CONTRAIL CIRRUS COVERAGE AND RADIATIVE FORCING

FROM SATELLITE DATA

Hermann Mannstein(1)

, Waldemar Krebs(1)

, Simon Pinnock(2)

, Frank Jelinek(2)

(1) Institut für Physik der Atmosphäre, Deutsches Zentrum für Luft- und Raumfahrt e.V. (DLR),

Oberpfaffenhofen, D-82230 Wessling, Germany, email: Hermann.Mannstein@dlr.de (2)

European Space Agency – ESRIN, Via Galileo Galilei, I-00044 Frascati (RM), Italy, email:

simon.pinnock@esa.int

(3) EUROCONTROL Experimental Centre, Centre de Bois des Bordes, BP 15, F-91222 Brétigny sur

Orge, CEDEX, France, email : frank.jelinek@eurocontrol.int

ABSTRACT

Within the ESA DUE project 'Contrails' the amount of

linear contrails has been analysed using ENVISAT

AATSR and NOAA AVHRR data. The impact of air-

traffic on cirrus coverage as well as the first time on

outgoing radiation flux density has been estimated by

using simultaneous observations from Meteosat Second

Generation (MSG) SEVIRI instrument and high spatial

and temporal resolution air-traffic data from

EUROCONTROL (European Organisation for the

Safety of Air Navigation). The study shows some

correlations between air-traffic density on the one side

and cirrus coverage, outgoing long and short wave

radiant flux densities on the other side. The relation

between air-traffic and radiation fluxes depends

strongly on daytime and season.

1 INTRODUCTION

The impact of aviation on climate follows several

pathways. Carbon dioxide and water vapour, both

effective greenhouse gases, are emitted as well as nitric

oxides, which influences the chemical composition of

the upper troposphere. Soot and sulphuric oxides add to

the ambient aerosol and have an impact on cirrus

formation and cloud microphysical properties. Since

the IPCC special report on "Aviation and the Global

Atmosphere" [1] it is known and widely accepted that

contrails and the cirrus clouds evolving out of them

have a climate impact comparable to the CO2 from the

combustion process. These additional, purely man-

made clouds change the radiative forcing of the earth-

atmosphere system: they reduce the incoming solar

radiation as well as the outgoing thermal radiation in a

way that the mean net balance at top of the atmosphere

is slightly positive – i.e. they add to the greenhouse

effect [2]. The role of contrail and cirrus formation

within the total impact of aviation on climate was

confirmed at the Aviation, Atmosphere and Climate

(AAC) Conference 2003 [3]. In [4] we find linear

contrails as the only explicitly mentioned radiative

forcing term from the traffic sector (see Figure 1) with

a very high range of uncertainty from 0.003 to 0.03

Wm-2

.

Figure 1: Radiative forcing components from [4]

2 THE DLR CONTRAIL DETECTION

ALGORITHM

The analysis of satellite data to infer contrail frequency

and coverage started at IPA in 1989. Several attempts

have been made to change from visual inspection of

images to an operational, automated system. In 1997

this development succeeded [5] and resulted in the

DLR contrail detection algorithm (CDA), which is

considered to be world wide the only operational

algorithm for this task. The development aimed at data

of the AVHRR thermal infrared split window channels,

and the algorithm has also been successfully used with

ATSR and MODIS data. It was applied in studies to

derive contrail coverage over Europe [5,6], SE and E-

Asia [7], USA [8].

A further development was the statistical interpretation

of the detected contrails and a first assessment of the

radiative forcing by [7]. Large data volumes allow to

estimate the false alarm rate, i.e. the part of false

positive detections, and to correct for variations in the

detection efficiency of the algorithm.

3 VALIDATION OF THE CDA

Validation was performed by comparison of the CDA

with human visual interpretation of satellite scenes.

_____________________________________________________

Proc. ‘Envisat Symposium 2007’, Montreux, Switzerland 23–27 April 2007 (ESA SP-636, July 2007)

Figure 2: Contrail detection in an ATSR2 scene by

three different human interpreters: in red, green and

blue; the mixed colours represent overlap of detected

contrails.

55 MERIS, 12 AATSR, 4 ATSR2, and 20 AVHRR

scenes have been independently analysed at least by

two, and some of them by three human interpreters.

Figure 2 indicates the differences in the interpretation

of the same scene, which is the major source of

uncertainty in the determination of the detection

efficiency (DE). Figure 3 shows an ENVISAT-MERIS

scene with many contrails in different stages of

development. In Figure 4 this scene is analysed for

contrails by the CDA and two humans.

Figure 3: ENVISAT-MERIS scene (2004-03-08,

Northern Spain).

Figure 4: Automated contrail detection from

AATSR(red) in combination with visual analysis from

MERIS (green) and AATSR IR channels in the right

image.

An overall detection efficiency of ~25% was

determined from all visual analyses.

The false alarm rate (FAR) of the CDA is obtained

from the results in comparison to air traffic data. All

detected contrails in regions without air-traffic within

one hour before the ENVISAT overpass are considered

to be false alarms. For the AATSR the FAR is 0.10%.

Figure 5: left – 2004 mean air traffic density within

one hour before each ENVISAT overpass, right mean

false alarms 2004

Figure 6: False alarm rate vs. monthly mean cirrus

coverage

More than 90% of the false alarms are natural cirrus

cloud bands. Thus we find a strong dependency of

FAR on cirrus coverage and approximate this

dependency with a function of the cirrus coverage -

c·(1-c) - as shown in Figure 6

4 CONTRAIL COVERAGE FROM AATSR

Linear contrails have been analysed with the DLR

CDA from all available ENVISAT AATSR scenes for

the year 2004 (see Figure 7 and Figure 9) and

corrected as far as possible against false alarm rate and

detection efficiency (Figure 8 and Figure 10).

Obviously the CDA misses many contrails in areas

with high air traffic density (see Figure 11 ). We

attribute this behaviour to the fact, that in these areas

the linear contrails quite often can be found in the

vicinity of or embedded in contrail cirrus.

Figure 7: Contrails analysed by the CDA for AATSR

daytime scenes 2004

Figure 8 Contrails analysed by the CDA and corrected

for FAR and DE for AATSR daytime scenes 2004

Figure 9: Contrails analysed by the CDA for AATSR

night-time scenes 2004

Figure 10: Contrails analysed by the CDA and

corrected for FAR and DE for AATSR night-time

scenes 2004

Figure 11: Contrail coverage analysed by the CDA

for all AATSRe scenes 2004 vs. air traffic density

5 CONTRAILS FROM AVHRR AND

(A)ATSR(2) DATA

NOAA AVHRR and ATSR data has been analysed

with the CDA for the years 1985, 1990, 1995, 2000,

and 2004 at DLR and the Dundee Satellite Receiving

Station (DSRS). The results have been sampled into

5°x5° boxes for 4 day time slots, 4 seasons and every

year and corrected against FAR and DE for every

satellite. After the correction a combined guess of

coverage by linear contrails and the associated

uncertainty was calculated with an Bayesian approach.

The results are shown in Figure 12 and Figure 13, the

data source composition in Figure 14

Figure 12: Mean contrail coverage of the area between

80° W and 50° E, 20° N and 75° N for the years 1985,

1990, 1995, 2000, and 2004(top), winter, spring,

summer and autumn of all years (middle) and different

times of the day (bottom) estimated from all analysed

data.

Figure 13 Mean contrail coverage from NOAA-9, -11,

-12, -14, -15, -16, -17, ATSR2 and AATSR for the years

1985, 1990, 1995, 2000, and 2004

Figure 14 Data source for the CDA: Number of pixels

analysed from NOAA-9, -11, -12, -14, -15, -16, -17,

ATSR2 and AATSR for the years 1985, 1990, 1995,

2000, and 2004

6 CONTRAIL CIRRUS COVERAGE

Air traffic initiates the formation of cirrus clouds in ice

supersaturated regions without cirrus coverage (cpot).

From METEOSAT8-SEVIRI infrared channels we

derive the cirrus coverage for the year 2004 with the

MeCiDa [9] algorithm and find a perfect agreement to

the conceptual model of random overlap filling of cpot.

Similar results have been already found based on an

analysis of 60 days with METEOSAT data [10]

Figure 15 Cirrus coverage derived from METEOSAT-

SEVIRI infrared channels vs. air traffic density for the

region indicated by the insert. The red curve indicates

the conceptual model

c(d)=co+cpot(1-exp(-d/d*))

cpot

d*

Figure 16 Cirrus coverage from different months of the

ECHAM4 climate model vs. air traffic density.

Figure 17 Sum of outgoing longwave and shortwave

radiative flux density derived from METEOSAT8

SEVIRI data

Figure 18 Outgoing shortwave (left) and longwave

radiative flux density derived from METEOSAT-

SEVIRI vs. air traffic density.

Cirrus coverage derived from METEOSAT data was

correlated to the air traffic density (sum of flight paths

over a region within a time interval) by [10], but the

interpretation of the significant correlation is

problematic, as the natural cirrus coverage, i.e. the

undisturbed case is unknown [11]. Analyses of the

cirrus coverage within the ECHAM4 climate model,

which has no information on air traffic (Figure 16,

[11]) indicate, that unavoidable spurious correlations

between air traffic and natural cirrus coverage restrict

the quantitative evaluation of this significant

correlation.

The same problems show up with the interpretation of

the correlation between SEVIRI-derived outgoing

radiances (Figure 17) and the air traffic density (Figure

18).

7 CONCLUSIONS AND OUTLOOK

The work on verification of the contrail detection

algorithm (CDA) demonstrated the vagueness of the

term 'linear contrails'. For this study the CDA was

tuned to reach a very low FAR. This resulted in a low

DE in areas with high air traffic densities like in

Central Europe. On a truly global scale the analysis of

linear contrails is still missing. The ESA-(A)ATSR(2)

data set provides a consistent base for further

evaluations over at least one decade. Such an analysis

could reduce the large uncertainties indicated in [4]

The correlation method to estimate the coverage and

forcing by contrail cirrus gave significant results, but

spurious correlation has an impact on the result. A final

quantification of the radiative forcing from contrails

and contrail cirrus is still not possible. Life cycle

studies of contrail cirrus which can shed more light on

this problem are ongoing. On a global scale a cirrus

climatology derived from (A)ATSR(2) with

sophisticated methods which are not sensitive to

instrument degradation could help.

Acknowledgements

This study was funded by ESA DUE. The project

‘Contrails’ was performed as a cooperation between

the ‘Deutsches Zenrum für Luft- und Raumfahrt e.V.’

(DLR), the ‘Dundee Satellite Receiving Station’

(DSRS), and the ‘Koninklijk Nederlands

Meteorologisch Instituut’ KNMI. We thank Knut

Dammann, Rüdiger Büll, Gerhard Gesell, Bernhard

Mayer,(from DLR), Hans Roozekrans, Paul DeValk,

Arnout Feijt ( from KNMI), Andrew Brooks, Neil Lonie

(from DSRS) James Smith, Ted Elliff, Andrew Watt

(from EUROCONTROL), Olivier Arino and Josef

Aschbacher (from ESA) for their work and support

8 LITERATURE

1 Intergovernmental Panel on Climate Change, 1999.

Aviation and the Global Atmophere. A special

Report of IPCC Working Groups I and III.

Cambridge University Press Cambridge

2 Meerkötter, R., Schumann, U., Doelling, D.R.,

Minnis, P., Nakajima, T., Tsushima, Y., 1999.

Radiative Forcing by Contrails. Annales

Geophysicae 17, 1080-1094.

3 Sausen, R., Fichter, C., Amanatidis G., (Eds.),

2004. Aviation, Atmosphere and Climate

(AAC). Proceedings of a European Conference,

Friedrichshafen, Germany, 30 June to 3 July

2003. European Commission, Air pollution

research report 83, ISBN 92-894-5434-2.

4 Intergovernmental Panel on Climate Change, 2007

WG1, Summary for policy makers,

http://www.ipcc.ch/WG1_SPM_17Apr07.pdf

5 Mannstein, H., Meyer, R., Wendling, P.:

Operational Detection of Contrails from

NOAA-AVHRR-Data., International Journal of

Remote Sensing, 20, 8, (1999), pp. 1641-1660,

6 Meyer, R., Mannstein, H., Meerkötter, R.,

Schumann, U., Wendling, P.: Regional

Radiative Forcing by Line-Shaped Contrails

Derived from Satellite Data.Journal of

Geophysical Research, 107, D10

(10.1029/2001JD000426), (2002), S. ACL 17-1-

ACL 17-15,

7 Meyer, R., Büll, R., Leiter, Ch., Mannstein, H.,

Pechtl, S. Oki, T., Wendling, P., Contrail

observations over Southern and Eastern Asia in

NOAA/AVHRR data and comparisons to

contrail simulations in a GCM, Internalional

Journal of Remote Sensing, (2007)

8 Minnis, P., Schumann, U. , Doelling, D.R., Gierens,

K., Fahey, D.W.: Global Distribution of

Contrail Radiative Forcing. Geophysical

Research Letters, 26, 13, (1999), S. 1853-1856

9 Krebs, W., Mannstein, H., Bugliaro, L., Mayer, B.:

A new day- and night-time Meteosat Second

Generation Cirrus Detection Algorithm

MeCiDA, 2007, submitted to ACP

10 Mannstein, H., Schumann, U.: Aircraft induced

contrail cirrus over Europe. – Meteorol. Z.

14(1), (2005), pp. 549–554.

11 Mannstein, H., Schumann, U.: Corrigendum to

Aircraft induced contrail cirrus over Europe. –

Meteorol. Z. 16(1), (2007), pp. 131–132.