Statistical Analyses of Extremes from a Regional Climate Model

1
Statistical Analyses of Extremes from a Regional
Climate Model
Chris Ferro Climate Analysis Group Department of
Meteorology University of Reading
Royal Meteorological Society, London, 21 January
2004
2
Overview

• Regional climate-change experiment
• Application of extreme-value theory
• Daily maximum temperature extremes
• Seasonality
• Clustering
• Apparent temperature
• Concluding remarks

3
Regional Modelling Experiment

• PRUDENCE
• 10 high-res. RCMs nested in global GCM
• 30-year control simulation, 1961-1990
• 30-year A2 scenario simulation, 2071-2100

4
Extreme-value Theory

• Aim quantify extremal behaviour
• Problems limited data, extrapolation
• Solution exploit statistical regularities
• Example

5
Seasonality London grid box
6
Seasonality statistical model
Estimate threshold quantile regression Excess
distribution generalised Pareto Estimate
parameters maximum likelihood
Davison and Smith (1990) J. Royal Statistical
Soc. B, 52, 393442
7
Seasonality London estimates
Scale (ese)
1.27 (0.1)
1.44 (0.2)
Shape (ese)
-0.11 (0.04)
-0.01 (0.07)
8
Seasonality London estimates
Scale (ese)
1.27 (0.1)
1.44 (0.2)
Shape (ese)
-0.11 (0.04)
-0.01 (0.07)
9
Times of annual maxima Europe
Control
Scenario Control
day of year
days
10
Clustering London grid box
11
Clustering London results
Mean cluster size (days) 90 confidence interval (days)
Control 3.2 (2.4, 3.9)
Scenario 4.0 (3.3, 4.7)
Ferro and Segers (2003) J. Royal Statistical Soc.
B, 65, 545556
12
Mean Cluster Size Europe
Control
Scenario / Control
days
13
Apparent Temperature London
14
Apparent Temperature model

• Univariate distributions GEV model for tails
• Dependence structure nonparametric estimate

Steadman (1984) J. Climate Applied Met., 23,
16741687
de Haan and Sinha (1999) The Annals of
Statistics, 27, 732759
15
Apparent Temperature results
16
Review

• Many applications of extreme-value theory
• Individual values (e.g. seasonality)
• Clusters (e.g. warm spells)
• Combinations (e.g. temp. and humidity)
• Preliminary Tmax analysis (London)
• Shifted annual cycle
• Longer warm spells
• Greater heat stress

17
Further Information

• PRUDENCE
• Climate Analysis
• Group
• E-mail address

• prudence.dmi.dk
• www.met.rdg.ac.uk/cag
• ? /extremes
• c.a.t.ferro_at_reading.ac.uk