Meteorol. Appl.

8, 95–105 (2001)

Prediction of hoar-frost by use of a Road

Weather information System
Maria Karlsson, Laboratory of Climatology, Physical Geography, Earth Sciences Centre,
Göteborg University, Box 460, 405 30 Göteborg, Sweden

This study measured the amount of hoar-frost on the road surface with the help of copper plates. The
amounts were compared to data collected at a station in the Road Weather information System (RWiS)
together with additional measurements of the temperature and humidity profile above the road surface.
The data were used to specify the weather parameters when hoar-frost forms on the road surface and
relate the measured amount of hoar-frost and resulting friction to measured weather parameters. The
aim was also to see whether there was an increased possibility of predicting the amount of hoar-frost
with additional measurements of the temperature and humidity profiles. The results of this study
confirm that the amount of hoar-frost deposited on the road surface increases with increasing average
wind speed, increasing maximum difference between dew point (Td) and road surface temperature (Tr)
and increased duration when Tr < Td, measured at the RWiS station. However, the difference between
Td and Tr failed to indicate hoar-frost on several occasions due to the location and poor accuracy of the
measuring equipment. Measurements of temperature and humidity in a profile up to 2.5 m above the
road surface showed that the air layer up to 0.4 m reflected conditions at the road surface whereas the
surrounding areas influenced the measurements at higher levels. A forward multiple regression
performed on the data showed that measurements of temperature and humidity at lower levels
improved predictions of deposition of hoar-frost on the road surface. Measurements of friction on the
road surface when hoar-frost was present showed that friction decreased with increasing difference
between Td and Tr and it also increased with the duration of Tr < Td. There was no correlation between
the amount of hoar-frost and friction at the road surface. The friction is probably influenced not only by
the amount of hoar-frost but also by the structure of the hoar-frost, which has not been determined in
this study.

1. Introduction vations to describe the distribution of rime in different

terrains.
Hoar-frost forming on the road surface is one type of
slipperiness that is hard to predict both in amount Factors important for the development of hoar-frost
deposited and in the resulting friction reduction. are discussed in several articles.
Theoretically hoar-frost is water vapour sublimating
on the road surface when the temperature is below zero (a) Wind speed, air temperature and humidity are con-
and the dew point temperature is higher than the road sidered the most important factors by several
surface temperature. Hoar-frost deposition on roads authors (e.g. Gocheva, 1990; Scherm & Van
was studied by Hewson & Gait (1992) who used phys- Bruggen, 1993; Stanev et al., 1987; Hewson & Gait,
ical reasoning and case studies to develop two methods 1992).
of predicting the formation of hoar-frost. Gustavsson (b) Scherm & Van Bruggen (1993) maintains that frac-
& Bogren (1990a) studied temperature differences tional cloud cover is important for the duration of
between road surface and dew point and their relation- hoar-frost events.
ship to hoar-frost occurring during the passage of a (c) Stanev et al. (1987) includes wind direction and
warm front. Hoar-frost also commonly forms on the Hewson & Gait (1992) the time of year as being
road surface when the road surface cools more quickly important.
than the air layer above. The amount of hoar-frost (d) Makkonen & Ahti (1995) and Stanev et al. (1987)
formed is usually higher during the former situation mention the importance that the duration of the
(Kromer, 1995). Takle (1990) used a survey of frost event has on the amount of hoar-frost.
occurrence from the highway maintenance people in (e) Takle (1990), Hewson & Gait (1992) and
order to describe the occasions when hoar-frost Gustavsson & Bogren (1990a) add that the preced-
formed. In a study of climatic mapping of ice loads, ing weather has an effect on the amount of hoar-
Makkonen & Ahti (1995) used airport weather obser- frost.
95
 14698080, 2001, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1017/S1350482701001086 by Czech Technical University, Wiley Online Library on [28/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
M Karlsson
(f) Gustavsson & Bogren (1990a) also proposed that They were taken at the Swedish National Road
the size of the temperature difference between road Administration test site at Road 45 near Göteborg. The
surface temperature (Tr) and dew point tempera- test site is situated alongside a section of a four-lane
ture (Td) could be proportional to the amount of highway at Surte and it comprises an asphalt plane
hoar-frost deposited. approximately 30 m long and 10 m wide in contact with
(g) Gustavsson (1991) studied the importance of local one lane of the highway. The site is equipped with a
climate variation for the duration and amount of monitoring system which measures temperature,
hoar-frost. humidity, wind speed and wind direction at several lev-
els. New measuring equipment is tested at the site and
The Road Weather information System (RWiS) in an original RWiS station is also present.
Sweden uses measurements of air temperature and
humidity at 2 m, wind speed at 5 m and road surface For this project data from the RWiS station were used
temperature to generate warnings of hoar-frost deposi- together with data from additional equipment measur-
tion on the road surface. Studies of temperature vari- ing temperature and humidity in a profile above the
ation in the air layer close to the road surface have shown road up to 2.5 m. The additional equipment took mea-
that large temperature differences in that layer occur surements every 10 seconds and stored them as an aver-
especially during weather situations when hoar-frost age in a logger every 10 minutes. Measuring levels and
usually develops (Chen et al., 1999; Karlsson, 1999). With instrument data are shown in Table 1. Before and after
large temperature and humidity differences in the air each measuring season the temperature and humidity
layer close to the road surface the warnings generated instruments were calibrated.
from measurements at 2 m can sometimes be misleading.
The RWiS data were used to analyse the relationship
The aim of this study is to analyse the importance of between parameters measured at the RWiS station and
temperature variations in the air layer close to the road the amount of hoar-frost deposited on the road surface.
surface for the formation of hoar-frost. The following The additional equipment, measuring temperature and
specific points are dealt with. humidity profiles, were used to analyse whether the
measuring of additional parameters at the RWiS station
(a) The amount of hoar-frost produced during differ- could improve the prediction of hoar-frost on the road
ent weather situations. surface.
(b) The weather situations when hoar-frost forms on
the road surface. To obtain an estimate of the amount of hoar-frost
(c) The correlation between amount of hoar-frost and deposited at the road surface, hoar-frost was measured
friction to the measured weather parameters. at three locations with the help of thin copper plates.
(d) The height of the air layer that reflects the The amounts from the three plates were then averaged.
processes close to the road surface. The plates measured 15 cm × 20 cm × 0.05 cm and were
(e) Improving hoar-frost prediction with additional painted black in order to produce a similar emissivity
measurements of temperature and humidity pro- to the road surface. The plates were weighed to calcu-
files. late the amount of hoar-frost deposited. The structure
of the hoar-frost and its distribution on the road
2. Study area and instrumentation surface were also visually observed on each occasion.

2.1. Measurements A portable skid resistance tester, developed at the

Swedish National Road and Traffic Research Institute
The measurements described in this paper were per- (Gustafsson, 1981), measured the friction at the road
formed during three winter seasons from 1996 to 1999. surface on some occasions when hoar-frost was pre-

Table 1. Instrument specifications for measuring equipment

Measurement Level (m) Time constant Accuracy Range Sampling Instrument

frequency
RWiS
Air temperature 2 – ± 0.3 °C –60 to +70 °C 30 min Lambrecht 8091
Humidity 2 – ± 3% 5 to 100% 30 min Lambrecht 8091
Wind speed 5 – ± 0.2 m s-1 0.4 to 75 m s-1 30 min Vaisala,WAA15
Wind direction 5 – ± 11.75° 0 to 360° 30 min Vaisala,WAV15
Road surface
temperature –0.002 – ± 0.3 °C –60 to +70 °C 30 min Pt 100
Road mast
Temperature 0.1, 0.4, 1.0, 1.5, 2.5 < 10 s ± 0.1 °C – 10 min Thermistor
Humidity 0.1, 0.4, 1.0, 2.5 15 s (90%) ± 0.5% 0 to 100% 10 min Vaisala, HMP 243

96
 14698080, 2001, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1017/S1350482701001086 by Czech Technical University, Wiley Online Library on [28/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Prediction of hoar-frost by use of a Road Weather information System
sent. The skid resistance is given as a friction coefficient 3. Results and discussion
between the frictional force and the normal force. The
friction coefficient is non-dimensional but due to the
3.1. Weather situations during formation of hoar-
low number of the coefficient it is usually given in µ
(10-6). frost
Hoar-frost was measured on 16 selected nights during
2.2. Measuring accuracy the study period. Earlier studies by Hewson & Gait
(1992) and Gustavsson & Bogren (1990a) showed that
Differences in emissivity and thermal properties hoar-frost especially tends to form during two weather
between the copper plates and the road surface result in situations. The most common is during clear nights
different cooling patterns which means that the when the surface cools below dew point. The other sit-
amounts of hoar-frost measured on the copper plates uation is when a warm front passes and low air tem-
cannot be directly used for estimations on the road sur- peratures are replaced with warm air by advection; the
face. Observations showed, however, that the distribu- road surface temperature then rises more slowly and
tion and structure of the hoar-frost on the copper stays below the dew point. Examples of the tempera-
plates and on the road surface were similar. Hoar-frost ture development and wind speed during these types of
was present on the asphalt surface on each occasion situations are shown in Figure 1. The examples are
when hoar-frost was measured on the copper plates. taken from two consecutive days where the first is a
The variation in amount of hoar-frost between the clear and calm night and the following night is clear fol-
plates was approximately ± 40 g m−2. lowed by warm air advection. During the first night,
26/27 January 1997, the wind speed varied between 1
Plates are commonly used in crop research to measure and 3 m s−1, the sky was clear and the road surface had
the amount of dew on the ground and at different cooled to below dew point (Figure 1(a)). During the
levels, see Jacobs et al. (1994). Kidron (1997) showed second night, 27/28 January 1997, the sky was also clear
that plates larger than 30 cm × 30 cm created their own and the wind speed was lower than 3 m s−1 but warm
microclimate resulting in larger amounts of dew and air advection increased the air temperature although the
plates smaller than 5 cm × 5 cm received smaller road surface had cooled below dew point (Figure 1(b)).
amounts due to the impact of soil thermal properties. On both these occasions large amounts of hoar-frost
The choice of plates measuring 15 cm × 20 cm in this were measured in the early morning hours.
study corresponded quite well to the amount of hoar-
frost formed on the road surface (visual estimate). The In this study, hoar-frost was measured on 13 occasions
plate was very thin (0.5 mm) and the high thermal when there was cooling of both the air and road surface
capacity of copper results in a quick adjustment to the temperatures, and on 3 occasions when warm air advec-
temperature of the underlying surface. The use of plates tion increased the air temperature whilst the road sur-
was a good alternative to other expensive measuring face temperature continued to cool for a few hours. The
devices available and it also had the advantage that weather conditions for these occasions are shown in
the amount of hoar-frost was confirmed with visual Table 2. The sky was clear with cloud cover less than 2
observations. oktas during some part of the night for all situations.
The intense cooling on clear calm nights increases the
To measure the relatively small differences in tempera- risk of hoar-frost and when this weather is followed by
ture between levels requires instruments which are well warm air advection the temperature differences between
calibrated with each other. The instruments used in this road and air increases thus increasing the risk of slipperi-
study had an accuracy of ±0.1 °C. A calibration per- ness (Gustavsson, 1991). The average wind speed was
formed in a climate chamber showed a maximum dif- below 3 m s−1 for all situations. The minimum road sur-
ference between instruments of 0.2 °C. The tempera- face temperature ranged from –2.2 to –12.2 °C and the
ture recordings at different levels showed nearly equal minimum air temperature varied from –1.2 to –14.3 °C.
temperature development during cloudy and windy sit-
uations. This indicates that the differences of up to The RWiS uses a simplified procedure to determine the
1.7 °C between different measuring levels in this study flow of water vapour to the road surface. One criteria
were accurate and not related to measuring errors. for the sublimation of hoar-frost is that the tempera-
ture difference between dew point (Td) and road sur-
The instrument used for measuring wind speed in this face temperature (Tr) is positive. When comparing the
study has a starting value of 0.4 m s−1. The wind speed recordings from RWiS and manual observations of
is measured every second and then averaged for each hoar-frost in Table 2, it can be seen that the maximum
10-minute period. At very low wind speeds this results difference between Td and Tr, with dew points calcu-
in the measured average wind speed possibly being lated from RWiS measurements at 2 m, is negative on
lower than the actual average wind speed. several occasions when hoar-frost is observed and mea-
sured on the road surface. A large humidity and tem-
perature difference in the air layer below the measuring

97
 14698080, 2001, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1017/S1350482701001086 by Czech Technical University, Wiley Online Library on [28/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
98

M Karlsson
Table 2. Weather description for situations when hoar-frost was measured, X indicates that the cloud cover where less than 2 oktas during some part of the
night, Y = hoar-frost indicated by RWiS, N= Hoar-frost not indicated by RWiS. All values consider the whole night except 27 January 1997b, that is a one-
hour value.

Level Cloud Average Min Tr Min Ta Max Max Temperature Hoar-frost RWiS Hewson & Gait Hewson & Gait
cover wind speed (°C) (°C) Td–Tr Ta–Tr difference (g m–2) Method 1 Method 2
<2/8 (m s–1) (°C) (°C) 2.5–0.1 m
(°C)
Cooling
24 Jan 1997 X 0.59 –5.1 –4.0 –0.6 1.1 1.3 – N 3 0
27 Jan 1997a X 1.66 –4.25 –2.2 2.1 2.3 0.9 181 Y 4 10.64
27 Jan 1997b X 2.73 –2.9 –1.2 0.8 1.5 0.3 24 – – –
29 Jan 1997 X 0.31 –3.3 –3.3 2.5 3.0 1.5 210 Y 6 5.50
30 Jan 1997 X 1.05 –2.3 –2.3 0.4 0.5 0.8 209 Y 5 0.26
31 Jan 1997 X 1.51 –4.7 –4.7 1.5 3.9 1.4 297 Y 4 4.71
27 Nov 1997 X Missing Missing Missing Missing Missing Missing 85 Missing Missing Missing
18 Dec 1997 X 0.90 –8.9 –8.0 –0.4 2.0 – 55 N 6 0
19 Dec 1997 X 1.27 –9.3 –9.0 –0.2 0.7 – 75 Y 5 0.28
22 Jan 1998 X 0.64 –9.5 –10.60 –1.7 0.9 1.0 132 N 4 0
29 Jan 1998 X 0.15 –7.5 –9.1 –1.9 1.5 1.7 82 N 3 0
17 Feb 1998 X 1.39 –3.5 –4.0 –1.7 0.7 1.3 186 N 2 0
30 Jan 1999 X 0.11 –12.20 –12.60 –0.8 1.5 – 82 N 3 0
11 Feb 1999 X 0.23 –9.4 –10.80 –0.1 0.6 *0.5* 120 N 3 0
Warming
28 Jan 1997 X 1.71 –2.2 –0.5 –1.8 –5.7 1.2 495 Y 5 8.81
29 Jan 1999 X 0.79 –11.20 –14.30 –2.3 –1.0 – 115 N 3 0
12 Feb 1999 X 0.99 –5.6 –5.0 –1.9 –2.6 *1.4* 94 Y 5 9.45

* Temperature difference over the distance 1 m to 0.1m.

 14698080, 2001, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1017/S1350482701001086 by Czech Technical University, Wiley Online Library on [28/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Prediction of hoar-frost by use of a Road Weather information System

Figure 1. Temperature development and wind speed during two typical situations when hoar-frost sublimates on the road sur-
face. (a) A cooling situation where the road surface cools more than the air. (b) A warming situation when the air temperature
is increased whereas the road temperature is lowered. On both occasions the road surface temperature is lowered below the dew
point.

height at 2 m makes it difficult to predict hoar-frost on skies, small dew point depression, high dew points,
the road surface. The maximum temperature difference strong winds, cold and clear conditions the preceding
between the air and the road surface is positive for all night, low road-bed temperature and a long night. The
occasions except one. Maximum air temperature gradi- second method multiplied the area between Tr and Td,
ents shows that there was an inversion in the lowest air in a graph, with the average wind speed (U in m s−1
layers on all occasions. where 1 unit is added to the average to account for the
stirring caused by traffic, i.e. U*= U + 1 m s−1).
3.2. Methods for predicting hoar-frost According to their results this figure was roughly pro-
portional to the depth of hoar-frost.
Several different methods are used to predict the pres-
ence and amount of hoar-frost deposited. As men- In this project the methods presented above were tested
tioned earlier the National Road Administration in to determine their hit rate for the days when hoar-frost
Sweden uses a simplified procedure with data from was observed (Table 2). The RWiS method and Method
RWiS stations to warn of hoar-frost on the road sur- 2 of Hewson & Gait is based on the difference between
face. The criterion for the dew point temperature is Td and Tr, therefore they could not predict the hoar-
lowered accordingly to avoid warnings of light hoar- frost for the occasions when the Td – Tr was negative
frost or inadequate measuring accuracy: (Table 2). It should be noted that the RWiS method is
only supposed to give a warning when hoar-frost is
Tr < 0 °C, Td – Tr > 0.5 °C and wind speed > 2 m s−1 present in amounts large enough to require salting or
gritting to prevent slipperiness. Even so, the method
Tr < 0 °C and Td – Tr > 1 °C when wind speed < 2 m s−1 still failed on some occasions to warn when large
amounts of hoar-frost were present on the road surface.
The warnings should be considered as an aid for the Method 1 of Hewson & Gait, where higher numbers
maintenance crew to decide whether or not to take indicates increased risk of hoar-frost, shows no corre-
action against slipperiness. Hewson & Gait (1992) used lation to the measured amount of hoar-frost shown in
two different methods to predict the amount of hoar- Table 2. This indicates clearly that there is a need for
frost. The first method was a checklist of factors which improvement in the methods used to predict the
increase the risk of hoar-frost. The factors were clear amount of hoar-frost deposited on the road surface.

99
 14698080, 2001, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1017/S1350482701001086 by Czech Technical University, Wiley Online Library on [28/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
M Karlsson
The following sections deal with the cause for the dis- duration of road surface temperature below dew point.
crepancy between measured and predicted hoar-frost Maximum difference between Ta and Tr and the time
and how measurements can be improved. period with a positive difference are also shown. In the
figure it can be seen that the amount of hoar-frost
increases with increasing wind speed. The increase in
3.3. Amount of hoar-frost related to wind speed results in more turbulence and transporta-
meteorological parameters tion of humidity to the road surface. There is a break-
point when increased turbulence and resulting smaller
Figure 2 shows the correlation between measured hoar- temperature differences reduce the deposition of hoar-
frost and factors measured at the RWiS station that are frost (Gocheva 1990). High minimum Tr increases the
mentioned as important in the literature. Factors amount of hoar-frost due to the non-linear correlation
shown in Figure 2 are average wind speed, minimum between temperature and saturation vapour pressure.
Tr, maximum difference between Td and Tr and the A temperature decrease at higher temperatures results

(a) (b)

(c) (d)

(e) (f)

Figure 2. Amount of hoar-frost correlated to parameters important for the development of hoar-frost. (a) Average wind
speed during the night. (b) Minimum Tr . (c) Maximum temperature difference between Td and Tr . (d) Time with a positive dif-
ference in Td – Tr . (e) Maximum temperature difference between Ta and Tr . (f) Time with temperature difference between
Ta – Tr .

100
 14698080, 2001, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1017/S1350482701001086 by Czech Technical University, Wiley Online Library on [28/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Prediction of hoar-frost by use of a Road Weather information System
in more humidity available for sublimation. The between amount of hoar-frost and friction. The largest
amount increases up to 0 °C where the sublimation amount of hoar-frost receives the smallest friction but
changes to dew deposition. The maximum difference with lower amounts of hoar-frost the friction is well
between dew point and Tr shows a tendency for the scattered and is probably determined by the hoar-frost
amount of hoar-frost to increase with increased differ- structure.
ence, but the fit is better when the difference between
Ta and Tr is used. The duration of a temperature differ-
ence, either Td – Tr or Ta – Tr indicates an increasing 3.4. Temperature and humidity gradient close to
amount of hoar-frost with increasing time. the road surface during hoar-frost situations
On some occasions the friction at the road surface was Hoar-frost is theoretically formed when the Tr is below
measured. These results are shown in Figure 3. The 0 °C, Tr is below the dew point and there is a flow of
friction against the road surface during hoar-frost humid air towards the road surface. However, the
events decreased with increased temperature difference results in Figure 2 and Table 2 show that the difference
between Td and Tr. The friction also decreased with between Td and Tr fails to indicate hoar-frost on some
increasing time with temperature difference between occasions. There could be several reasons for this, most
air temperature and Tr. There is a very small correlation of them related to the measuring technique at the RWiS
station. A study by Bergendahl (1998) showed that the
Tr could differ by up to 2.6 °C between measurements
(a) made with the infrared technique and a probe situated
2 mm below the road surface. Measurements with IR-
techniques have also shown that the road surface tem-
perature can vary by up to 1.5 °C across the road
(Gustavsson & Bogren, 1990b). Karlsson (1999)
showed that the temperature during occasions with
hoar-frost could differ by up to 1.7 °C between mea-
surements at 2.5 m and 0.1 m above the road surface.
Depending on the measuring equipment or their loca-
tion, these differences in temperature are probably the
reason why hoar-frost could not be predicted on all
occasions.
(b)
The air layer affected by the surface processes can be
determined from temperature and humidity profile
measurements taken above the road. These were mea-
sured at three levels, 0.1 m, 1.0 m and 2.5 m from
January to May 1998, and at 0.1 m, 0.4 m and 1 m dur-
ing January and February 1999. Results from two occa-
sions when hoar-frost developed on the road surface
are shown in Figure 4. The measurements from 29
January 1998 when temperature and humidity were
measured up to 2.5 m showed that the temperature
increased vertically to 1 m above the road surface by a
maximum of 2.2 °C m−1 and thereafter decreased by
(c) a maximum of 0.47 °C m−1 (Figure 4(a)). These tem-
perature gradients can be compared with Figure 4(b)
where the profiles were measured up to 1 m on
12 February 1999. There it is seen that the increase in
temperature and absolute humidity is between 0.4 m
and 1 m, with a maximum temperature gradient of
1.5 °C m−1 and only 1 °C m−1 between 0.1 and 0.4 m.
Humidity profiles show the same pattern.
Measurements of temperature and humidity close to
the surface reflect the surface processes better and are
therefore able to predict more accurately the amount of
hoar-frost. It is, however, impractical to measure too
close to the road surface owing to spray from the traf-
Figure 3. Friction at the road surface correlated to (a) time
fic and the likelihood of equipment being covered dur-
with temperature difference between Ta and Tr , (b) maximum
difference between Td – Tr and (c) amount of hoar-frost. Skid ing snowfalls. However, the results show that measure-
friction in µ. ments up to 0.4 m could be used to represent the con-

101
 14698080, 2001, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1017/S1350482701001086 by Czech Technical University, Wiley Online Library on [28/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
M Karlsson

Figure 4. Temperature and absolute humidity profiles above the road surface for two occasions when hoar-frost develops on the
road surface. (a) 29 January 1998 profile measured at three levels up to 2.5 m. (b) 12 February 1999 profile measured at three
levels up to 1 m.

ditions close to the road surface better than those mea- Comparison of the temperature and humidity change
surements taken at 2 m. Measurements at a height of can give a rough estimate of the occurrence of advec-
0.4 m are also less affected by spray from the road tion, evaporation and condensation at different levels.
surface than measurements at lower levels. The correlation between 10-minute changes in absolute
humidity and temperature at 0.1 m, 0.4 m, 1 m and 2.5
The temperature and humidity change at different lev- m for three occasions when hoar-frost developed on
els provides information on the amount of humidity the road surface is shown in Figure 5. The measure-
sublimated when there is a temperature decrease. ments at 0.1 and 0.4 m show a good correlation

102
 14698080, 2001, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1017/S1350482701001086 by Czech Technical University, Wiley Online Library on [28/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Prediction of hoar-frost by use of a Road Weather information System

Figure 5. The 10 minute absolute humidity change plotted against temperature change for 0.1 m, 0.4 m, 1.0 m and 2.5 m from
five occasions when hoar-frost sublimates on the road surface.

between the temperature and humidity changes levels or the differences between the original equip-
whereas the correlations at 1.0 and 2.5 m are poorer. A ment and these levels can be used to increase the accu-
decrease in air temperature without any exchange with racy of hoar-frost predictions.
other air masses results in a humidity increase since the
volume of the air mass is reduced when the air temper-
ature is lowered. However, at the lower levels a tem- 3.5. Multiple regression based on RWiS data and
perature decrease is associated with a humidity temperature profiles
decrease, i.e. moisture is deposited at the road surface
as hoar-frost. The pattern is more scattered at higher A forward multiple regression was performed on the
levels indicating that temperature and humidity data set to determine if the additional data measured in
changes are not closely associated with the processes at a profile above the road could increase the chance of an
the road surface. An increase or decrease in tempera- accurate hoar-frost prediction. The amount of data is
ture is also associated with a larger increase/decrease in small, but only the significant results (where p < 0.0005)
humidity at lower levels. A temperature decrease of 0.4 are presented. The input data to the regression were
°C at the 0.1 m level corresponds to a humidity retrieved from an RWiS station: wind speed, tempera-
decrease of 0.1 g m−3 compared with a decrease of ture difference between Ta and Tr, temperature differ-
0.06 g m−3 at 2.5 m, i.e. the amount of humidity that ence between Td and Tr, time with Tr < Td, minimum
sublimates is higher at lower levels. The trend for 0.1 m Tr, minimum Ta and temperature fall. Among these,
and 0.4 m is very similar. Measurements from these maximum temperature difference between Ta and Tr,

103
 14698080, 2001, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1017/S1350482701001086 by Czech Technical University, Wiley Online Library on [28/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
M Karlsson

Figure 6. Observed versus predicted values of hoar-frost from a forward multiple regression performed with data from (a)
RWiS and (b) RWiS and temperature profiles. The dotted line indicates the 95% confidence interval.

time with Tr < Td, maximum difference between Td and tions at the road surface, whereas the surrounding
Tr and minimum Ta were chosen to explain the varia- areas influenced the measurements at higher levels.
tion in amount of hoar-frost deposited. Together they (d) A forward multiple regression performed on the
produced a correlation coefficient of 0.82. When data data showed that measurements of temperature
from measuring profiles were included the correlation and humidity at lower levels increased the ability
coefficient increased to 0.93. The following factors to detect deposition of hoar-frost on the road sur-
were chosen to explain the variation in amount of face.
deposited hoar-frost: maximum temperature difference (e) Measurements of friction on the road surface, on
between Ta and Tr, time with Tr < Td, maximum differ- occasions with hoar-frost, showed that the friction
ence between Tr and Td, maximum temperature gradi- decreased with increasing difference between dew
ent above the road, time with a strong inversion above point and road surface temperature and increasing
the road and wind speed. In Figure 6 the predicted duration with Tr < Td . There was no correlation
amount of hoar-frost compared with the observed between amount of hoar-frost and friction at the
amount is shown for the two multiple regressions. Data road surface. The friction is probably not only
that can be retrieved from an RWiS are used in Figure influenced by the amount of hoar-frost but also the
6(a). Both RWiS and data from measuring profiles are structure of the hoar-frost, which has not been
used in Figure 6(b). The results show that the correla- determined in this study.
tion coefficient is high between the observed and pre- (f) The ability to estimate amount of hoar-frost with
dicted amount of hoar-frost, with data from the RWiS the help of measured weather parameters will
station used as input. However the chance of predicting increase if measurements are performed at lower
hoar-frost increases when additional measurements at levels.
lower levels are added to the input data.

Acknowledgements
4. Conclusions
Thanks to Professor S. Lindqvist, Dr T. Gustavsson,
The following conclusions can be made: Dr J. Bogren and my colleague Mr J. Norrman for
valuable comments on my work. I would also like to
(a) The amount of hoar-frost deposited on the road thank the staff at Bergab (Mr A. H. and Mr B. E.) who
surface increased with increasing average wind helped me to retrieve data from the test site. The
speed, maximum difference between dew point research has been funded by the Swedish National
(Td) and road surface temperature (Tr), the air min- Road Administration. Grants have also been provided
imum temperature and the duration when Tr < Td by the Swedish Society of Anthropology and
which has been shown in other studies. Geography (SSAG), the Anna Ahrenberg Foundation,
(b) The difference between Td – Tr failed to indicate the Hierta Retzius Foundation, the Lars Hierta
hoar-frost on several occasions due to the location Foundation and the Adlerbertska Foundation.
and poor accuracy of the measuring equipment.
(c) Measurements of temperature and humidity in a
profile up to 2.5 m above the road surface showed
that the air layer up to 0.4 m reflected the condi-

104
 14698080, 2001, 1, Downloaded from https://rmets.onlinelibrary.wiley.com/doi/10.1017/S1350482701001086 by Czech Technical University, Wiley Online Library on [28/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Prediction of hoar-frost by use of a Road Weather information System
References Jacobs, A. F. G., Van Pul, A. & El-Kilani, R. M. M. (1994).
Dew formation and the drying process within a maize
Bergendahl, S. (1998). Studier av yttemperaturvariationer canopy. Bound. Layer Meteorol., 69: 367−378.
med hjälp av IR-teknik, B-Series. In Swedish, Abstract in Karlsson, I. M. (1999). Temperature differences in the air
English. Available from Earth Sciences Center, Physical layer close to the road surface. (Published in Ph.D thesis,
Geography, Box 460, 405 30 Göteborg, Sweden. A42, ISSN-1400-3813, available from: Department of
Chen D., Gustavsson, T. & Bogren, J. (1999). The applicabil- Earth Sciences, Box 460, 405 30 Göteborg, Sweden.)
ity of similarity theory to a road surface. Meteorol. Appl., Kidron, G. J. (1998). A simple weighing method for dew and
6: 81−88. fog measurements. Weather, 53: 428−433.
Gocheva, A. (1990). Statistical distribution of air temperature, Kromer, I. (1995). Hungarian icing activity survey. Atmos.
relative humidity and wind velocity during rime-icing for Rec., 36: 311−319.
the non-mountain part of the territory of Bulgaria. Makkonen, L. & Ahti, K. (1995). Climatic mapping of ice
Proceedings of 5th International Youth School on loads based on airport weather observations. Atmos. Rec.,
Meteorology and Hydrology, 4: 84−89. 36: 185−193.
Gustafsson, K. (1981). Halka på olika väguppbyggnader. VTI Scherm, H. & Van Bruggen, A. H. C. (1993). Sensitivity of
rapport nr 216 (available from National Road and Traffic simulated dew duration to meteorological variations in dif-
Research Institute, 518 01 Linköping, Sweden). ferent climatic regions of California. Agric. For. Meteorol.,
Gustavsson, T. & Bogren, J. (1990a). Road slipperiness dur- 66: 229−245.
ing warm air advection. Meteorol. Mag., 119: 267−270. Stanev, S., Moralijski, E. & Velchev, K. (1987). Rime icing of
Gustavsson, T. & Bogren, J. (1990b). Infrared thermography the power lines in mountain regions. Bulgarian academy of
in applied road climatological studies, Int. J. Remote sciences, problems of meteorology and hydrology, 2 (avail-
Sensing, 12: 1811−1828. able from Institute of Hydrology and Meteorology, BAS
Gustavsson, T. (1991). Analyses of local climatological fac- Sofia, Bulgaria).
tors controlling risk of road slipperiness during warm-air Takle, E. S. (1990). Bridge and roadway frost occurrence and
advections. Int. J. Climatol., 11: 315−330. prediction by use of an expert system. J. Appl. Meteorol.,
Hewson, T. D. & Gait, N. J. (1992). Hoar-frost deposition on 29: 727−735.
roads. Meteorol. Mag., 121: 1−21.

105