Content uploaded by Esbjorn Olsson
Author content
All content in this area was uploaded by Esbjorn Olsson on Feb 01, 2019
Content may be subject to copyright.
1. Introduction
Icing is a significant meteorological hazard for aviation
(see Pike 1995 for an example of an icing aviation acci-
dent). In a joint project between the Swedish Armed
Forces and the Swedish Meteorological and Hydro-
logical Institute, a small working group of meteorolo-
gists has made a study of aircraft icing and forecasting.
The output from that project, which ended in 1999, was
an updated forecaster’s guide (in Swedish) for use when
forecasting icing on aircraft and helicopters within and
below clouds. A new algorithm for icing in clouds was
also developed by the working group.
An icing index from a numerical model is a useful tool
for estimating icing conditions, but it is important to
recognise that such an index will not solve the whole
forecasting problem. The forecaster must still be
trained to recognise those weather situations that are
likely to cause ice accretion, as this remains the only
method for forecasting icing below clouds.
2. A new icing algoritm
In the late 1990s cloud-water was introduced as a vari-
able in the numerical models and a test with a simple
index including water content from the model was
developed at the Swedish Meteorological and Hydro-
logical Institute:
Index A = 5 + ln cw
where ln is the natural logarithm and cw is the total
amount of cloud-water (g/kg) from the model regard-
less of whether it is frozen or liquid.
The index was calculated for every grid-point in the
Swedish HIRLAM model (22×22 km and 31 levels) and
gave a number from 1 to 5 where the amount of cloud-
water was high enough. The number could then be
used as an indicator of the expected icing intensity in a
suitable temperature interval, which was chosen as 0 to
–15°C.
Index A was a fairly good indicator of areas where icing
conditions might be expected, although the horizontal
extension sometimes seemed too large and the area
with the strongest icing intensity was not always well
defined. The index had the same tendency as the fore-
caster to overestimate the icing. The reason for this is
not fully explained but one factor might be the use of
the total cloud condensate from the model instead of
the liquid water content. The true liquid water content
was not available from the model. The index was sup-
posed to represent the worst case; hence the total cloud
condensate was used instead of an amount of liquid
water according to a statistical formula based on tem-
perature.
However, the working group took this further and
tried to take into account, directly or indirectly, all the
main meteorological variables for icing:
• temperature
• amount of liquid water
• size of the droplets.
The total cloud condensate continued to be taken as
representative of the amount of liquid water. The verti-
cal velocity was chosen as a substitute for the droplet
size, which is not available from the numerical models.
Index A was modified to an Index B using the vertical
velocity, w, from the model according to Table 1.
Meteorol. Appl. 10, 111–114 (2003) DOI:10.1017/S1350482703002020
A new algorithm to estimate aircraft icing in the
HIRLAM model
Bernt Olofsson, Swedish Armed Forces Headquarters, SE-107 85 Stockholm, Sweden
Esbjörn Olsson, Swedish Meteorological and Hydrological Institute, SMHI Sundsvall,
SE-830 60 Sörberge, Sweden
Sven Andersson, Thomas Mårtensson & Ebba Mårtensson, Swedish Armed Forces’
Forecasting Center, Box 420, SE-746 29 Bålsta, Sweden
A new index to estimate aircraft icing in clouds from operational meteorological models has been
developed by Swedish meteorologists. Although rather simple it takes into account, directly or
indirectly, all the principal meteorological variables for icing. The index has been evaluated during
three winter seasons and is now operational in the Swedish HIRLAM model. A graphical
representation of the index is presented.
111
There are two reasons for giving vertical velocity a rel-
atively greater weighting in the index:
• positive vertical velocity increases the likelihood
that most of the cloud-water is supercooled; and
• updrafts in the order of tens of cm/s favour the
production of big cloud drops up to the size of
drizzle, which we know increases the intensity of
the icing.
Index B has now been operational in Sweden for four
years. It is calculated at every grid-point where the
HIRLAM model indicates cloud-water and where the
temperature at the same time is below 0 °C. It gives a
value from 0 to 9 depending on cloud-water and verti-
cal velocity, which has been calibrated to indicate icing
intensity.
3. Presentation
It is very important that the index values are clearly dis-
played for the forecaster so that the information is easy
to interpret. The working group suggested a plot
model, which is now in use in Sweden. The highest
intensity at any level is plotted as a coloured symbol at
the grid-point. The lower and upper limits of the icing
layer as well as the level of maximum intensity are also
plotted. In this way the colours indicate the horizontal
distribution and the maximum intensity of the icing
risk while the numbers indicate the vertical extent of
the icing layer. An example of the numerical icing fore-
cast is shown in Figure 1.
4. Evaluation
Undertaking a thorough evaluation of icing indices is
an enormous task. In the USA much work has been
done on this (see, for example, Brown et al. 2002). A
subjective evaluation of the Swedish index has been
made in real time by the aviation forecasters on duty at
military and civil forecasting offices using all available
information. The general judgement was that Index B
gave a more accurate and detailed picture of the
expected icing conditions than Index A. However, it
was noticed that both indices had a tendency to overes-
timate the icing intensity at high altitudes. This is prob-
ably due to a combination of operational and meteoro-
logical reasons. Aircraft flying at high altitudes are gen-
erally operating at high speed and the icing risk is
reduced by the frictional heating of the airframe. The
indices do not discriminate between cloud liquid and
cloud ice. Low temperatures increase the likelihood
that the cloud-water will turn into a frozen state and
thus lower the risk of icing.
Although the index was originally supposed to con-
sider only the meteorological factors it was adjusted to
take into account these operational experiences. An
easy way to do this was to reduce the amount of liquid
water linearly between –15 and –40 °C. Thus Index A
was modified slightly, with the total amount of cloud-
water (cw) being replaced by a reduced amount (cwr)
when the temperature is below –15 °C. The formula is:
cwr = cw * (40+T)/25
where T is the temperature (°C).
According to experiences from three winter seasons in
Sweden, Index B can roughly be translated into the
ICAO definitions of icing intensity according to Table
2. The index is calibrated to estimate the maximum
icing intensity that is likely to be experienced by an air-
craft in a particular weather situation.
It should be noted that the index has been evaluated
only in the present version of the HIRLAM model in
Sweden. It should not be used operationally in other
models and in other climatic zones without proper test-
ing. However, the Swedish aviation forecasters believe
that the index provides a forecast that is as good as, or
sometimes even better than, an observational forecast
in situations with lows, fronts and orographical
updrafts. In convective situations the index does not
always add very much extra information. The vertical
resolution of the model is still not good enough to
Bernt Olofsson et al.
112
Table 1. Index A modified to an Index B using the vertical velocity, w, from the model (positive w corresponds to
ascent).
w (cm/s) < 0 0–10 10–20 20–30 > 30
Index B A–1 A+1 A+2 A+3 A+4
Table 2. Correlations between Index B and ICAO icing intensity according to evaluation in the Swedish version
of the HIRLAM model.
ICAO No icing Light Light–moderate Moderate Moderate–severe Severe
Index B 0–1 2–3 4 5–6 7 8–9
Estimating aircraft icing in the HIRLAM model
113
Figure 1. An example of a numerical icing forecast. The highest intensity at any level is plotted as a coloured symbol in the
grid–point. Only index values of 4 or more are used. Index 4 is plotted with a green cross, 5–6 with a yellow square, 7 with an
orange square and 8–9 with a red square. In each square the level of maximum icing intensity is plotted in thousands of feet
(nearest to the level of the model). Above and below the square the highest level and the lowest level of index 4 are plotted. In
this way the colours indicate the horizontal distribution and the maximum intensity of the icing risk while the numbers indi-
cate the vertical extent of the icing layer.
allow forecasting of severe icing in inversions, which
may sometimes be dangerous in winter, at least in
Sweden. Furthermore, freezing precipitation is not
covered by this method as precipitation water is not
included in the present version of the index.
5. Summary and conclusion
An icing index can be constructed in many ways but it
is important that all the main meteorological variables
governing the intensity of the icing are included. To
estimate icing conditions in clouds, an index derived
from a numerical model can be a useful tool but it can-
not solve the whole forecasting problem. A forecaster’s
guide on icing and close co-operation with the pilots
are also essential to ensure good forecasting. An icing
index from a model should always be considered as a
forecasting aid and not a readymade forecast.
References
Brown, B. G., Mahoney, J. L. and Fowler, T. L. (2002)
Verification of the In-flight Icing Diagnostic Algorithm
(IIDA). Proceedings of the 10th American Meteorological
Society Conference on Aviation, Range and Aerospace
Meteorology, Portland, Oregon, May 2002.
Pike, W. S. (1995) Extreme warm frontal icing on 25 February
1994 causes an aircraft accident near Uttoxeter. Meteorol.
Appl. 2: 273–279.
Bernt Olofsson et al.
114