Auxiliary material for Paper 2009GL038557

Statistical uncertainty of changes in winter storms over the North Atlantic and 
Europe in an ensemble of transient climate simulations

Paul M. Della-Marta
Federal Office for Meteorology and Climatology, MeteoSwiss, Zurich, Switzerland

Winter Storm Research, Partner Reinsurance Company, Zurich, Switzerland

Joaquim G. Pinto
Institute for Geophysics and Meteorology, University of Cologne, Cologne, 
Germany


Della-Marta, P. M., and J. G. Pinto (2009), Statistical uncertainty of changes 
in winter storms over the North Atlantic and Europe in an ensemble of transient 
climate simulations, Geophys. Res. Lett., 36, L14703, doi:10.1029/2009GL038557.

Introduction

This auxiliary material contains five figures, one table, and two text files. 
Figures S1, S2, and S3 are self explanatory and are referred to in the main 
text. Table S1 is also referred to in the main text. Texts S1 and S2 refer to 
Figures S4 and S5, respectively. These sections contain detailed results on the 
sensitivity of results to the choice of threshold value and the influence of 
rapidly deepening cyclones on the changes in storm intensity under A2 scenario.

1. 2009gl038557-fs01.eps
Cyclone track density over the North Atlantic region for the present climate 
conditions (1960-2000) with the ECHAM5 GCM as ensemble average [cyclone 
days/winter]. Data corresponds to extended winter October-March. Also shown are 
the selected regions for the analysis of the return periods of storms. Region 1: 
green box; region 2: blue box; region 3 red box. Cyclone distribution adapted 
from Pinto et al.[2007, Figure 1c].

2. 2009gl038557-fs02.eps
Return periods of storms in terms of (a, c and f) CP for Box 1, 2 and 3 
respectively and (b, d and f) VOR for Box 1, 2 and 3 respectively. Blue curve 
corresponds to ECHAM-20C storms, grey curve to ERA40 storms, and black curve to 
NCEP storms. Individual values are shown as single dots. Dot-dashed lines 
indicate the 90% CI limits. The units are hPa and hPa/(deg. lat.^2) for CP and 
VOR respectively. Horizontal dashed gray line denotes the GPD threshold. Note 
the logarithmic x-axis.

3. 2009gl038557-fs03.eps
As for Figure 2, change in 20C 1 year RP based on moving 40 year climatology for 
CP in Box 2.

4. 2009gl038557-ts01.txt
Comparison of the annual average storm frequency (Lambda) in terms of VOR 
(hPa/(deg.lat^2)) ECHAM5 20C with A1B and A2 storm frequency including 90% CI 
estimates for each region of the North Atlantic (Boxes 1,2 and 3). Columns Lower 
and Upper denote the lower and upper bounds of the 90% CI respectively. Only 
Lambda for under A2 in Boxes 2 and 3 are significantly different at the 1% level 
from the corresponding 20C Lambda. Percentages in brackets in the left column 
indicate the threshold for the GPD.
4.1 Column "Region", denotes which Box is used and the GPD threshold chosen for 
each box. Refer to Figure S1 and the main text for more details.
4.2 Column "ECHAM5-20C", contains three sub-columns denoted "Lower", "Lambda" 
and "Upper". Columns "Lower" and "Upper" denote the lower and upper bounds of 
the 90% CI respectively. Whereas "Lambda" is the estimate of the annual average 
storm frequency.
4.3 Column "ECHAM5-A1B", contains three sub-columns denoted "Lower", "Lambda" 
and "Upper". Columns "Lower" and "Upper" denote the lower and upper bounds of 
the 90% CI respectively. Whereas "Lambda" is the estimate of the annual average 
storm frequency.
4.4 Column "ECHAM5-A2", contains three sub-columns denoted "Lower", "Lambda" and 
"Upper". Columns "Lower" and "Upper" denote the lower and upper bounds of the 
90% CI respectively. Whereas "Lambda" is the estimate of the annual average 
storm frequency.

5. 2009gl038557-txts01.txt
The sensitivity of results to the threshold chosen for fitting the generalized 
Pareto distribution. This text refers to Figure S4.

6. 2009gl038557-fs04.eps
The shape parameter of the GPD distribution as a function of the threshold used 
to fit the GPD to (a) A1B VOR Box 1, (b) A1B VOR Box 2 and (c) A1B VOR Box 3 for 
the 2060-2100 period based on all three ensemble members. The vertical black 
lines denote the 95% CI calculated using a parametric resampling technique 
[Della-Marta et al., 2009]. The numbers aligned vertically indicate the number 
of storms identified in each box and threshold. The lowest threshold value in 
each plot is the 70th percentile The vertical black lines represent the 
respective 97.5, 95 and 95% 20C quantiles (a,b and c respectively), which were 
used as thresholds for A1B to fit the GPD distributions (shown in Fig 1b and d 
red lines).

7. 2009gl038557-txts02.txt
Testing the relevance of rapidly deepening cyclones near the British Isles for 
the identified shortening of the RP over Western Europe. This text refers to 
Figure S5.

8. 2009gl038557-fs05.eps
Same as Figure 1 (c and d respectively) with British bomb cyclones removed. Blue 
curve corresponds to 20C storms, red curve to A1B storms (See Figure 1 caption).


