UNDERSTANDING, DETECTING AND COMPARING EXTREME PRECIPITATION CHANGES OVER MEDITERRANEAN USING CLIMATE MODELS Dr. Christina Anagnostopoulou Department of Meteorology-Climatology, School of Geology Aristotle University of Thessaloniki Greece
To assess the ability of RCMs datasets to simulate extreme daily precipitation To produce estimates of predicted changes in return levels by future time periods ( and ) Detection of extreme precipitation assuming that model predictions are accurate Aim
Data and methods Results for selected grid points Spatial distribution of the extreme precipitation indices Differences of the extreme precipitation indices between future and reference time period Outline
KNMI Data C4I RCMs data for Mediterranean region Window: 10 o W – 35 o E 31 o N - 45 o N
Description of RCMs used KNMI-RACMO2: Royal Netherlands Meteorological Institute ( KNMI, Lenderink et al., 2003; van den Hurk et al., 2006 ) Parent ECHAM5 Time period SRES A1B Physical parameterizations of ΕCMWF (European Centre for Medium – Range Weather Forecasts) used also for ERA-40 ( Spatial Resolution 25x25km.
Description of RCMs used C4IRCA3 : Community Climate Change Consortium for Ireland (C4I). Parent ECHAM5 Time period SRES A2 RCA3 the third version of the Rossby Centre Atmospheric model (Kjellström et al., 2005) Spatial Resolution 25x25km.
Methodology Geveralized Extreme Value Distribution μ: location parameter σ: scale parameter ξ: shape parameter Return level for ξ 0 for ξ = 0
Estimation for GEV distribution 1. Maximum Likelihood Estimation-MLE 2. Bayesian Method
Methodology Reference period: year period: year period: Indices P m : medianP m (t)=X 0.5 (t) P 20 : 20-year return valueP 20 (t)=X 0.95 (t) P 100 : 100-year return value P 100 (t)=X 0.99 (t)
Western Mediterranean Central Mediterranean Eastern Mediterranean
Maximum Likelihood Estimation-MLE KNMI C4I Eastern Mediterranean Central Mediterranean Western Mediterranean
Bayesian Method Eastern Mediterranean Central Mediterranean Western Mediterranean location scaleshapeReturn level
Spatial distribution of maximum annual precipitation Max Min Mean
Spatial distribution of the extreme precipitation indices KNMI-MLE
KNMI - MLE
Spatial distribution of the extreme precipitation indices C4I - MLE
Spatial distribution of the extreme precipitation indices C4I - MLE
Spatial distribution of the extreme precipitation indices KNMI-Bayes
KNMI - Bayes
Spatial distribution of the extreme precipitation indices C4I - Bayes
Spatial distribution of the extreme precipitation indices C4I - Bayes
Differences of the extreme precipitation indices between the two time period ( & ) and the reference period ( ) KNMI-MLE Differences of the extreme precipitation indices between the two time period ( & ) and the reference period ( ) KNMI-MLE
Differences of the extreme precipitation indices between the time period ( ) and the reference period C4I-MLE
Differences of the extreme precipitation indices between the two time period ( & ) and the reference period ( ) KNMI-Bayes
Differences of the extreme precipitation indices between the time period ( ) and the reference period C4I-Bayes
Concluding remarks The two RCMs datasets simulate reasonably well the extreme annual daily precipitation Pm index presents no change or a slight decrease for the future time period, in Mediterranean region P 20, an index that locates in the tail of the GEV distribution, present increase especially in central Mediterranean The two estimators (MLE and Bayesian) present similar results for the reference period but different for the future time-period. The Bayesian method present a practical advantage.