Methods for diagnosing extreme climate events in gridded data sets

1
Methods for diagnosing extreme climate events in
gridded data sets

• D. J. Steinskog
• D. B. Stephenson, C. A. S. Coelho and C. A. T.
  Ferro

Mines Paris, Fontainebleau, 20 March 2007
2
Outline

• What are extremes in climate?
• Short info about R and RCLIM
• Methods for looking at extremes in gridded
  datasets
• Future development
• Conclusions

3
Climate extremes
4
What is an extreme in meteorology?

• Large meteorological values
• Maximum value (i.e. a local extremum)
• Exceedance above a high threshold
• Record breaker (thresholdmax of past values)
• Rare event
• (e.g. less than 1 in 100 years p0.01)
• Large losses (severe or high-impact)
• (e.g. 200 billion if hurricane hits Miami)
• risk p(hazard) x vulnerability x exposure

5
Examples of wet and windy extremes
Convective severe storm
Hurricane
Extra-tropical cyclone
Polar low
Extra-tropical cyclone
6
Examples of dry and hot extremes
Drought
Wild fire
Dust storm
Dust storm
7
IPCC 2001 definitions

• Simple extremes
• individual local weather variables
• exceeding critical levels on a continuous
• scale
• Complex extremes
• severe weather associated with particular
• climatic phenomena, often requiring
• a critical combination of variables
• Extreme weather event
• An extreme weather event is an event
• that is rare within its statistical reference
• distribution at a particular place.
• Definitions of "rare" vary, but an extreme
• weather event would normally be as
• rare or rarer than the 10th or 90th percentile.
• Extreme climate event

8
Future changes in extremes?
IPCC 2001 Possible scenarios of extremes
9
R and RCLIM
10
R Short intro

• RCLIM make use of R, a powerful statistical tool.
• R is freely available, and can be used on most
  computer platforms
• It is a huge community working with and on R.
• R can be downloaded from
• www.r-project.org

11
RCLIM-initiative

• Part of Workpackage 4.3 ENSEMBLES Understanding
  Extreme Weather and Climate Events
• Progress
• Spring 2005 Initiative started
• March 2006 Delivery finished and methods made
  public
• Future More methods to be included, especially
  for daily datasets.

12
RCLIM-initiative

• Main motivation
• Climate analysis requires increasingly good
  statistical analysis tools.
• Aims
• Develop statistical methods and write user
  friendly functions in the R language for
  describing and exploring weather and climate
  extremes in gridded datasets, making efficient
  use of the already existing packages.
• Webpage
• http//www.met.reading.ac.uk/cag/rclim/

13
RCLIM-initiative

• The RCLIM initiative will develop functions for
• Reading and writing netcdf gridded datasets
• Exploratory climate analysis in gridded datasets
• Climate analysis of extremes in gridded datasets
• Animating and plotting climate analysis of
  gridded datasets
• Team
• David Stephenson, Caio Coelho, Chris Ferro and
  Dag Johan Steinskog

14
Statistical methods
15
European heat wave 2003
Estimated total mortality 35000-50000
This extreme wheather was caused by an
anti-cyclone firmly anchored over the western
European land mass holding back the rain-bearing
depressions that usually enter the continent from
the Atlantic ocean. This situation was
exceptional in the extended length of time (over
20 days) during which it conveyed very hot dry
air up from south of the Mediterranean.
2003 event can be used as an analog of future
summers in coming decades (Beniston, GRL
2004) It is very likely (confidence level gt90)
that human influence has at least doubled the
risk of a heatwave exceeding this threshold
magnitude (Stott et.al., Nature 2004)
Effects on crops, both negative and positive
16
Data used in this presentation

• Monthly mean gridded surface temperature
  (HadCRUT2v)
• 5 degree resolution
• January 1870 to December 2005
• Summer months only June July August
• Grid points with gt50 missing values and SH are
  omitted.
• Special focus on the 2003 summer heat wave in
  Europe

17
Mean temperature
Central Europe (12.5ºE, 47.5ºN)
18
Standard Deviation
Standard Deviation
19
Model for tails peaks-over-threshold
For sufficiently large thresholds, the
distribution of values above a sufficiently large
threshold u approximates the Generalized Pareto
Distribution (GPD)
Shape -0.4 upper cutoff Shape 0.0
exponential tail Shape 10 power law tail
Probability density function
20
Example Central England Temperature

• n 3082 values
• Min -3.1C
• Max 19.7C
• 90th quantile 15.6C

21
GPD fit to values above 15.6C

• Location parameter u15.6C
• Maximum likelihood estimates
• Scale parameter 1.38 /- 0.09C
• Shape parameter -0.30 /- 0.04C
• ? Upper limit estimate

22
1870-2005 time series of summer
(June-July-August) monthly mean temperatures for
a grid point in Central Europe (12.5ºE, 47.5ºN)
2003 exceedance
75th quantile (uy,m 16.2ºC)
Excess (Ty,m uy,m)
15.2ºC
Long term trend (Ly,m)
23
Time varying threshold
JJA pts trendseasonal terms
Excesses
? Flexible approach that gives exceedances 25 of
months
24
Time mean of 75 threshold
25
Mean of the excesses
? Large over extra-tropical land regions
26
GPD scale parameter estimate
? Large over extra-tropical land regions
27
GPD shape parameter estimate
Generally negative ? finite upper temperature
limit
28
Upper limit for excesses
? Largest over high-latitude land regions
29
Return periods for August 2003 event
? Central Europe return period of 133 years (c.f.
Schar et al 46000 years!)
30
The role of large-scale modes
? ENSO effect on temperature extremes in NH
31
Teleconnections between extremes
32
1-point association map for extreme events
? association with extremes in subtropical
Atlantic
33
Future development of RCLIM and methods

• Methods for data with high temporal correlation
  will be introduced (e.g. daily dataset)
• Quantile regression to estimate the thresholds?
• Improve the plotting procedure filled contours
  and projections
• Feedback on other methods that could be included
  is wanted!

34
Conclusions

• Huge potential of doing extremes on gridded
  datasets
• Simple extremes can be analysed using
  peaks-over-threshold methods
• Extremes do not have a unique definition
• Future work include testing the methods on daily
  datasets and develop new methods for data with
  high autocorrelation with special focus on Arctic
  region

35
Reference

• Coelho, C. A. S., C. A. T. Ferro, D. B.
  Stephenson and D. J. Steinskog Exploratory tools
  for the analysis of extreme weather and climate
  events in gridded datasets, Submitted to Journal
  of Climate
• Contact info
• David Stephenson, d.b.stephenson_at_reading.ac.uk
• Dag Johan Steinskog, dag.johan.steinskog_at_nersc.no

36
Thank you for your attention!