Content uploaded by Patrick Minnis
Author content
All content in this area was uploaded by Patrick Minnis on Jun 07, 2014
Content may be subject to copyright.
RSPSoc 2009, 8-11th September 2009, Leicester, UK Page
The use of Meteorological Data to Improve Contrail
Detection in Thermal Imagery over Ireland.
Gillian M. Whelan1, Fiona Cawkwell1, Hermann Mannstein2 and Patrick
Minnis3
1Department of Geography, University College Cork, College Road, Cork, Ireland
Email: whelan.gillian@student.ucc.ie
2DLR Institute for Atmospheric Physics, Oberfaffenhofen, Germany
3NASA Langley Research Centre, Virginia, USA.
Summary
Aircraft induced contrails have been found to have a net warming influence on the climate
system, with strong regional dependence. Persistent linear contrails are detectable in 1 Km
thermal imagery and, using an automated Contrail Detection Algorithm (CDA), can be
identified on the basis of their different properties at the 11 and 12 µm wavelengths. The
algorithm’s ability to distinguish contrails from other linear features depends on the
sensitivity of its tuning parameters. In order to keep the number of false identifications low,
the algorithm imposes strict limits on contrail size, linearity and intensity. This paper
investigates whether including additional information (i.e. meteorological data) within the
CDA may allow for these criteria to be less rigorous, thus increasing the contrail-detection
rate, without increasing the false alarm rate.
1 Introduction
Contrails are artificial linear ice-clouds that form in the wake of jet aircraft, when the hot and
moist exhaust gases mix with much colder ambient air. Under suitable meteorological
conditions, contrails can persist for several hours and trigger additional cirrus cloud
formation. Previous studies into the climate impact of linear contrails have found that
contrails produce a small but significant net warming effect on the climate system (Stuber et
al., 2006). The IPCC’s fourth assessment report (2007) assigned a global Radiative Forcing
(RF) of +0.010 Wm-2, while Stuber et al. (2006) quote a regional value over a high traffic
location in the UK of +0.23 Wm2 – one order of magnitude higher than the global estimated
value. However, the climate impact of contrail-cirrus may be between 2 to10 times larger than
this. The high regional dependence of this effect suggests that for countries such as Ireland,
who have a high density of air-traffic, these effects could be even greater than is outlined in
the latest IPCC report. Lee et al. (2009) present an updated estimate of global linear contrail
RF of +0.0118 Wm2 for 2005 based on updated air-traffic operations data. No best-estimate
for the RF of induced-cirrus-cloudiness is presently available, although Lee et al. (2009)
postulate that it could be 0.033Wm2, although with large uncertainties and based on a very
low level of scientific understanding.
Furthermore, aviation demand is reportedly increasing, with the total number of aircraft
expected to double in the next twenty years, and annual aviation fuel use to increase by 3.9%
(Whitelegg, 2006). From 2002-2008, overflights in Irish air-space increased substantially,
such that the maximum number of overflights in 2008 was double that of the 2002 minimum
(Figure 1).
RSPSoc 2009, 8-11th September 2009, Leicester, UK Page
Figure 1 Monthly total overflights through Irish air-space.
Most of the North Atlantic flight Tracks (NATs), are located directly to the west of Ireland,
with a large number of high altitude (above 24,000ft) overflights crossing Ireland every day
peaking during the early (eastbound traffic) and late (westbound traffic) morning. Overflights
increased from 2002-2008 but values for 2009 (January to April) are consistently less than for
other years (Figure 1).
Persistent spreading contrails can induce additional cirrus cloud formation. Palle and Butler
(2001) observed a 15% increase in Irish cloud cover and a corresponding 20% drop in annual
sunshine hours at four ground stations from 1881-1998. What proportion, if any, of this cloud
cover increase can be attributed to increasing aviation activity over Ireland? To what extent
are Irish skies and climate modified by high altitude overflights? In order to answer these
questions, an objective evaluation of contrail coverage trends over Ireland is needed. Satellite
imagery is the only source of data that allows the objective production of a cloud and contrail
climatology over the course of a whole year. Presently, the Contrail Detection Algorithm
(CDA) that operates on the dual thermal channels of A/ATSR or AVHRR imagery is
deliberately tuned to have a low false alarm rate, but this results in a low contrail detection
efficiency also. This paper investigates the possibility of including meteorological
information within the CDA to improve its detection efficiency without increasing its false
alarm rate.
2 Aims and Objectives
Both meteorological data (with air-traffic data) and satellite imagery can be used
independently to investigate contrail occurrence and persistence; but this research aims to
evaluate whether integrating upper atmospheric pressure, humidity and temperature data
(from radiosondes), in conjunction with known meteorological contrail formation conditions,
as part of the decision-making criteria within the CDA, (originally developed by Mannstein et
al. (1999) at DLR) improves contrail detection without increasing false alarms.
RSPSoc 2009, 8-11th September 2009, Leicester, UK Page
3 Contrail Detection from Satellites
The low brightness temperature of contrails in the thermal infrared and their characteristic
linear shape allow them to be detected by passive remote sensing methods. Linear contrails
are visible in the 11 and 12 µm channels of AVHRR and AATSR, but due to their higher
transmissivity at the shorter thermal wavelength, they show up more clearly in 11-12 µm
temperature difference images. Mannstein et al. (1999) proposed a fully automated algorithm
for evaluating contrail coverage in AVHRR 11 and 12 µm thermal imagery, and the same
method is applicable to A/ATSR imagery.
The CDA has two main inputs; the 11-12 µm temperature difference (TD) image and the 12
µm image. Normalisation and filtering techniques are applied to these images (Mannstein et
al., 1999), and linear features which are potentially contrails extracted. In order to identify
which of these objects are contrails, the algorithm subjects each to a series of threshold
‘checks’ to reject those which are definitely not contrails. Unfortunately, a lot of actual
contrails are also eliminated by this approach. The contrail-checks examine the size, linearity
and intensity of each object, and the thresholds set within the algorithm for each of these
parameters are tuneable. Optimising the CDA to limit the misdetection rate is an iterative
process. Using this technique, Mannstein et al. (1999) successfully identified approximately
30 to 50% of those contrails visibly recognisable in the original TD image, keeping the false
alarm rate at just 0.1%.
The CDA has been applied to AATSR imagery of Ireland and the surrounding coastal waters,
as shown in Figure 2a for 17/03/2009 at 11:18UTC. Figure 2b shows the 11-12 µm
temperature difference for this region, and Figure 2c the linear features extracted by the CDA
prior to checking. Figure 2d shows the features identified by the CDA as contrails, with a
coverage in this area of 2.061%. The CDA was run with default (stringent) threshold criteria,
therefore this value of ~2% contrail-coverage, although high for the area, is nonetheless
considered to be a conservative estimate of the contrail-coverage in this scene. Based upon
preliminary results, this value of 2% from the imagery is very unusual. Initial evaluations for
contrail-coverage (when present) generally range from 0.1 to 0.6%; which is more consistent
with the value of 0.5% obtained by Mannstein et al. (1999) for central Europe in 1996 (which
was regarded as a conservative estimate).
Figure 2: AATSR image for 17/03/2009 at 11:18UTC. a) True colour image, b) 11-12 µm temperature
difference image, c) extracted linear features and d) identified contrails (~2% coverage).
The timing of the image shown in Figure 2 coincides with one of the heaviest periods of
aircraft traffic crossing Ireland, predominantly in a westward direction from continental
Europe and the UK. If conditions have been conducive for contrail formation for some time, it
is possible that the cirrus coverage seen across the centre of Ireland is contrail-induced-cirrus.
RSPSoc 2009, 8-11th September 2009, Leicester, UK Page
Radiosonde measurements of ambient atmospheric pressure, temperature and humidity allow
for this hypothesis to be further explored.
4 Combining Meteorological Data with CDA
Aircraft contrails, or ‘condensation-trails’, form as a result of the mixing of hot and humid
exhaust air with much colder ambient air below a critical temperature threshold, as defined by
the ‘Schmidt-Appleman’ criterion (Figure 3) (see Schumann, 1996 for more details).
Figure 3 Schmidt-Appleman criterion for contrail formation. The red line represents the state of the
exhaust air as it mixes with environmental air. Where this line crosses the saturation vapour pressure
curve for water, a contrail will from. If ambient air is saturated with respect to ice, contrails will persist.
The critical temperature for contrail formation is dependent upon ambient pressure, relative
humidity and the aircraft’s contrail factor, with the critical temperature calculated for each
atmospheric layer (Schumann, 1996) from radiosonde data. Figure 4 shows a radiosonde
ascent from Valentia at 11:15UTC coincident with the AATSR overpass discussed above,
with the contrail-susceptible layer of the upper atmosphere shown to occur centred around the
200hPa pressure level (which corresponds to approximately 35,000ft – a typical cruising
altitude for transatlantic flights). In this instance, the radiosonde’s path might not take it
directly through one of the contrails shown on the image in Figure 2 however it does indicate
the potential for upper atmospheric conditions to support contrail formation at this time.
RSPSoc 2009, 8-11th September 2009, Leicester, UK Page
Figure 4 Radiosonde sounding indicating contrail susceptible atmospheric layers for the same date and
approximate time as the AATSR imagery in Figure 2.
If the atmospheric relative humidity is above ice-saturation, contrails that form will persist
under these conditions, as hypothesised for the image in Figure 2. Several studies (Schumann
1996; Steufer et al., 2005; Stuber et al., 2006; Rädel and Shine, 2007) have successfully
predicted contrail persistence using these criteria with reasonable accuracy. Based on these
results, a modification of Mannstein’s CDA is proposed to include an additional check that
would only identify an ‘object’ as a contrail if the Schmidt-Appleman criterion was met and
the atmospheric relative humidity was ice-supersaturated. If the meteorological conditions for
contrails to form and persist are not met then the ‘object’ on the image is unlikely to be a
contrail. By this means more non-contrail objects can be excluded, thus allowing the other
contrail-identification checks to be more inclusive, without increasing the false alarm rate. A
flow diagram outlining our approach is below (Figure 5).
Figure 5 Outline of approach to modify CDA with meteorological data.
The effect of this modification will be ascertained using an interactive software tool
developed by researchers at NASA, and described by Minnis et al. (2005) which allows the
user to manually add missed contrails and remove false alarms. The software tracks these
‘corrections’ to gauge the efficiency of the algorithm. By comparing the output from the
‘pure’ CDA and the ‘meteorological’ CDA the impact of including meteorological data on the
overall detection efficiency will be assessed.
RSPSoc 2009, 8-11th September 2009, Leicester, UK Page
5 Discussion and Conclusion
As demonstrated by the sample image shown here (Figure 2), there are occasions when the
meteorological conditions in the upper atmosphere over Ireland support unusually high
contrail formation and persistence, with eventual dispersion into high level cirrus cloud. As
shown by Figure 4 the atmospheric conditions are conducive to contrail formation at an
altitude commensurate with transatlantic aircraft, supporting the results of the CDA. On the
date shown in the image here, clear skies over Ireland would have allowed the high level
contrail coverage to be observed from the ground, but it is not currently known how prevalent
such extreme contrail coverage incidents are, or whether their number and duration is
increasing as the amount of aircraft has increased. This research aims to use satellite imagery
to gain a greater insight into the long term trends in contrail-coverage over Ireland, and
subsequently to provide an estimation of their RF contribution and the regional effects of air-
traffic on the Irish climate.
6 Acknowledgements
This work is being funded under the EPA’s STRIVE programme (2007-2013) project code
2007-PhD-ET-4. Meteorological data has been provided by Met Éireann. Irish air-traffic
information and statistics have been provided by the Irish Aviation Authority. AATSR data
have been provided by ESA, under the IRE-LUX AO, project code 4753. Software was also
provided by DLR (CDA) and NASA (Interactive Assessment Tool). The authors of this paper
would also like to thank Dr. Una Ni Chaoimh from UCC for many useful discussions and
assistance.
References
LEE, D. S., FAHEY, D. W., FORSTER, P. M., NEWTON, P. J., WIT, R. C. N., LIM, L. L.,
OWEN, B., SAUSEN, R., 2009. Aviation and global climate change in the 21st
century, Atmospheric Environment, 43:3520-3537.
MANNSTEIN, H., MEYER, R., WENDLING, P., 1999. Operational detection of contrails
from NOAA-AVHRR-data, Int. J. Remote Sensing, 20:1641-1660.
MINNIS, P., PALIKONDA, R., WALTER, B. J., AYERS, J. K. and MANNSTEIN, H., 2005.
Contrail properties over the eastern North Pacific from AVHRR data, Meteor. Z.,
14:515-523.
PALLE, E. and BUTLER, C. J., 2001. Sunshine records from Ireland: cloud factors and
possible links to solar activity and cosmic rays, International Journal of Climatology,
21:709-729.
RÄDEL, G., and SHINE, K. P., 2007. Evaluation of the use of radiosonde humidity data to
predict the occurrence of persistent contrails, Q. J. R. Meteorol. Soc., 133:1413-1423.
SCHUMANN, U., 1996. On conditions for contrail formation from aircraft exhausts. Meteor.
Z., 5: 4-23.
STEUFER, M., MENG, X. and WENDLER, G., 2005. MM5 Contrail Forecasting in Alaska.
Monthly Weather Review, 133(12):3517.
STUBER, N., FORSTER, P., RÄDEL, G., SHINE, K., 2006. The importance of the diurnal
and annual cycle of air-traffic for contrail radiative forcing. Nature. 441:864-867.
WHITELEGG, J., 2006. Managing the Environmental Impacts of Globalisation on Transport:
Aviation Emissions. Organisation for Economic Co-operation and Development
(OECD), ENV/EPOC/WPNEP/T(2006)8.
