Extreme events and Euro-Atlantic Extreme events and Euro-Atlantic atmospheric blocking in present and atmospheric blocking in present and future climate simulations future climate simulations Jana Sillmann Jana Sillmann Max Planck Institute for Meteorology, Hamburg International Max Planck Research School on Earth System Modelling Paris, SAMA seminar, 20 th January 2009
42
Embed
Extreme events and Euro-Atlantic atmospheric blocking in present and future climate simulations Jana Sillmann Max Planck Institute for Meteorology, Hamburg.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Extreme events and Euro-Atlantic Extreme events and Euro-Atlantic atmospheric blocking in present and atmospheric blocking in present and
Potential Vorticity (PV) - based Potential Vorticity (PV) - based blocking indicatorblocking indicator
Blocking detection method (Schwierz et al. 2004):
• Identification of regions with strong negative PV anomalies between 500-150hPa
• PV anomalies which meet time persistence (> 10 days) and spatial criteria (1.8*106km2) are tracked from their genesis to their lysis
Atmospheric blockingAtmospheric blocking
Representation in present and future Representation in present and future climateclimate
Blocking events > 10daysDJF
ERA-40 re-analysis model
Blocking frequency in %
1961-2000
2160-2199
Atmospheric blockingAtmospheric blocking
European blockings
(15°W-30°E,50°N-70°N)
%
Blocking frequency for DJFBlocking frequency %
1961-2000
Atmospheric blockingAtmospheric blocking
Correlation of European blockings with Correlation of European blockings with winter (DJF) minimum temperaturewinter (DJF) minimum temperature
Significant Spearman’s rank correlation coefficient to the 5% significance level
1961-2000 2160-2199
Extreme eventsExtreme events
Methods for extreme value analysis
Identification of extreme events in Identification of extreme events in climate dataclimate data
Statistical modeling of extreme values
• GEV – Generalized Extreme Value distribution
• parametric approach to characterize the distribution of extreme events
• calculation of return values
Indices for climate extremes
Stationary GEVStationary GEV
Generalized Extreme Value (GEV) Generalized Extreme Value (GEV) distributiondistribution
with parameters (location), (scale) and (shape)
Stationary GEVStationary GEV
Parameters for DJF minimum temperatureParameters for DJF minimum temperature
ERA-40
20C
A1B –20C
location scale shape
Non-stationary GEVNon-stationary GEV
Can we use the association between Can we use the association between extreme events and atmospheric blocking in extreme events and atmospheric blocking in the statistical modeling of extreme events?the statistical modeling of extreme events?
stationary GEV non-stationary GEV
COV – time dependent covariate
Atmospheric blocking as covariate derived from the PV-based blocking indicator (CAB)
where nllh0(M0) is the neg. log-likelihood of simple model
nllh1(M1) is the neg. log-likelihood of more complex model
nllh
353349348
example
*
* degrees of freedom
Non-stationary GEVNon-stationary GEV
Model selection for Model selection for minimum temperature extremes in winterminimum temperature extremes in winter
model model model
Non-stationary GEVNon-stationary GEV
Slope of the location parameterSlope of the location parameter
ºC/blocking freq. %
Non-stationary GEVNon-stationary GEV
Grid-point example at 9ºE, 53ºWGrid-point example at 9ºE, 53ºW
GEV distribution for the stationary and non-stationary model 1
Non-stationary GEVNon-stationary GEV
Return values at grid point 9ºE, 53ºWReturn values at grid point 9ºE, 53ºW
T-year return value … is the (1-1/T)th quantile of the GEV distribution
median
20-year return value
90% confidence interval
Non-stationary GEVNon-stationary GEV
20-yr return values for minimum 20-yr return values for minimum temperature extremes in wintertemperature extremes in winter
20C A1B
Significant differences between RV20 of stationary and non-stationary GEV distribution
ERA40
SummarySummary
• Is the model able to capture observed patterns of climate extremes?
• What changes in extremes can we expect under anthropogenic climate change?
increase of temperature and precipitation extremes as well as dry periods
regional and seasonal distinguished changes of extremes in future climate
SummarySummary
• Can we find associations between climate extremes and atmospheric blocking?
atmospheric blocking favors extreme cold nighttime temperatures in Europe
association remains robust in future climate, but influence of blocking events diminishes due to decreasing blocking frequency
• Can we use these associations in the statistical modeling of extreme events?
atmospheric blocking implemented as covariate in the GEV can explain more of the variability in the underlying data
modeling of colder return values possible
OutlookOutlook
Improvement of the statistical modeling:
• longer climate simulations (500-year control run) to further test the statistical robustness of the results
• apply Generalized Pareto distribution
• use other or more covariates
Usage of this methodology for statistical downscaling:
• limit region of interest, e.g. to northern, southern Europe
• find appropriate covariate for that region
• test method with observations
Thank you very much!Thank you very much!
Indices for extremesIndices for extremes
Is the model able to capture observed Is the model able to capture observed patterns of climate extremes?patterns of climate extremes?
HadEX dataset: indices for extreme events calculated on the basis of a worldwide weather observational dataset from the Hadley Centre (3.75° x 2.5° horizontal resolution) (Alexander et al. 2006)
Testing the method for Testing the method for El Nino El Nino and its impact on and its impact on precipitationprecipitation for for 1961-2000 winter (ONDJFM) winter (ONDJFM)
Modeling DiagnosticModeling Diagnostic
Best model
0 1 2 3 4 5
model #
xCOV(t)
(mm
)
5 4 3 2 1 0 -1 -2 -3 -4
Model Diagnostic at Grid Point [9E, 53N] for min.Tmin (ONDFM)
Model DiagnosticModel Diagnostic
0.0 0.2 0.4 0.6 0.8 1.0empirical
mo
de
l
0.0
0
.2
0.4
0
.6
0.8
1
.0
Probability Plot
Quantile Plot
-1 0 1 2 3 empirical
mod
el
0
2
4
Statistical modelingStatistical modeling
Block maxima approach
Daily minimum temperature data are blocked into sequences of length n, generating a sequence of block minima to which the GEV distribution can be fitted
• select block size (e.g., 1 season, 1 month)
• choose smallest event in each block (month or season)
• fit GEV distribution to selected extreme events
• estimation of GEV parameters for each global grid point via Maximum-Likelihood
Generalized Extreme Value (GEV) Generalized Extreme Value (GEV) distributiondistribution