Transcript Click to edit Master title style • • • • • Click to edit Master text styles Long-term trend analysis of Second level aerosol parameters at the Third level Fourth levelJungfraujoch Fifth level Martine COLLAUD COEN, MeteoSwiss, Switzerland Ernest WEINGARTNER, Stephan NYEKI.

Click to edit Master title style
•
•
•
•
•
Click
to edit Master
text styles
Long-term
trend
analysis of
Second level
aerosol
parameters
at
the
Third level
Fourth levelJungfraujoch
Fifth
level
Martine COLLAUD COEN, MeteoSwiss, Switzerland
Ernest WEINGARTNER, Stephan NYEKI and
Urs BALTENSPERGER, Paul Scherrer Institute,
Switzerland
[email protected]
1
1
The Sphynx station at the Jungfraujoch
Click to edit Master title style
•
•
•
•
•
Click to edit Master text styles
Second level
• 3580 m asl
• GAW station
Third level
• Partially in free
troposphere (FT)
Fourth level
• Influenced by PBL
Fifth level
• Remote, aged particles
• 40% in-cloud
2
2
Aerosol parameters
to edit Master
title style
at Click
the Jungfraujoch:
1995-2005
•Scattering
Click coefficient
to edit Master text
styles

sp
at  = 450, 550, 700 nm
•Backscattering
Second level
coefficient
 bsp
at

=
450,
550,
700
nm
• Third level
fraction
b   bsp /  sp
•Backscattering
Fourth
level
at  = 450, 550, 700 nm

•Scattering
Fifth level
exponent
  a *
Absorption coefficient
at 7  : 370 nm  950 nm
Condensation Nuclei
 ap
CN
Nephelometer
(TSI 3563)
Nephelometer
(TSI 3563)
ratio
fit
Spectrum
Aethalometer (AE31)
CPC (TSI 3010)
3
3
10.5 years of measurement
Click to (1995-2005)
edit Master title
•
•
•
•
•
Click to edit Master text styles
Second level
Third level
Fourth level
Fifth level
style
Daily median
Monthly RM
Yearly RM
4
4
Seasonal Mann-Kendall test
Click to edit Master title style
The Mann-Kendall test is based on rank and allows to determine if a
• Click totrend
edit
Master
textconfidence
styles limit.
exists
at a chosen
• Second level
• Third level
•Seasonality are taken into account.
•
Fourth
level
•Missing values, ties in time (several measurements per season) and ties
in•values
arelevel
allowed
Fifth
•It is a non parametric test that can therefore be applied to all distributions
•The covariance is corrected by the Dietz and Killeen estimator.
•The variance is corrected for data autocorrelation by the procedure
described by Hamed and Rao (1998).
5
5
Sen‘s
slopeMaster
estimatortitle style
Click
to edit
Sen‘s slope estimator is a non parametric estimate of the linear trend.
•
•
•
•
•
•
•
Click to edit Master text styles
It allows missing values and ties.
Second level
The Sen‘s slope is given by the median of all Aij, with
Third level
(Ci  C j )
FourthA level
j>i , ti<>tj
ij 
(ti  t j )
Fifth level
• Confidence limits at 90% have been evaluated (Gilbert 1998).
6
6
Least mean square fit and number of
Click
to edit to
Master
title
style
years
necessary
detect the
trend
•Y(t)
Click
styles
= to+ edit
St +Master
(t/12)text
+ Ut
+ Nt, t = 1,…n,
• Second level
Seasonal
Stationnary
auto-regressive
noise
AR(1):
LMQ fitcomponent:
has been applied
on the
log of the data
when
0
if
t the
 T data had a

=
constant
Intervention
trend
slope
U 
St=• sin(2t/365.25)
+ sin(4t/365.25)+
cos(2t/365.25)
ThirdLinear
level
lognormal
Nt =with
Nt-1
+ distribution.
1 if t  T
•Number
Fourth
level
of years
necessary to detect the estimated trend (Weatherhead,
2000) :
• Fifth level
0
t
0
 (2  z  )
1 
n 
N

1   
 
*
2/3
*
1
1  3 (1   )
1/ 2
7
7
LMQ fit of the monthly median of the
Click to edit
Master
title
scattering
coefficient
at 700
nmstyle
Ln(scattering coef)
•
•
•
•
•
Click to edit Master text styles
Second level
Third level
Fourth level
Fifth level
June-August
NovemberDecember
Slope= 4%/year
Time
8
8
Results of the seasonal
test title
Click Mann-Kendall
to edit Master
Months
1
2
3
4
5
6
7
8
9
10
• Click to edit Master text styles
• Second level
Backscatt
• Third level
CN
Abs
• Fourth level
Abs
• Fifth level
Scat
white
11
style
12
Significant at 90%
Positive slope
Significant at 95%
Positive slope
Significant at 90%
Negative slope
Significant at 95%
Negative slope
Bfraction
Scat exp
9
9
Seasonal cycle of the scattering
Click to edit
Master
title
style
coefficient
• Click to edit Master text styles
• Second level
coef. level
• Abs.
Third
coef level
• Scat.
Fourth
• Fifth level
10
10
Sen’s slope of the scattering
Click to coefficients
edit Master title
style
Scattering coefficient
Slope in %/year
• Click to edit Master text styles
12.00
• Second level
8.00
• 4.00
Third level
• 0.00
Fourth level
• -4.00
Fifth level
450 nm
550 nm
700 nm
UCL 700 nm
LCL 700 nm
-8.00
1
2
3
4
5
6
7
Months
8
9
10 11 12
Significant trends at 90%
Significant trends at 95%
11
11
Sen’s slope of the backscattering
Click to edit
Master title style
fraction
Sen's slope [%/year]
•
•
•
•
•
backscattering fraction
Click to edit Master text styles
Second
level
0.00
Third level
-4.00
Fourth level
-8.00
Fifth level
450 nm
550 nm
700 nm
UCL 700 nm
LCL 700 nm
-12.00
1
2
3
4
5
6
7
Months
8
9
10 11 12
Significant trends at 90%
Significant trends at 95%
12
12
Results of the LMQ:
Click
to edittrends
Master
title
Significant
at 95%
•
•
•
•
•
style
*
Slope
of
ln(data)
Slope
of
data
Nb
of
years
n
edit Master text styles
%/year
%/year
Click to
Second
level
Scattering
coef.
Third level
Fourth level
Backscattering
coef.
Fifth level
Backscattering
fraction
Scattering exp.
0.21
2.9
10.2
0.33
4.0
7.6
0.35
3.9
7.5
0.18
3.2
8.9
0.24
4.0
7.6
0.17
3.5
8.8
-2.3
6.5
-2.6
6.6
-2.8
6.6
5.0
3.7
13
13
Click to edit
Conclusions
Master title style
The scattering coefficients have a positive significant trends of 3•☺Click
to edit Master text styles
4% yr-1.
• Second level
☺ The autumn and winter are the periods with the most significant
trends.level
• Third
There is level
no trend in the summer months with the greatest PBL
•☺Fourth
influence.
• Fifth level
☺
The particle size in the free tropospheric air masses decreases.
☺
Increase of aerosol background concentration.
[email protected]
14
14