Eva Plavcová, Jan Kyselý, Petr Štěpánek: Links between circulation indices/types and precipitation in Central Europe in RCMs
1. Links between circulation
indices/types and precipitation in
Central Europe in RCMs
Eva Plavcová, Jan Kyselý, Petr Štěpánek
e-mail: plavcova@ufa.cas.cz
kysely@ufa.cas.cz
2. Motivation
Regional climate models (RCMs) – driven by GCMs, re-
analysis
Tool to simulate climate at the regional and local scale and
future climate change scenarios
Biases in their simulations of recent climate – important to
identify sources of these errors and to address them in
further developing the models
Particularly useful to identify links between individual
variables and their biases (dependence of errors on
atmospheric circulation, soil moisture characteristics, cloud
amount, etc.)
May bring insight into physical processes affecting model’s
performance or reveal potential common deficiencies
3. Data
Daily precipitation amounts (PREC), daily mean sea level pressure
40-yr period (1961–2000), summer (JJA) and winter (DJF) seasons
Observations (OBS)
PREC - taken from the gridded dataset (GriSt) interpolated from a
high-density station network operated by the CHMI
Pressure data from the ERA-40 re-analysis
Models
5 RCMs (~25km) from the ENSEMBLES project, runs driven by the
ERA-40 re-analysis (‘perfect’ boundary condition, comparison)
Institution RCM Reference
DMI (Danish Meteorological Institute) HIRHAM Christensen et al. (1996)
KNMI (Royal Netherlands Meteorological Institute) RACMO Lenderink et al. (2003)
ICTP (Abdus Salam International Centre for
RegCM Giorgi et al. (2004)
Theoretical Physics)
MPI (Max-Planck Institute) REMO Jacob (2001)
SMHI (Swedish Meteorological and Hydrological Kjellström et al. (2005),
RCA
Institute) Samuelsson et al. (2011)
4. Methods
Links investigated for PREC in 3 regions in the Czech Republic:
CL – Central lowland (Polabí, 18 gridpoints) – 257 m a.s.l.
SH – Southern highland (Šumava, 7 gridpoints) – 790 m a.s.l.
EH – Eastern highland (Jeseníky, 7 gridpoints) – 597 m a.s.l.
Precipitation amounts area-averaged in most cases
5. Atmospheric circulation
Represented by 3 circulation indices (Jenkinson & Collison 1977) derived from gridded MSLP by
means of equations given in Plavcová & Kyselý 2011
flow direction (DIR)
flow strength (STR)
flow vorticity (VORT)
The circulation indices are used to produce a simple classification of 27 circulation types (Jones et
al 1993):
STR and VORT < 6: unclassified U
|VORT| ≥ 2∙STR & VORT>0: strongly cyclonic C
|VORT| ≥ 2∙STR & VORT<0: strongly anticyclonic AC
Directional types:
|VORT| < 1∙STR, DIR is divided into 8 quadrants:
N, NE, E, SE, S, SW, W, NW
Hybrid types (combination of directional and
non-directional):
CN, CNE, CE, CSE, CS, CSW, CW, CNW
AN, ANE, AE, ASE, AS, ASW, AW, ANW
6. Atmospheric circulation
Predominate westerly flow (especially in winter)
Relative frequency < 0.5%:
4 types in winter (CS, CSW, CSW, ANE)
3 types in summer (CS, CSW, ANE)
RCMs driven by the ERA-40 re-analysis
Overestimation of STR
More strong cyclonic flow at the expense of anticyclonic flow
7. Seasonal mean PREC in RCMs
• Mean seasonal precipitation amount:
[mm/season] DJF JJA Differences larger
than ±10% are
CL SH EH CL SH EH highlighted by colour
observation 103 181 167 213 284 287
HIRHAM 144 328 244 178 385 311
RACMO 130 208 187 150 289 220
RegCM 187 270 212 231 290 274
REMO 139 137 124 235 235 210
RCA 134 246 189 233 425 361
• Biases due to biases in circulation/links between circulation and PREC
• In winter: all RCMs (except REMO in highlands) simulate more
precipitation
• In summer:
RCA overestimates PREC
RACMO underestimates PREC in CL and EH
HIRHAM enhances differences between lowland and highlands
RegCM vs. REMO: (un)realistic orography x PREC in highlands
(RegCM – the ‘best’ model in summer, REMO does not capture
differences for highlands and lowlands)
8. Mean precipitation – OBS
Mean observed PREC on days falling into each index bin and circulation type
To ensure representativeness, means are only calculated for bins with samples of at least 15 days throughout the
period, grey lines represent relative frequencies of circulation type/index bin
Links are more pronounced for highlands (SH, EH)
In summer, strong links between VORT and PREC: large mean precipitation amounts during
cyclonic flow days
Stronger flows – larger PREC amounts in both seasons
Generally larger mean PREC for westerly (Atlantic influence) and northeasterly flows (cyclones
over central and eastern Europe, Mediterranean cyclones) in comparison to southerly flow; in CL,
these links are weaker
9. Mean precipitation – RCMs
RCMs are able to capture some basic features:
links more pronounced for highlands (except REMO)
larger mean precipitation amounts during cyclonic flow days than anticyclonic in both seasons
stronger flows – larger PREC amounts in both seasons
RCMs have difficulties to capture some features:
unrealistic links to DIR in summer – larger PREC for SW flow, for northerly flow large PREC only in
EH (x CL and SH)
CN in EH – to some extend reproduced in some RCMs
CSW - only in EH x in all/no regions
JJA
10. Probability of wet day - OBS
Wet day = when PREC > 0.3 mm
Links are better pronounced than for mean PREC
Relationships are analogous in both (all) seasons
The link to DIR is larger in winter and for highlands (northerly flow – more wet days) – orography
particularly important in inducing PREC for flow from the NE quadrant
The link to VORT is very strong in summer, when the probability of wet days is almost equal to 1
for strongly cyclonic days
Generally, more wet days are for westerly flow, higher STR and cyclonic flow
more dry days: southerly flow, low STR and anticyclonic flow
11. Probability of wet day - RCMs
RCMs are generally able to capture the links
more dry days for southerly flow, low STR and anticyclonic flow
All RCMs overestimate probability of wet days in winter, the overestimation is largest for CL
Some RCMs do not capture the lower probability of wet days for NE flow in CL in comparison to
highlands
12. Probability of wet day - RCMs
The link to VORT:
in winter, RCMs overestimate probability of wet days for both strong AC and C flow (except AC in
REMO)
in winter, all RCMs except RCA overestimate the differences between probability of wet days for
strong AC and C flow days (REMO most)
in summer, RegCM and RCA simulate more wet days among strong AC days, and that is why the
differences between strong AC and C days are much smaller than in the observed data
Probability of wet days for 2% of days with strongest anticyclonic/cyclonic
flow, and their differences; averaged over 3 regions
DJF
PROB DJF JJA
wet AC C C-AC AC C C-AC
0.22 0.85 0.63 0.18 0.92 0.74
observed
0.31
0.96
0.65
0.18
0.93
0.76
HIRHAM
0.27
0.97
0.70
0.09
0.88
0.79
RACMO
0.28
0.99
0.71
0.42
0.92
0.50
RegCM
0.12 0.95 0.84 0.12 0.90
0.77
REMO
0.38 0.96 0.58 0.52 0.97
0.45
JJA
RCA
13. Probability of heavy rainy day - OBS
Heavy rain = mean area-averaged PREC amount over all grids in a region > 90% quantile in
given season
The probability is high especially for
strong flow days
strong cyclonic flow
westerly (winter) and northeasterly (summer) flow
14. Probability of heavy rainy day - DJF
Winter:
very low probability for anticyclonic types – all RCMs
simulate less heavy rainy days for AC types
higher probability for westerly types
RCMs reproduce these links
Highest probability of heavy rainy day
in SH – CW type – RegCM, REMO
HIRHAM – W, CNW; RACMO – W, RCA – CNW
in EH – CN, W
RACMO and RCA – CSW, CSE
in CL – W, CE, CS
HIRHAM – W; RACMO – W; RegCM – C, CW;
REMO – CSW, W; RCA – W, SW
15. Probability of heavy rainy day - JJA
Summer:
high probability for strongly C types
Highest probability of heavy rainy day
in CL – C, SW
HIRHAM – CS, C; RACMO – CW (0.5); RegCM –
CSW; REMO – CW, W; RCA – CW, CSW
in SH – C, CS type – HIRHAM
RACMO and RegCM – CSW; REMO – C; RCA -
CNW
in EH – CN, CNE, CSW
HIRHAM – CN; RACMO – CW, CN, CSW; RegCM
and REMO – CSW; RCA – CSW, CN
16. Types with extreme PREC - DJF
Extreme PREC = in at least one grid over the
region, grid box PREC higher than the 99%
quantile of PREC distribution in that grid
Barplot – total number of events, crosses – relative
frequency for given circulation type
OBS
largest probability for CSW and W types in
CL and SH, while for CS type in EH
in SH, almost all events occur during
westerly types
never for CW, CNW, AN, ANE, AE, ASE
RCMs
simulate extreme PREC for CW and CNW
types
17. Types with extreme PREC - JJA
OBS
largest probability for C and CNE types in all
regions (U and A are the most frequent
circulation types in summer)
in CL, CS ~30%, SW
in EH also CN
never for AN, ANE, AE, ASE
RCMs
are relatively able to capture the types
during which extreme precipitation events
occur/do not occur
18. Summary
RCMs overestimate mean seasonal precipitation amount in winter, and some models have biases
in summer, too
There are quite strong links in the observed data between atmospheric circulation and mean PREC,
probability of wet days and days with heavy and extreme precipitation
RCMs are generally able to capture most of the links
However, RCMs often do not capture the differences of the links for the regions (highland x
lowland) – RCM outputs more useful/realistic at a larger scale than several gridboxes
If the links between large-scale atmospheric circulation and precipitation are realistically reproduced
in a climate model, biases in simulation of total precipitation amounts could be attributed to biases
of large-scale circulation itself
• (e.g. HIRHAM simulates more strong flow days + strong flow days are linked to higher PREC
=> HIRHAM overestimate mean PREC)
References:
Jenkinson AF, Collison FP (1977) An initial climatology of gales over the North Sea, Synoptic Climatology Branch Memorandum No. 62,
Meteorological Office, Bracknell, United Kingdom
Jones PD, Hulme M, Briffa KR (1993) A comparison of Lamb circulation types with an objective classification scheme. International Journal of
Climatology 13: 655–663
Plavcová E, Kyselý J (2011) Evaluation of daily temperatures in Central Europe and their links to large-scale circulation in an ensemble of
regional climate models. Tellus 63A: 763–781
Editor's Notes
Motivace pro vsechny jasna
Most widely used tools for simulating climate e.g. with respect to dependence of errors on atmospheric Such studies focusing on links of surface climate characteristics to atmospheric circulation and/or other variables may bring insight into physical processes affecting model’s performance or reveal potential common deficiencies if an ensemble of climate model simulations is examined.
Central Europe Dobre porovnani kdyz rizene ERA-40
CL region represents the main agricultural area main mountain ranges with windward precipitation effects
purely
Nejvice zapadni proudeni, v zime AC, v lete U a AC caused