Thermal mapping a technique for road climatological studies

Thermal mapping a technique for road climatological studies Meteorol. Appl. 6, 385 394 (1999) Torbjörn Gustavsson, Laboratory of Climatology, Department of Earth Sciences, University of Göteborg, Box 460, SE 405 30 Göteborg, Sweden The present study deals with the thermal mapping technique in a road climatological perspective. The temperature mapping technique is analysed in relation to instrument set-ups, data collecting and possible sources of errors in converting measured radiation to surface temperature values. Furthermore, analyses of thermal mapping recordings are discussed. Factors that give rise to a varying temperature pattern are presented (i.e. road-bed material, radiation and advection), and also how the thermal mapping technique can be used to determine the relative importance of these factors. The conclusion of the study is that the technique is very useful but it is very important to be aware of the limits and different source of errors involved. 1. Introduction Micro- and local climatological processes vary in relation to small-scale geographical features. Consequently measurements that aim to cover these variations must be properly and carefully designed. For local climatological studies it is common to use either a stationary network of field stations or vehicles equipped with temperature sensors. Mobile measurements have a long tradition; for example Schmidt (1927; 1930) and Peppler (1929) published some early work using mobile platforms to describe climate variations. Since then this approach has been used in a large number of studies concerned with topo-climatology as well as urban climatology (e.g. Huovila, 1964; Hocevar & Martsolf, 1971; Oke & Maxwell, 1975; Söderström & Magnusson, 1995). Also Lomas et al. (1969) discussed mobile surveys in agro- and topo-climatological studies and especially the possibility of using mobile measurements to describe spatial variations in temperature in areas close to roads. An advantage of using a mobile platform is that a large area can be covered in a relatively short period and that the same instruments are used. This results in a temperature pattern that can be analysed in relation to, for example, topographical factors whilst avoiding differences related to instrumental errors or calibration. Mobile temperature measurements, or thermal mapping, have been used in applied road climatological studies since the middle of the 1970s. Lindqvist (1976) published an early report where methods for detecting road sections with a high frequency of ice formation were discussed. Other early work within this area is that by Sugrue et al. (1983) and Thornes (1985). Thermal mapping has since been developed and is today a method which is usually used prior to an installation of a Road Weather Information System (RWIS). Thornes (1991) describes the thermal mapping technique in great detail and also gives examples of how the temperature data varies according to weather as well as other important factors. The thermal mapping technique has also been used in a number of studies where the influence of different topographical objects was studied (e.g. Bogren & Gustavsson, 1989, 1991). An important development of the traditional mobile measurements was the inclusion of surface temperature detectors which could be used for road climatological studies. By use of infrared detectors the important factors influencing the local risk of road icing could be measured. Traditionally thermal mapping is used as a method to detect locations which differ in temperature compared to neutral areas; for example these could be areas where no gathering of cold air takes place. These locations are further compared with field observations and analyses of topographical maps to identify the most suitable locations for field stations in an RWIS. The present study emanates from a long tradition in using the thermal mapping technique for climatological studies. The need for the evaluation of the technique, both with respect to instrumentation and ways of analysing the recordings, has grown during the last few years. It is very important that thermal mapping is performed in a correct way since the technique is very sensitive to errors. This study focuses on the measuring technique and how to analyse the temperature recordings in order to determine factors that control the variation in surface temperature. The results from this study will provide recommendations about how the thermal mapping technique could be developed in order to give high-quality data for analyses of road climate variations. 385

T Gustavsson 2. Data and analysis In Sweden thermal mapping has been used for more than 20 years as a tool for road climatological studies. All major roads have been thermally mapped during different weather conditions to analyse the influence of local topography and other factors that determine the road surface temperature (RST). Many roads have also been mapped on a number of occasions, with several years in between. These recordings have been used to analyse the influence of changes in road coatings and the clearance of forests close to the road etc. All the measurements have been saved on data disks, which constitute a very large data bank available for analyses of variations in road climate conditions. The examples of thermal mappings used in this study are carried out along road no. 47 between Falköping and the road crossing no. 47/no. 48, in the county of Skaraborg, in the south-western part of Sweden. This particular road has been chosen as a test road as it passes through a large number of different local climate environments. The road is furthermore covered with a uniform asphalt layer, which is essential for the interpretation of the RST. For the study of road-bed material a test road outside Luleå, in the northern part of Sweden, was used. A homogeneous asphalt layer also covers this road, but the construction material differs along the road. Each stretch composed of a different material from the standard type (gravel) is marked on the roadside, so making it possible to perform detailed analyses of the temperature variation. During thermal mapping the air temperature is measured at two levels: 0.3 m and 2.0 m above the road surface. The surface temperature is measured with a Heimann KT15 radiation thermometer, which has a spectral sensitivity in the interval 8 14 µm and a response time of 1 second (Heimann Instruction Manual, Heimann GMBH, Wiesbaden, Germany). Each parameter is logged every 10 m during the measurement and the driver can feed manual markings to the computer about factors that can help in the analysis of the recordings, such as change in surface conditions, screening by bridges etc. Before and after each measurement the prevailing weather conditions are noted together with information about geographical location, time and date of the measurement. The weather conditions for the examples used in this study are discussed in each section. However, situations have carefully been selected in order to find the most suitable conditions for each of the aspects that are examined. 3. Measuring technique Surface temperature is a very difficult parameter to measure. The surface is what one could characterise as the transition zone between two different media. Large temperature differences often occur in this zone owing 386 to the change from turbulent to laminar flows close to the surface, as well as other factors. One way of performing measurements of the surface temperature is to register the energy flux and relate that to the temperature by use of the Stefan Boltzmann law: I = ε σt 4 where I is the energy flux density, σ is the Stefan Boltzmann constant, ε is the emissivity and T is the temperature of the surface given in degrees kelvin. Infrared (IR) scanners are instruments which have sensors that are sensitive to a specific wavelength-band. The interval 8 12 µm is most commonly used since it corresponds to the atmospheric window (i.e. where the atmosphere is nearly transparent for energy flow). This is also the range which is of most interest in studies of applied climatology. Using the IR technique is not trouble-free. For example the emissivity factor can be especially difficult to handle, since a small variation can cause a large variation in temperature calculations. In Figure 1 the temperature variation is shown for a calculation using an emissivity of 0.95. A change from 0.95 to 0.85 gives a surface temperature difference of approximately 8 C, given that I = 300 W m 2. The emissivity of a surface varies with type of material, texture etc. Tabulated values (Marshall, 1981) give an emissivity for concrete of 0.92 0.94 and asphalt 0.967. Another parameter of special importance in road climatology is whether the surface is wet or dry, which will significantly affect the emissivity of the surface. Measurements using an IR sensor have shown that the RST can differ by approximately 0.8 C between a dry and a moist asphalt surface. Therefore it is of great importance to carry out measurements during situations with constant surface status, either dry or moist. It is very difficult to interpret temperature recordings produced during situations with alternating surface status; therefore these situations should be avoided. The radiation received by the IR sensor comes mainly from three different sources, and the total radiance (I tot ) can be calculated using the following equation: I tot = I t-object τε+ I t-amb τ (1 ε) + I t-atmos τ (1 ε) where I t-object is the radiation from the road surface, I t-amb is the reflected radiation from surrounding objects, I t-atmos is the radiation from the atmosphere (controlled by the temperature of each object), τ is the transmission and ε is the emissivity. A problem to be overcome during thermal mapping measurements is to minimise the influence of the last two factors in the equation for I tot. One way of doing that is to have the scanner mounted in a nadir position and to install the sensor close to the surface. If the scanner

Using thermal mapping for road climatological studies Figure 1. Plot of the effect of variation in the emissivity for determining surface temperature. The surface temperature is expressed as a difference from a reference value (ε = 0.95). Points I V are values from different studies (see text for more information). is mounted with a view angle differing from nadir, a number of problems occur. First of all the size of the measuring area increases, which in a way can be an advantage, but if the oblique angle view is used, the importance of radiation from the surroundings and the atmosphere increases. Furthermore the emissivity changes with view angle and thereby the risk of misinterpretation increases. The intensity of the radiation also differs in relation to the view angle, and if an oblique set-up is used, a recalculation must be carried out in order to compare measurements using different set-ups. The marked points in Figure 1 refer to variations in emissivity owing to a varying view angle (I = 20, II = 40, III = 60 ). Data from Lagourade et al. (1995) have been used to give the relation between the apparent emissivity and the view angle. However, detailed studies performed over an asphalt surface have shown a much larger variation (Bergendahl, 1998). A view angle of 45 gives an apparent emissivity of 0.87 (IV in Figure 1), which has a pronounced effect on the calculated surface temperature, i.e. the RST will be 6 C lower compared with a nadir mounting. Scott (1986) has calculated that an increased view angle will result in a decrease in the apparent emissivity of 4% (V in Figure 1) using a view angle of 40. This will give a temperature difference of 3 C. Further studies are needed in this area to fully understand the effect from a varying view angle on the RST. In Figure 2, repeated temperature measurements along the test road (no. 47) are shown. Measurements were carried out four times during the same clear and calm night in order to follow the temperature development of the stretch of road and also to enable analyses of the quality of the instruments used. If the instrument is of high quality and the measurements are carried out in a proper way, the correlation between the individual measurements should be high, which is the case for the example shown in the figure (there is a correlation coefficient of 0.77 between runs 2 and 3). The magnitude of the temperature differences changes through the night, which will affect the correlation in a negative way. However, the temperature pattern stays the same through the night, and this is clearly seen in Figure 2. Other factors that can influence the quality of the measurements are whether the scanner is kept at constant temperature or not, whether the lens of the scanner is cleaned, and whether the instruments are calibrated. All these factors have a large influence on the accuracy of the measurements. 4. Analysis of the thermal mapping recordings Three main factors can be distinguished which determine the road surface temperature: (a) road-bed materials, (b) radiation, and (c) advection and stagnation of cold air. In the following sections the importance of these factors is discussed and examples are presented which show how the parameters can be studied by the 387

T Gustavsson I II III IV Figure 2. Recordings of surface temperature along road no. 47 on four different occasions during the same night (I = 20.00, II = 23.00, III = 01.00, IV = 03.00) for 27 March 1994. use of thermal mapping. The analysis of the influence on the RST of the three factors is carried out in such a way that the studied parameter should be the main factor determining the variation in temperature. The influence of road-bed material, for example, is studied along a stretch of road with very small topographical variations and open roadsides, so that no obstruction of the radiation balance occurs as a result of nearby obstacles. 4.1. Road-bed materials The heat flow and storage in the road bed is determined by the thermal properties of the material used in the road s construction. This factor influences the temperature reaction during cooling as well as during warming. Information about the type of material used is not always available, and therefore it is very important to perform the thermal mapping in such a way that this factor is covered. An example of the large influence of the construction material is presented in Figure 3. The measurement was carried out during a cloudy, windy night (4/5 December 1997) along the test road outside Luleå, previously described. The road passes through open arable land, and there are no large variations in altitude (<15 m). The temperature recording shown in Figure 3 is from one of the test sections, located between 800 and 940 m from the start point. This section has construction material consisting of iron-rich sand, a waste material from mining. In the sections adjacent to the test section the material is gravel. A homogenous asphalt layer covers the entire stretch. During the measurement the road surface was dry, but there was snow along the hard shoulders. It might be 388 expected that drifting snow on the road surface could have an effect. However, the influence of this factor is considered marginal because the temperature differences are consistent and well correlated with the variation in road-bed material. The examples discussed below are from a test stretch where the influence of different built-up materials is examined. It is not common to have such large variations along an ordinary road, but this can occasionally occur, for example, between gravel and solid rock. These two materials can alternate along roads which pass through rock cuttings where the in situ rock is used near the cuttings and gravel in between. In Figure 3 the air and surface temperatures from measurements carried out at 22.30 (local time) and the surface temperature from 07.00 are shown. The air temperature along this test section is constant ( 13 C), but the surface temperature shows a very large variation associated with the change in construction material, being approximately 2.5 C lower than that of the surrounding stretch at 22.30 and 2.0 C lower at 07.00. As the change is very distinctive and is repeated in a similar way between the two measurements, the material used in the road bed must cause the pattern. The air temperature showed no variation along the stretch on both occasions. This is a fact that also indicates that the material is the main factor causing this variation in surface temperature. 4.2. Radiation Incoming solar radiation and outgoing long-wave radiation from the surface are major influences on the RST.

Using thermal mapping for road climatological studies Figure 3. Variation in surface and air temperatures along a test road outside Luleå in the northern part of Sweden on 4/5 December 1997. Objects that screen off the sun influence the radiation balance both during the day and night (by obstructing the long-wave radiative cooling). That screening off the sun has a pronounced effect on the RST has been shown by, for example, Bogren (1991). Previous studies (e.g. Bärring et. al., 1985; Gustavsson, 1995) have documented that the sky view factor (SVF) has a strong correlation with the temperature, since this factor is a measure of the degree of sky reduction. The SVF can, for example, be determined from fish-eye photographs. However, this is a very time-consuming method, and information is only obtained from the restricted place where the photograph was taken. As the radiation condition is a very important parameter to have information about, there is a need for a method that can be used for entire stretches of road. A surface temperature recording from road no. 47 during late afternoon (18.00) just before sunset is shown in Figure 4(a). The weather situation during the day and the following night was a clear sky ( 1 oktas) with a weak wind (< 2 m s 1 ). From 16 km onwards the road passes through a forest, alternating between open and tree-covered sections. The orientation of the road is north north-east to south-south east, giving a pronounced screening effect during the morning and afternoon. Open and screened sections of the road can be distinguished by means of the temperatures: open areas have a surface temperature of approximately +10 C and the screened ones of approximately +4 C. This type of temperature recording gives information about the road s surroundings or more precisely about the obstruction of the radiation that occurs on that specific section of road. Analysis of daytime recordings was carried out in order to determine if this type of measurement gives enough information about the radiation conditions or SVF along a road stretch. The mean surface temperature was calculated for sections of the road showing a consistent temperature level. These mean temperatures were further compared with SVF photographs from selected places and scaled by use of the maximum and minimum temperatures and SVF values, respectively (see Figure 4(b)). The open area had SVF = 1.0 and the lowest SVF (0.60) was obtained at 21 km. By comparing these values with the air temperature pattern from a recording carried out during a clear night, it possible to verify if this technique is useful or not. In Figure 5 the sky view values are plotted against the difference in air temperature for a clear, calm night. The differences in air temperature are calculated using open, neutral areas as a reference level. The correlation between the two variables is high, showing that the sky obstruction is an important factor influencing the variation in temperature in forested areas. The first 16 km of the measured route in Figure 2 pass through open, undulating terrain. Analysis of the temperature pattern from these sections of road shows that the radiation plays an important role in the open terrain as well as in the forested part, as previously discussed. The surface temperature shows a very diversified pattern that cannot be explained by variation in the air temperature; rather the pattern is associated with the 389

T Gustavsson (a) (b) Figure 4. (a) A daytime recording at 18.00 on 27 March 1994 of the surface temperature in a forested part of the study area. (b) Calculation of the sky-view factor from the surface temperature recording presented in (a). Figure 5. Plot of calculated sky view values versus variation in air temperature at 20.00 on 27 March 1994. detailed orientation of the road. In valleys the bottom and the valley slope facing the east will receive less short-wave radiation during the afternoon and evening. And as shown in Figure 2, this causes large variation in surface temperature (3 C at sunset). 390 By comparing the surface temperature pattern from sunset onwards, it becomes possible to determine the period for which this variation in received radiation has an effect (see Figure 2). The repeated measurement along road no. 47 shows that temperature differences decrease in intensity. However, after 5 hours, the after-

Using thermal mapping for road climatological studies noon radiation is still an important factor influencing the surface temperature. This is very important to take into consideration if the thermal mapping is used to find the most suitable location for field stations in an RWIS. Mapping must be carried out during the winter period when the intensity of the incoming radiation is weak to be able to fully evaluate the influence of the local topography etc. Another aspect is that this factor must be taken into consideration since it controls the surface during autumn and spring. The selection of station sites and detailed documentation of the areas and of the reasons for choosing that specific site are very important aspects of the establishment of an RWIS. 4.3. Advection and stagnation of cold air Advection, especially cold air flows, and stagnation of cold air in valleys and low points can significantly influence the road surface temperature. If the air temperature is measured during thermal mapping, the influence of low temperatures in cold air pools and cold (a) hollows can be analysed. Studies by Gustavsson (1995) and Gustavsson et al. (1998) have shown that a very diversified air temperature pattern develops in undulating terrain shortly after sunset. It has also been shown that a significantly lower than average surface temperature can be found in valleys during clear nights (e.g. Bogren & Gustavsson, 1991). A temperature recording during a clear and calm situation along road no. 47 is shown in Figure 6. The recording was carried out during 15 16 February 1994 at 18.00 (i.e. approximately one hour after sunset). In the figure the temperatures are presented as deviations from the neutral temperature. The neutral temperature represents the temperature of areas where no gathering of cold air takes place. By use of this type of presentation the cold air pools can clearly be identified. Along this specific route, four different valleys could be identified where the accumulation of cold air occurs. These are located at 5 6.5 km, 10 11.3 km, 11.9 13 km and 14 14.8 km from the starting point. As shown in Figure 6 there is a close correspondence between the air tem- (b) Figure 6. Recordings of (a) air temperature and (b) surface temperature along the test road for 18.00 on 15/16 February 1994. The temperatures are expressed as deviations from a reference temperature. 391

T Gustavsson peratures and the surface temperature: the lowest temperature are found in valleys where stagnation as well as pooling of cold air leads to low air temperatures. A second measurement was carried out along the same route two hours later (20.00). By comparing these two measurements it is possible to draw conclusions about the temperature development and especially the effect that an accumulation of cold air will have on the surface temperature. In Table 1 the deviation from the neutral temperature for both the air and the surface temperature is shown. The air temperature deviation shows that intense cold air pools are established shortly after sunset. A measurement carried out just before sunset showed very small temperature variations along the stretch. In a study by Gustavsson et al. (1998) it was concluded that this could be explained by stabilisation of the surface-cooled air in sheltered locations such as valley bottoms and in forests which are not very dense. The difference between the two measurements (I, II) indicates that the intensity does not increase to any great extent; the effect of a longer cooling period is rather the extension of the cold air pools. The surface temperature shows quite a different development. The surface is more inert, and therefore a longer time period is needed in order to adjust to the low air temperature in valleys. This can be seen in Table 1 from the fact that the difference between the neutral temperature and the valley-surface temperature increases between the two measurements even though the air temperature does not show the same clear trend. By plotting the air and surface temperature deviation from the reference temperature for both measurements it possible to determine the relative importance of the advection-stabilisation of cold air. Thereby it is also possible to compare this factor with the other ones that can influence the surface temperature pattern. In Figure 7 the result of such a comparison is shown. A linear fit is calculated for both situations, and the square of the correlation coefficient (R 2 ) increases from 0.45 to 0.67; this indicates that the surface temperature is adjusting to the variations in air temperature along the route. 5. Conclusions paid to a number of factors during the measurement. This holds especially for measurements using IR sensors. Particularly important is the mounting of the IR sensor and awareness of the factors controlling the radiation received by the sensor as well as the quality of the type of sensor used. Analysis of the recordings in order to select the most suitable locations for field stations in an RWIS is also an important task. It is crucial that all parameters that control the road surface temperature are measured in order to interpret the recordings. In the present study it was concluded that three major factors could be distinguished which together control the variation in RST. (a) Road-bed material. Measurements along a test route showed that the variation in road construction material had a significant effect on the RST. The influence of this factor can be determined by use of detailed information about the materials used or from analyses of the temperature reaction during cooling or warming episodes. (b) Radiation. The obstruction of the radiation that occurs in forested or built-up areas can be determined from daytime measurements of the surface temperature and furthermore can be related to the night-time temperature differences. (c) Accumulation of cold air. Cold air accumulation in valleys and stabilisation in wind-sheltered locations will significantly influence the RST. The influence of this factor will increase during the night owing to the inertia of the road surface. Thermal mapping measurements should be planned and performed in such a way that the influence of the above-discussed factors is determined. It may also be necessary to study road construction plans regarding the variation in road-bed materials. The obstruction of the radiation can be determined either by direct measurements or in a general way by the method described in this paper. For the actual locations of field stations it is important to select different areas where these factors vary in importance and also to describe the station sites very carefully. Thermal mapping is a very useful technique in applied climatological studies. However, attention must be Table 1. Cold air pool intensity and lowering of surface temperature in the valleys compared with the general temperature. Measurement I was carried out at 18.00 and II at 20.00 on 15 February 1994. Valley number Measurement I Measurement II Air Surface Air Surface 1 10.1 3.7 10.1 4.2 2 6.7 2.5 8.1 3.1 3 7.0 2.7 7.6 3.0 4 8.5 2.3 7.3 2.8 392

(a) Using thermal mapping for road climatological studies (b) Figure 7. (a) Plot of lowering surface temperature owing to an accumulation of cold air in valleys along the test road, at 18.00 on 15 February 1994. (b) As (a) but at 20.00. Acknowledgement This study was supported by the Swedish National Road Administration. Comments on the draft of this paper by Dr Jörgen Bogren and Maria Karlsson, MSc are also greatly appreciated. References Bärring, L., Mattsson, J. O. & Lindqvist, S. (1985). Canyon geometry, street temperatures and urban heat island in Malmö, Sweden. J. Climatol., 5: 433 444. Bergendahl, S. (1998). A study of surface temperature variations with the use of IR-technique. Earth Sciences Centre: 393

T Gustavsson B143, University of Göteborg. (In Swedish with an English summary.) Bogren, J. & Gustavsson, T. (1989). Modelling of local climate for prediction of road slipperiness. Physical Geography, 10: 147 164. Bogren, J. & Gustavsson, T. (1991). Nocturnal air and road surface temperature variations in complex terrain. Int. J. Climatol., 11: 443 455. Eliasson, I. (1996). Urban nocturnal temperatures, street geometry and land use. Atmosph. Environ., 30: 379 392. Gustavsson, T. (1995). A study of air and road surface temperature variations during clear windy nights. Int. J. Climatol., 15: 919 932. Gustavsson, T., Karlsson, M., Bogren, J., & Lindqvist, S. (1998). Development of temperature patterns during nocturnal cooling. J. App. Meteorol., 37: 55 571. Hocevar, A. & Martsolf, J. D. (1971). Temperature distribution under radiation frost conditions in a central Pennsylvania valley. Agricult. Meteorol,. 8: 371 383. Huovila, S. (1964). On precautions against crop damage due to radiation frost within hilly regions. Helsinki: SSF XXIX, 22 pp. Lagourade, J. P., Kerr, Y. H. & Brunet, Y. (1995). An experimental study of angular effects on surface temperature for various plant canopies and bare soils. Agricult. Forest Meteorol. 77: 167 190. Lindqvist, S. (1976). Methods for detecting road sections with high frequency of ice formation. GUNI-Report 10, Department of Physical Geography, University of Göteborg. Lomas, J., Shashoua, Y. & Cohen, A. (1969). Mobile surveys in agrotopoclimatology. Meteorologische Rundschau, 22: 96 101. Marshall, S. J. (1981). We need to know more about infrared emissivity. SPIE, 313: 119-127. Oke, T. R. & Maxwell, G. B. (1975). Urban heat island dynamics in Montreal and Vancouver. Atmos. Environ., 9: 191 200. Peppler, A. (1929). Das Auto als Hilfsmittel der Meteorologischen Forschung. Das Wetter, 46: 305 308. Schmidt, W. (1927). Die Verteilung der Minimumtemperaturen in der Frostnacht des 12.5.1927 im Gemeindegebeit von Wien. Fortschritte der Landwirtschaft 2. H., 21 pp. Schmidt, W. (1930). Kleinklimatische Aufnahmen durch Temperaturfahren. Meteorologische Zeitschrift, 47: 92 106. Scott, J. R. (1986). Incorporation of angular emissivity effects in long wave infrared image models. SPIE, 687:.44 53. Söderström, M. & Magnusson, B. (1995). Assessment of local agroclimatological conditions a methodology. Agricult. Forest Meteorol., 72: 243 260. Sugrue, J. G., Thornes, J. E. & Osborne, R. D. (1983). Thermal mapping of road surface temperatures. Phys. Technol, 14: 212 213. Thornes, J. E. (1985). The prediction of ice on roads. J. Institution of Highways and Transportation, Aug/Sept. Thornes, J. E. (1991). Thermal mapping and road weather information systems for highway engineers. In Highway Meteorology, ed. by A. H. Perry, and L. J. Symons, E. & F. N. Spon, London. 394