess the thermal comfort of the outdoor environment. 2.7. Mean radiant temperature estimation This work primarily concerns the use of remotely sensed data, airborne LIDAR DSM and thermal images, to estimate Tmrt. Six methods were used in various combinations to estimate Tmrt using the climatology calculating model RayMan and SkyHelios. The estimates of Tmrt were then analyzed to identify the combination methods whose results most closely matched the surface measurements of Tmrt. Method 2 was used as the standard for estimating Tmrt from surface measurements, and its accuracy was compared with that of other combined methods to demonstrate their feasibility of the application. First, the air temperature (Tam) and land surface temperature (LSTm) were measured by using a thermocouple. Globe temperature (Tg), relative humidity (RHm) and sky view factor (SVFm) were obtained from fisheye photographs. Second, air temperature (Taw), relative humidity (RHw) and global solar radiation (G) data were measured at a weather station. Third, the sky view factor (SVFa) was obtained from airborne

equipment, calculated by the LIDAR DSM in SkyHelios model, and land surface temperature (LSTa) was obtained from TIR images. The six methods that were used to estimate Tmrt are described below (Fig. 2). Method 1 uses measurements of Tam, RHm and Tg, which were imported into an ISO formula to estimate Tmrt1. Method 2, in which measurements and weather station data for Tam, RHm, LSTm, SVFm, and G, are imported into a RayMan model, was used to estimate Tmrt2. Method 3, in which measurement data from the weather station and airborne data on Tam, RHm, LSTm, SVFa, and G, are imported into the RayMan model, was used to estimate Tmrt3. Method 4, in which data on Taw, RHw, G and LSTa were extracted by thermal images based on data from the weather station, then imported into the RayMan model, was used to estimate Tmrt4. Method 5, in which data from the weather station and airborne data on Taw, RHw, G, and SVFa are imported into the SkyHelios model, was used to estimate Tmrt5. Finally, Method 6, in which weather station data and airborne data on Taw, RHw, G, LSTa and SVFa, are imported into the RayMan model, was used to estimate Tmrt6. Method 1 uses the empirical formula of Ashrae 55, which does not consider the shielding factor. Methods 2 and 3 differ in the data source for SVF. Both of these methods use meteorological measurements to compare SVFm and SVFa to improve the accuracy of the obtained SVF from LIDAR. The meteorological data in Methods 4, 5 and 6 are all from the weather station and different combinations of airborne data, which are compared with surface measurements to find the best way to estimate Tmrt without physical contact with the surface. Method 4 uses LSTm to analyze the impact of LST, and to determine whether only LSTm and data from the weather station are good enough to estimate Tmrt accurately. Method 5 uses SVFa to analyze the impact of the shading factor on Tmrt and to determine whether LST is important in estimating Tmrt. Method 6 uses both the LSTa and the SVFa to estimate the Tmrt, to determine whether complete information increases accuracy. 3. Results 3.1. Initial output and correction from airborne data The LIDAR DSM and TIR images, based on airborne data in this study, include two kinds of high-resolution information. To increase the accuracy of these data in terms of the actual urban thermal environment of Banqiao, the initial outputs were corrected comparing the SVF values, and calibrated them using surface measurements. After these correction and calibration processes, the airborne data could be adopted to estimate Tmrt in Banqiao. The calibration results include the SVF comparison and the LST correction map. In this study, airborne and surface measurements were compared to establish the reasonableness of using airborne information to elucidate an urban thermal environment. Highly flexible aerial photography enables information about human biometeorological conditions information to be obtained immediately, so that the serious phenomenon of increasing temperature in an urban environment can be controlled. 3.1.1. Sky view factor comparison Chapter 2.4 presents the method of calculating the sky view factor. The SVF obtained from the photographs taken using the fish-eye camera at the 13 measurement points was compared with the SVF simulated by importing LIDAR DSM into the SkyHelios model, demonstrating that LIDAR can accurately describe the shelter status of an urban environment. The differences of result between the two methods varied from 0.3% to 17%, and the average difference was approximately 5.1%, indicating that the results were very close to each other. The differences in the results were caused mainly by the limitations of the LIDAR DSM data, which refers only to the objects at the highest altitude at a resolution of 1 m. Therefore, LIDAR DSM did not properly

Please cite this article as: Chen, Y.-C., et al., Modeling of mean radiant temperature based on comparison of airborne remote sensing data with surface measured data, Atmos. Res. (2016), http://dx.doi.org/10.1016/j.atmosres.2016.01.004

Y.-C. Chen et al. / Atmospheric Research xxx (2016) xxx–xxx

5

Fig 2. Various Tmrt estimation methods.

consider the porous nature of the tree crown and the shaded space under the trees, but the difference between SVFm and SVFa remains small, such as at points C and E. However, the SVF determined only by buildings was completely accurate, such as at points F and H in Table 2. An SVF distribution map of the demonstration area was generated to show the shielding (Fig. 3). The SVF was lower at points in the

Table 2 Comparison of SVF by fisheye photo and LIDAR.

area that are surrounded by the buildings or in regions with dense planting, such as parks, than in open space, such as on roads or in squares. This result indicates that the LIDAR DSM data can be used in research because they capture the urban environment with high accuracy.

3.1.2. Land surface temperature map Chapter 2.5 describes the LST method for correcting airborne thermal images. LIDAR DSM data were imported into SkyHelios model to obtain the shading at 2014/9/14 8:45 AM. Each of the 13 surface measurement points was identified by GPS positioning to be in a shaded or a non-shaded area. The shaded area has ten points, and the non-shaded area comprises three points. Calibration using two corresponding shading correction equations makes the surface temperatures obtained from thermal images closer to those measured using a thermocouple in the surface measurement. The calibration can consider the association difference between obtained LSTs, and the irradiation status of the locations. Therefore the result can be used extensively for locations other than the 13 measurement points. Accordingly, the effect of the constructed environment on the urban thermal environment and the impact of the shading effect on the LST can be determined. The analysis of the surface temperature distribution map of the demonstration area revealed many buildings with many floors. The high angle of the sun in the morning yielded a large shaded area in the model. In the original, uncorrected LST map obtained using the TIR imager, LST values of 20–25 °C were found to the northwest of the buildings and vegetation. Non-shaded areas reached 30–35 °C. The corrected LST map after the LST was calibrated by applying the correction equation for shading revealed LST values in the shaded area of 30–35 °C, while the non-shaded area reached 45–50 °C (Fig. 4). The difference between the LST in the original map (Fig. 4a) and in the corrected map (Fig. 4b) was as high as 10–15° in non-shaded areas. This analytical result indicates that airborne thermal imaging significantly underestimates the surface temperature, especially in nonshaded areas, revealing the importance of correcting the LST using TIR images, and the necessity of considering shade in urban thermal environments.

Please cite this article as: Chen, Y.-C., et al., Modeling of mean radiant temperature based on comparison of airborne remote sensing data with surface measured data, Atmos. Res. (2016), http://dx.doi.org/10.1016/j.atmosres.2016.01.004

6

Y.-C. Chen et al. / Atmospheric Research xxx (2016) xxx–xxx

Fig 3. SVF distribution map in demonstrated area.

3.2. Mapping of Tmrt in demonstration area 3.2.1. Correlations among Tmrt values obtained using six estimation methods The feasibility of assessing human biometeorological conditions by combining airborne LIDAR and thermal imaging was investigated. The Tmrt index considers Ta, RH, G and LST, and thus yields more convincing results than only applying Ta to describe thermal environment. Therefore, Table 3 compares the accuracy of the six estimation methods discussed in Chapter 2.7. Method 2 was applied at 13 surface measurement points in the demonstration area, which are associated with five varieties of pavement and various sky view factors, to yield the actual value of Tmrt. The correlations between Tmrt2 and the other five results of Tmrt obtained using the other estimation methods were analyzed. The computed Tmrt values closest to the actual Tmrt were those of Tmrt6 estimated using Method 6, which combines LIDAR and TIR imaging and weather data from Banqiao weather station. The correlation coefficient was 0.653. Method 6 is the most accurate method in capturing the effect of radiation because it considers both the LST and the shading factor. Method 6 uses the most complete information related to Tmrt estimation, indicating that the combination of these two remote sensing methods effectively elucidates the urban thermal environment. The second most accurate method is Method 4, which uses only the LST data and climate data from Banqiao weather station. Method 4 does

not consider the shading factor, because it does not reduce the radiation at the points in the shaded area owing to the scarcity of SVF data from LIDAR. All points within the demonstration area are regarded as being under the same shielding conditions, regardless of shading. Hence, the obtained Tmrt in the shaded area under direct sunlight was too high, and the correlation coefficient is only 0.601. Method 5 applies only LIDAR SVF data and climate data from Banqiao weather station. No LST information is used in the estimation. The method does not take into account the radiant heat from the surface, and considers the entire region as having the same surface material, so its Tmrt does not reflect the various surface types. Therefore, the correlation coefficient is only 0.594. Method 1 is based on the ISO formula of Ashrae 55, and it considers Ta, Tg and wind speed. The correlation is high because the method draws on many historic data, and the correlation coefficient is 0.564. However, this method of estimating Tmrt can be only utilized at points where actual surface measurements were made. Accordingly, the method cannot be used to obtain the Tmrt values throughout the research measurement point. The main difference between Methods 2 and 3 is that Method 2 uses the SVF obtained from fish-eye photographs to estimate Tmrt, while Method 3 adopts the SVF obtained from the LIDAR DSM. Since LIDAR cannot accurately capture the details of the vegetation, such as the tree crown and trunks, the shielding is incorrectly assessed at many points. Therefore, Method 3 miscalculates Tmrt and has a correlations coefficient of 0.718. Method 3 thus has a higher coefficient than the other methods, but it cannot calculate Tmrt in the points without surface measurement, and so cannot be applied in a wide range of areas. These analytical results indicate that Method 6, which combines LIDAR and TIR, is the most accurate method for estimating Tmrt. The Tmrt estimated by Method 6 is closest to that obtained from the surface measurements by Method 2, as it incorporates the effect of radiation on the temperature. The shading factor has a stronger impact on Tmrt than the LST factor. The heat associated with LST is from the reflection of radiation, which is weaker than direct radiation from the sun. Therefore, the effect of the LST on the estimate of Tmrt is lower than the shading factor. 3.2.2. Tmrt distribution map This study indicates that Method 6, which combines airborne LIDAR and TIR, is the most accurate way to estimate Tmrt. Therefore, the estimates from Method 6 were used to generate the Tmrt distribution map in the demonstration area (Fig. 5). The Tmrt distribution map is based on Method 5, and obtained using SkyHelios, shows that the Tmrt reached 48–50 °C in the shaded areas, and 51–52 °C in the non-shaded areas (Fig. 5a). The Tmrt distribution map obtained from Method 6 and using RayMan shows that Tmrt was

Fig 4. Land surface temperature map in demonstrated area of original (a) and correction (b).

Please cite this article as: Chen, Y.-C., et al., Modeling of mean radiant temperature based on comparison of airborne remote sensing data with surface measured data, Atmos. Res. (2016), http://dx.doi.org/10.1016/j.atmosres.2016.01.004

Y.-C. Chen et al. / Atmospheric Research xxx (2016) xxx–xxx

7

4.1. Three-dimensional model of Tmrt

Table 3 MRT estimate method correlation coefficient comparison. Y X

The estimate equation for Tmrt 2 by various methods

R square

Tmrt1 (estimated by Method 1) Method 3 (estimated by Method 3) Method 4 (estimated by Method 4) Method 5 (estimated by Method 5) Method 6 (estimated by Method 6)

Tmrt 2 = 0.34 ∗ Tmrt 1 + 30.9

0.56

Tmrt 2 = 0.77 ∗ Tmrt 3 + 10.0

0.71

Tmrt 2 = 1.27 ∗ Tmrt 4 − 4.8

0.60

Tmrt 2 = 0.48 ∗ Tmrt 5 + 27.1

0.59

Tmrt 2 = 0.52 ∗ Tmrt 6 + 25.9

0.65

31–34 °C in the shaded areas, and is 49–52 °C in the non-shaded areas (Fig. 5b), while the Ta recorded by Banqiao weather station was 30.6 °C, and G = 97 W/m2. Method 6 provides the most detailed information about the effect of pavement material on the LST. To the northwest of buildings and vegetation, shading reduces radiation from the sun. Regions with natural coverage, such as parks, and those that contain many shaded areas, have a lower Tmrt than others. Areas like plazas and roads, which are covered by artificial pavement and are non-shaded, have a higher Tmrt. The largest difference in Tmrt was approximately 25 °C according to Method 6, and 4 °C according to Method 5. This result confirms that the LIDAR precisely captures the real conditions of an urban environment, and that the TIR image after correction provides accurate information that increases the accuracy of the estimated Tmrt.

4. Discussion This study identified hot spots to elucidate human biometeorological conditions. The results obtained by the methods in Chapter 3.2.2 are presented in a two-dimensional plane map, in which buildings are displayed only as black blocks. The altitude data are not presented. A three-dimensional model is required to represent both Tmrt and the buildings information in the urban environment simultaneously. Tmrt is not sufficiently complete as an index of human biometeorological conditions to describe the human thermal comfort balance, so a commonly used thermal comfort index, PET, was compared with Tmrt to elucidate the thermal comfort range more accurately in Taiwan.

Many investigations of mapping involve two-dimensional maps (O'Donohue et al., 2008), which contain no elevation information. Tmrt is strongly correlated with the shade provided by high buildings and vegetation. A two-dimensional map is less useful than a threedimensional model in elucidating shading, as it cannot represent the heights of buildings and vegetation. The three-dimensional Tmrt distribution model in this research comprises two layers. One layer represents the basic Tmrt data that are estimated by Method 6, and the other represents base altitude data from the LIDAR DSM. This three-dimensional model represents both the built environment and the Tmrt distribution, and can even show building façades for ease of visualization and elucidation of the spatial characteristics of hot spots (Fig. 6). Such a model helps to elucidate the effect of shading on Tmrt. Most urban building data in the past have had several shortcomings, such as the use of old data that fail to represent the current urban built environment, and poor accuracy, associated with low image resolution. LIDAR is a highly precise and highly flexible technology, which can be used to make a wide range of measurements, and to generate threedimensional models. The combination of the LIDAR DSM and Tmrt data to generate a three-dimensional Tmrt distribution model helps urban planners to identify hot spots in urban areas and elucidate variations of temperature with sun direction and shading factor.

4.2. Comparison of thermal comfort indices This study considered the effects of shading and surface temperature on radiant temperature, and adopted Tmrt as an index of the heating effects of radiation. However, to improve the effectiveness of Tmrt in explaining human biometeorological conditions, PET, which is a national standard index in Germany and has been widely used in multiple outdoor thermal comfort studies, was also compared with Tmrt. PET is a thermal comfort index of the heat balance of the human body in a complex outdoor environment; in a typical indoor setting (Ta = Tmrt, VP = 12 hPa, v = 0.1 m/s), the heat budget of the human body is equilibrated to the same core and skin temperature as in complex outdoor conditions. Thus, PET enables comparison of the effects of complex thermal conditions outside with experiences indoors (Höppe, 1999; Mayer & Höppe, 1987). As PET integrates Ta, RH, wind speed, G and physical conditions by human, it is commonly used to assess the thermal stress caused by high temperatures. In this study, the PET was obtained from Tmrt to

Fig 5. Tmrt distribution map in demonstrated area of Method 5 (a) and Method 6 (b).

Please cite this article as: Chen, Y.-C., et al., Modeling of mean radiant temperature based on comparison of airborne remote sensing data with surface measured data, Atmos. Res. (2016), http://dx.doi.org/10.1016/j.atmosres.2016.01.004

8

Y.-C. Chen et al. / Atmospheric Research xxx (2016) xxx–xxx

5. Conclusions

Fig 6. Three dimensional Tmrt map in demonstrated area.

represent thermal stress in the current environment. The Tmrt values in the study area were classified based on previous research on outdoor thermal comfort in Taiwan (Lin & Matzarakis, 2008). The proposed Tmrt assessment method can be applied more widely and the hot spots in Banqiao could be identified more accurately when the Tmrt is compared with PET. The RayMan model was used to calculate the PET from the Ta, RH and wind speed data from the Banqiao weather station, all of which were also utilized to assess Tmrt by Methods 4, 5 and 6. Method 6 yielded the most accurate Tmrt value. The thermal comfort range in Table 4 was obtained from previous research on different perceptions of PET in Taiwan, so a thermally comfortable definition of Tmrt can be obtained. Analytical results indicate that the lowest Tmrt obtained by Method 6 was 30 °C, which equals a PET of 31.2 °C, revealing that residents in Taiwan perceive their environment to be slightly hot, which condition is accepted as comfortable. The highest Tmrt was 56 °C, which equals a PET of 43.4 °C, which is very hot and falls within the range of high thermal stress. The average Tmrt was around 50 °C, which equals a PET of 40 °C, which is in the thermal range classified as moderate thermal stress. Since previous research demonstrates that the thermal discomfort threshold for Taiwan is a PET of 35 °C, in summer mornings areas other than shaded areas exceed the threshold for thermal discomfort, and even reach a condition of high thermal stress. The thermal stress condition of Banqiao is in extreme discomfort, and may affect the efficiency of work, and various health, environmental, societal and economic factors (Papanastasiou et al., 2015). Discussions of issues associated with urban thermal environments require elaboration of thermal comfort with reference to integrated meteorological parameters. Such an approach can improve understanding of the relationship between thermal environment and urban development, which includes height of buildings, surface coverings and planting. Therefore, urban thermal stress can be reduced.

The rising global temperature is strongly affecting urban thermal environments and outdoor thermal comfort. Although existing satellite telemetry methods conveniently reveal geographical information, they cannot accurately estimate detailed distributions of radiation-related properties in a complex urban environment in three-dimensions, owing to the low resolution and lack of vertical information that is provided by such methods. Therefore, satellite telemetry methods cannot provide sufficient information about biometeorological conditions in urban areas. This study utilizes an innovative method to observe the urban thermal environment; it couples airborne LIDAR with TIR and incorporates synchronous climate measurement at ground level. The airborne survey adopts a high-resolution LIDAR with a resolution of 1 m × 1 m to image the terrain and surface obstacles. This LIDAR can also identify building heights, forms of vegetation and the height of planimetric. The TIR can be used to estimate surface temperatures and levels of radiation emitted from horizontal and vertical surfaces. In the surface measurements, Ta, RH and LST are all measured simultaneously in the surveyed areas. The weather information is recorded by Banqiao weather station. Combining these three information sources enables the Tmrt values to be estimated using various methods, and then compared with each other to find the most accurate estimate method. Analytical results show that Tmrt estimated using LIDAR and TIR matches closely the measurements made on the surface. The Tmrt is calculated and represented as a distribution map and model, and it is compared to PET. The hotspots in the survey area, which has high-density buildings and generates a large amount of anthropogenic heat, can be identified using the two-dimensional map and the three-dimensional model, indicating that in Banqiao, the human biometeorological conditions are poor, even reaching conditions of high thermal stress. The analytical results demonstrate that the use of coupled LIDAR and TIR rapidly provides an accurate understanding of urban human biometeorological conditions. Therefore, this quickly measured and calculated method can in the future be used for regional thermal environmental assessment in other cities that are threatened by urban heat island or thermal stress. The current study is also being performed in Taiwan's third metropolitan city, Taichung, in order to combine the estimation method in this study with the thermal perception of local people there. An accurate thermal environment estimate model is developed through consideration of a variety of meteorological parameters, human thermal preference and the use of different indicators. Acknowledgments The authors would like to thank the Ministry of Science and Technology, Taiwan, for financially supporting this research under Contract No. MOST 103-2633-E-006-001-MY2 and Project-Based Personnel Exchange Programme between the NSC and DAAD. References

Table 4 Thermal comfort range in Taiwan. Ta (°C)

RH (%)

WS (m/s)

MRT (°C)

PMV

PET (°C)

SET (°C)

Comfort range in Taiwan

31.8 31.8 31.8 31.8 31.8 31.8 31.8

61 61 61 61 61 61 61

0.5 0.5 0.5 0.5 0.5 0.5 0.5

30 35 40 45 50 55 60

2.2 2.5 2.9 3.3 3.6 4 4.5

31.2 33.4 35.8 38.2 40.7 43.4 46.1

26 27.9 29.8 31.7 33.6 35.6 37.7

Slightly warm Slightly warm Warm Hot Hot Very hot Very hot

Aniello, C., Morgan, K., Busbey, A., Newland, L., 1995. Mapping micro-urban heat islands using LANDSAT TM and a GIS. Comput. Geosci. 21 (965–967), 969. ASHRAE, 2001. ASHRAE fundamentals handbook 2001, SI ed. American Society of Heating, Refrigerating, and Air-Conditioning Engineers. GA, Atlanta. Bowen, Z.H., Waltermire, R.G., 2002. Evaluation of light detection and ranging (LIDAR) for measuring river corridor topography. J. Am. Water Resour. Assoc. 38, 33–41. Chen, L., Ng, E., An, X.P., Ren, C., He, J., Lee, M., Wang, U., 2012. Sky view factor analysis of street canyons and its implications for intra-urban air temperature differentials in high-rise, high-density urban areas of Hong Kong: a GIS-based simulation approach. Int. J. Climatol. 32, 121–136. Cook, B.D., Corp, L.A., Nelson, R.F., Middleton, E.M., Morton, D.C., McCorkel, J.T., Masek, J.G., Ranson, K.J., Ly, V., Montesano, P.M., 2013. NASA Goddard's LiDAR, hyperspectral and thermal (G-LiHT) airborne imager. Remote Sens. 5, 4045–4066. Dozier, J., 1981. A method for satellite identification of surface temperature fields of subpixel resolution. Remote Sens. Environ. 11, 221–229.

Please cite this article as: Chen, Y.-C., et al., Modeling of mean radiant temperature based on comparison of airborne remote sensing data with surface measured data, Atmos. Res. (2016), http://dx.doi.org/10.1016/j.atmosres.2016.01.004

Y.-C. Chen et al. / Atmospheric Research xxx (2016) xxx–xxx Höppe, P., 1999. The physiological equivalent temperature — a universal index for the biometeorological assessment of the thermal environment. Int. J. Biometeorol. 43, 71–75. Li, F., Jackson, T.J., Kustas, W.P., Schmugge, T.J., French, A.N., Cosh, M.H., Bindlish, R., 2004. Deriving land surface temperature from Landsat 5 and 7 during SMEX02/SMACEX. Remote Sens. Environ. 92, 521–534. 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., Matzarakis, A., 2008. Tourism climate and thermal comfort in sun moon lake, Taiwan. Int. J. Biometeorol. 52, 281–290. Lin, T.P., de Dear, R., Hwang, R.L., 2011. Effect of thermal adaptation on seasonal outdoor thermal comfort. Int. J. Climatol. 31, 302–312. 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. Matzarakis, A., 2012. Linking urban micro scale models — the models RayMan and SkyHelios. 8th International Conference on Urban Climates, Dublin, Ireland. Matzarakis, A., Matuschek, O., 2011. Sky view factor as a parameter in applied climatology — rapid estimation by the SkyHelios model. Meteorol. Z. 21, 1–7. Matzarakis, A., Rutz, F., Mayer, H., 2007. Modelling radiation fluxes in simple and complex environments — application of the RayMan model. Int. J. Biometeorol. 51, 323–334. 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. Matzarakis, A., Rutz, F., Mayer, H., 2010. Modelling radiation fluxes in simple and complex environments: basics of the RayMan model. Int. J. Biometeorol. 54, 131–139. Mayer, H., Höppe, P.R., 1987. Thermal comfort of man in different urban environments. Theor. Appl. Climatol. 38, 43–49. Neale, C.M.U., Sivarajan, S., Akasheh, O.Z., Jaworowski, C., Heasler, H., 2009. Monitoring geothermal activity in Yellowstone National Park using airborne thermal infrared remote sensing. P. Soc. Photo-Opt. Ins. 7472, 747210.

9

O'Donohue, D., Mills, S., Bartie, S., Park, P., David, A., 2008. Combined thermal–LiDAR imagery for urban mapping. 23rd International Conference on Image and Vision Computing New Zealand. New Zealand, Christchurch. Oke, T.R., 1981. Canyon geometry and the nocturnal urban heat island: comparison of scale model and field observations. J. Climatol. 1, 237–254. Oke, T.R., 1987. Boundary Layer Climates. second ed. Routledge, London. Papanastasiou, D.K., Melas, D., Kambezidis, H.D., 2015. Air quality and thermal comfort levels under extreme hot weather. Atmos. Res. 152, 4–13. Price, J., 1983. Estimating surface temperatures from satellite thermal infrared data: a simple formulation for the atmospheric effect. Remote Sens. Environ. 13, 353–361. Shan, J., Sampath, A., 2005. Urban DEM generation from raw Lidar data: a labeling algorithm and its performance. Int. J. Remote Sens. 71, 217–222. Sohn, G., Dowman, I., 2007. Data fusion of high-resolution satellite imagery and LIDAR data for automatic building extraction. ISPRS J. Photogramm. Remote Sens. 62, 43–63. Sun, D., Kafatos, M., 2007. Note on the NDVI-LST relationship and the use of temperaturerelated drought indices over North America. Geophys. Res. Lett. 34, L24406. Thorsson, S., Lindberg, F., Eliasson, I., Holmer, B., 2007. Different methods for estimating the mean radiant temperature in an outdoor urban setting. Int. J. Climatol. 27, 1983–1993. Vallet, J., 2008. High precision LiDAR mapping for complex mountain topography, In: Proceedings of the Mountain Mapping and Visualization. 6th ICA Mountain Cartography Workshop. Lenk im Simmental, Switzerland. VDI, 1998. Methods for the Human Biometeorological Evaluation of Climate and Air Quality for the Urban and Regional Planning. Part 1: Climate. VDI Guideline 3787. Part 2. Beuth, Berlin. Yuan, F., Bauer, M.E., 2007. Comparison of impervious surface area and normalized difference vegetation index as indicators of surface urban heat island effects in Landsat imagery. Remote Sens. Environ. 106, 375–386. Zhang, Z., Ji, M., Shu, J., Deng, Z., Wu, Y., 2008. Surface urban heat island in Shanghai, China: examining the relationship between land surface temperature and impervious surface fractions derived from Landsat ETM+ imagery. Int. Arch. Photogramm. Remote. Sens. Spat. Inf. Sci. 37, 601–606.

Please cite this article as: Chen, Y.-C., et al., Modeling of mean radiant temperature based on comparison of airborne remote sensing data with surface measured data, Atmos. Res. (2016), http://dx.doi.org/10.1016/j.atmosres.2016.01.004