Olivier Mestre Météo-France France Ingebor Auer ZAMG Austria Enric Aguilar U. Rovirat i Virgili Spain Paul Dell

1. ADVANCES IN HOMOGENISATION METHODS OF CLIMATE SERIES: AN INTEGRATED APPROACH Olivier Mestre Météo-France France Ingebor Auer ZAMG Austria Enric Aguilar U. Rovirat i Virgili Spain Paul Della-Marta MeteoSwiss Switzerland Vesselin Alexandrov NIMH Bulgaria Sylvie Jourdain Météo-France France COST OC-2006-32

2. "HOME"

3. Juin 1912 PAU-LESCAR (EN) 2005 PAU-UZEIN (AERO) Homogenisation : WHY?Exemple of PAU-UZEIN temperatures COST OC-2006-32

4. Station Silistra COST OC-2006-32

5. Station Dobrich COST OC-2006-32

6. Anomalies of annual minimum temperature in Kjustendil (raw data) COST OC-2006-32

7. Raw series of annual Maximum Temperatures (TX) PAU-UZEIN COST OC-2006-32

8. « before » « after » Homogenisation of Pau Maximum Temperatures COST OC-2006-32

9. Decomposition of climatological series COST OC-2006-32

10. Usual method: relative homogeneity PRINCIPLE : removing the climatic signal to put into evidence artificial shifts in the series Tested series minus COMPARISON series Reference series COST OC-2006-32

11. Two problems • DETECTION OF SHIFTS • CORRECTION OF SHIFTS COST OC-2006-32

12. Many solutions ! • DETECTION Visual, Craddock test, Student t-test, Likelihood ratio test (SNHT), Potter test, Bayesian procedures, Local contrast test, Pettitt test, penalized likelihood, MASH… • CORRECTION Composite reference series, interpolated reference series, multiple non-homogeneous series (MASH), ANOVA (Mestre)… COST OC-2006-32

13. Statistical problems • Related to detection power and level of the procedures, which are the best ones? • Related to correction how to ensure unbiased corrections? COST OC-2006-32

14. Practical problems • Data requirements, depending on parameters: spatial correlation • Nature of correction: raw monthly estimated coefficients, annual coefficients? • What about close shifts? • What about gaps? • Trust metadata or statistics? • Trust relative homogenisation or parallel measurements? COST OC-2006-32

15. Why COST? • Many procedures, many authors, that might lead to different results • Few intercomparisons COST is an ideal platform for the exchange of experiences, harmonisation of approaches and the development of joint methods COST OC-2006-32

16. Objectives To achieve a general method for homogenising climate and environmental datasets • Provide practical rules for the implementation of homogenisation. • Provide tools for comparison and evaluation of different methods. • Analyse the strengths and weaknesses of the methods for different applications. • Provide methods for evaluating uncertainties resulting from homogenisation. COST OC-2006-32

17. Scientific Program • Inventory of existing detection and correction methods • Compilation of a benchmark dataset to be used across the Action • Selection, comparison and evaluation of existing detection methods • Selection, comparison and evaluation of existing correction methods • Documentation of practical recommendations • Selection, comparison and evaluation of existing correction methods for daily data • Presentation and release of the new common method COST OC-2006-32

18. Is this ambitious? YES COST OC-2006-32

19. Why ? • Because we want to make a synthesis of the most appropriate methods, based on objective comparisons and sharing of practical experiences • Because we feel that the corresponding software has to be released – just editing recommandations would be useless COST OC-2006-32

20. Is this TOO Ambitious? NO COST OC-2006-32

21. Why ? • We have a clear view of what has to be done • Benchmark dataset conception (simulated and real cases) • Coding and testing the procedures is rather easy, and may be done by students • To test real cases and edit practical recommandations, we get together highly trained climatologists COST OC-2006-32

22. Structure of the proposal • Working groups Benchmark Report WG3 Correction WG5 Implement Report Method Final report Software Benchmark Report WG2 Detection WG1 inventory Report Benchmark WG4 Daily values Report COST OC-2006-32

23. Dissemination plan • Mailing list: costh@meteo.fr • Web site: domain name « homogenisation.org » is available, with free software and documentation • Publications • Workshops • Training sessions (ENM) COST OC-2006-32

24. Benefits • Standardisation of homogenisation procedures in Europe • Results of climate studies based on homogenised series could be easily compared • Provide a benchmark dataset that can be used to test future methods • Significant advances in daily data homogenisation • All this will result in an increased confidence in European (and global) assessments of mean and extreme Climate Change COST OC-2006-32

25. Benefits • Links to European projects : CIRCE, MACE, ECA&D… • Economical impacts: risk assessment (extremes: insurance companies) • Calibration of weather derivatives COST OC-2006-32

26. Participants • From 17 countries • Around 60 participants • 60% plan to participate actively to the Action • 40% are mainly interested by the results • 45% may have students working on the Action COST OC-2006-32

27. Conclusion • Many procedures • Few intercomparisons • Daily data correction has to be developed  Need of a COST concerted Action COST OC-2006-32