Past and projected future changes
of North Atlantic
polar low frequency
Matthias Zahn
Institute of Coastal Research
Helmholtz-Zentrum Geesthacht
Matthias.Zahn@hzg.de
March 11, 2015
Visiting Uni Res./Bjerknes Ctr, 11 Mar 2015, Bergen, Norway
Polar lows, characteristics
◮
◮
◮
◮
◮
◮
diameter < 1000km
strong winds > 13.9 ms
heavy precipitation
poleward Polar Fronts in
winter
spiral cloud and cloud free
core
comma shape cloud pattern
c
Dundee
Satellite Receiving Station
Long-term: climatology
Research questions:
◮
◮
◮
how do frequencies change?
do areas of occurrence shift?
(how do intensities change?)
Requirements:
◮
◮
◮
long in time
high in spatial detail
homogeneous
Problem: DO NOT EXIST
Solution: dynamical downscaling
low-res global-domain
❄
Model grid
◮
high-res RCM-domain
◮
global data (re-analysis or GCM) long
in time and homogeneous
downscaled by means of Regional
Climate Model (RCM)
COSMO-CLM-2.4.6.
Example of a reproduced PL case
15 Oct 1993, 6:00, The Swan
Example of a reproduced PL case
15 Oct 1993, 6:00, The Swan
Next Problem: detection
Decades of data
◮
◮
◮
eye detection
infeasible for
long-term data
humans decide
subjectively
automated
procedure needed
Solution:
◮
◮
counted cases of 2 yrs, Blechschmidt (2008), GRL
spatial band-pass filter
separates information at different scales
Example of band-pass filtered MSLP
200-600km retained, PLs emerge as distinct minima
Detection and tracking scheme
1. record all locations at which filt. MSLP minimum ≤ −1hPa
2. combine detected positions to individual tracks, distance to
next (3h) pos ≤ 200km
3. checking further constraints along tracks:
◮
◮
◮
◮
◮
filt. minimum ( ≤ −2hPa once along track)
wind speed (≥ 13.9 ms once along track)
air-sea temperature difference (SST − T500hPa ≥ 43K )
no northward direction of track
limits to allowable adjacent grid boxes
OR: minimum in band-pass filtered MSLP ≤ −6hPa once
Tracks of 3 (among >3300) polar lows
Reproduced and detected after more than 40 years simulation time
Annual numbers of PLs (1949-2005)
Number of PLs per Polar Low Season (PLS), one PLS from July until June next year
Comparison with observed cases
bias, but qualitative
similarity to observation
data
black: our data (Zahn and v.Storch, 2008)
adjusted to observation data
red: MetNo (pers. comm.)
Wilhelmsen (1985)
Blechschmidt (2008)
Spatial density of PLs (1949-2005)
80˚
280˚
70
˚
˚
290
60
˚
0˚
350˚
340
˚
0˚
33
0˚
˚
32
40
31
0˚
˚
30˚
20˚
10˚
Number of PLs per area unit
50
0˚
30
5.5
5.0
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Future projections
90˚
280˚
80˚
70
˚
0˚
29
60
˚
0˚
30
˚
50
0˚
350˚
˚
340
0˚
33
0˚
˚
40
30˚
20˚
Model grid
10˚
32
◮
0˚
◮
driven by ECHAM5/MPI-OM
C20: control with GHG 1960-1990
B1,A1B,A2: GHG for 2070-2100 (AR4)
31
◮
Projected cumulative frequency of PLs in
IPCC-scenarios and annual cycle
significant decrease in the
number of PLs per winter
14
Blechschmidt
Wilhelmsen
REA
C20
B1
A1B
A2
average number of polar lows per month
12
10
8
6
4
2
0
Jan
Mar
May
Jul
month
Sep
Number of PLs per month
Nov
Spatial density distribution, differences
northward shift of genesis region
Polar Lows and projected vertical stability
◮
◮
atmosphere warms stronger
than ocean surface
proxy for frequency of
favourable PL conditions
decreases
Evolution of vdT (SST and T500hPa )
Polar Lows and projected vertical stability
Zahn and v.Storch (2010), Nature
◮
◮
Area and time-averaged icefree SST − T500hPa
over maritime northern North Atlantic
proxy for frequency of
favourable PL conditions
decreases
large inter model bias, but same
direction of change
Evolution of vdT ( SST and T500hPa )
Linkage of vertical stability changes to
ocean circulation
Spatial response vertical
stability
Regression with ocean
circulation
Multi-model mean response (A1B - C20) in
vertical stability S = T500hPa − SST (= −vdT )
over ice free ocean (Black iso-lines: SST
response)
S regressed on the AMOC response across
CMIP3 model (A1B - C20) space over ice free
ocean (Black iso-lines: regression SST-AMOC)
Frequency changes of Polar Lows
No change in recent past, but high
interannual variability
Significant decrease of annual number
in response to global warming
Decrease linked to more stable mean
conditions
Currently: global downscaling 1948 - 2014
low-res global-domain
❄
CliSAP Research Topic B-4: Regional
Storms and their Marine Impacts
◮ apply a global model - ECHAM6
◮ dynamical downscaling to global
domain (M. Schubert-Frisius)
◮ spectral nudging
◮
◮
high-res global-domain
global data at T255 (50-70km)
resolution
will be finished soon (in March)
Comparison to observed typhoons in 2004:
Identify storm tracks each of the EH6 experiments
◮
◮
◮
apply spectral filter to vorticity, T13-T65 (1500-200km)
detect locations of th1 >= 0.00006 1s and merge them to
tracks
dismiss tracks with:
◮ wind speed below 13.9 m/s
◮ shorter then 4 time steps
◮ land along more than 50% of track
Compared with observation data of storms
◮
IBTrACS-WMO : Tropical Cyclone observation data
provided by WMO
https:\\www.ncdc.noaa.gov\ibtracs\
Simulating 2004 tropical cyclones
Tracks of the 2004 season
30˚
60˚
50˚
40˚
60˚
90˚
120˚
150˚
180˚
210˚
240˚
270˚
300˚
330˚
Track with counterpart
Track without counterpart
surplus tracks
0˚
NCEP
China
25˚
25˚
Taiwan
30˚
20˚
10˚
20˚
20˚
0˚
-10˚
Phillipines
-20˚
15˚
-30˚
120˚
990
1000
15˚
130˚
1010
1020
hPa
1030
-40˚
EH6
China
-50˚
25˚
25˚
Taiwan
2004 named IBTRACS WMO tracks,
red/black: without/with detected counterpart in EH6
◮
◮
◮
◮
can improve TC representation
69 tropical cyclones named
51 corresponding tracks detected
next step: discriminate these from
non-TC detections
20˚
20˚
Phillipines
15˚
120˚
990
1000
130˚
1010
1020
15˚
hPa
1030
NOCK-TEN in NCEP-SP and
EH6-MSLP, 24 Oct 2004, 0:00
black crosses: observed,
green cross: detected
Separation by linear discriminant analysis
histograms of max_ws along tracks
ws > 33
MSLP < 990
score > −0.5
3
48
0
366
87.0%
20
31
40
326
81.6%
30
21
3
363
92.7%
0.10
0.15
ws > 18
33
18
146
220
59.7%
0.00
0.05
Density
0.20
0.25
mean max ws
ideal
51
0
0
366
100%
5
10
15
20
25
30
25
30
◮
maximum 10m wind speed [m/s]
0.15
0.10
0.00
0.05
Density
0.20
0.25
min max ws
5
10
15
20
◮
maximum 10m wind speed [m/s]
0.05
0.10
0.15
◮
0.00
Density
0.20
0.25
max max ws
5
10
15
20
25
30
maximum 10m wind speed [m/s]
with CP
without CP
without CP in tropics
◮
separates objects of different
classes based on their statistical
properties
here: with/without counter part
uses several variables at different
vertical levels
improves tracking compared to
standard configurations
Thank you very much for your attention
Mange takk for oppmerksomheten
http://coast.hzg.de/staff/zahn/