1. Reconstruction of highly resolved atmospheric forcing fields for Northern Europe since 1850 AD Frederik Schenk & Eduardo Zorita EMS Annual Meeting & European Conference on EMS Annual Meeting & European Conference on Applied Climatology, 16-09-2010

2. PAGE 2 Working Packages ANALOG RECONSTRUCTION 1850 - 2009 1957 - 2007

3. PAGE 3 Motivation I. Climatic aspects  extention back to 1850: „Rebound from Little Ice Age“  long period prior to large human impact II. Methodical aspects  high spatial (0.25°) and temporal (daily) resolution realistic variability + extremes  realistic variability + extremes  linear regression << 50% variability (von Storch et al. 2004)‏  non-linear approach  Analog-Method von Storch et al. (2004): Reconstructing Past Climate from Noisy Data. Science, Vol. 306, No. 5696, pp. 679-682.

4. PAGE 4 Outline I. Reconstruction Method II. Data III. Reconstruction Skills IV. Take Home

5. The Analog-Method as upscaling tool 1) Generate consistent Analog fields = numerical downscaling 2) Find Analog fields for station data = statistical upscaling

6. PAGE 6

7. PAGE 7 … … Analog-Method: find for : Reconstruction ANALOGS PREDICTOR (station data)‏ 1850 2009 1957-01-012007-11-30

8. PAGE 8 Settings Test: Cross-wise cal/val for 25 years  Predictand = daily analog fields from RCAO model  Predictor = daily SLP (N=23 stations)‏ Increase sample size for analogs: analogs in M {m-1, m, m+1}  days of month m  analogs in M {m-1, m, m+1}  allows seasonal shifts if forced by predictor Reconstruction: cal. 1958-2007  1850-2009

9. Data Atmospheric Fields (Analogs)‏ Station Data

10. PAGE 10 Analogs of Atmospheric Fields Atmospheric Fields for: -Sea-Level-Pressure [Pa] -U- and V-Wind [m/s] -Relativ Humidity [%] -Total Cloud Cover [%] -Precipitation [mm] -Temperature [K] Source RCAO Swedish Regional Climate Model with Coupled Ocean-Model for Baltic Sea

11. PAGE 11 Daily SLP Station Data EMULATE EMULATE Mean Sea Level Pressure data set (EMSLP)  provides ~ 20 stations for RCAO-domain 1850 - 2002  partly covers 1850 - 2002, updates by WMO, ECA&D Ansell, T. J. et al. (2006) Daily mean sea level pressure reconstructions for the European - North Atlantic region for the period 1850-2003', Journal of Climate, vol 19, No. 12, pp 2717-2742.

12. Reconstruction Skills Field Correlation Ratio of Variance

13. PAGE 13 Reconstruction Skills WS Taylor Diagram JAN 1958-2007 Taylor Diagram JUL 1958-2007

14. Reconstruction Skills for Wind Speed Field correlation & Variance Wind speed distribution 99% treshhold values

15. PAGE 15 SLP-Reconstruction Fieldcor for JAN 1958-1983 Fieldcor for JUN 1958-1983 Calibration: 1984-2007

16. PAGE 16 U-Wind-Reconstruction Fieldcor for JAN 1958-1983 Fieldcor for JUN 1958-1983 Calibration: 1984-2007

17. PAGE 17 V-Wind-Reconstruction Fieldcor for JAN 1958-1983 Fieldcor for JUN 1958-1983 Calibration: 1984-2007

18. PAGE 18

19. PAGE 19 Wind speed distribution number of events January (N = 1550)‏

20. PAGE 20 99% treshold for wind speed [m/s] Deviation of 99% treshold [m/s] 99 Percentile of wind speed

21. PAGE 21 99 Percentiles of Wind Speed 99% treshold values for daily wind speed for JANUARY (1958-2007)‏ RCAO RECONSTRUCTION

22. PAGE 22 99 Percentiles of Wind Speed Deviation of 99% treshold values for daily wind speed (REC – RCAO)‏

23. PAGE 23 Historical Storm Surges? W/SW-Baltic Sea 12/13 th of November 1872  Before: Storm from SW, filling up from North Sea  12/13th: increase to 120 km/h, change to NE (+3,5 m)‏ The big Hamburg Flood 16/17 th of February 1962  Extreme sustained wind speeds from NW (180 km/h)‏

24. PAGE 24 1872-11-12 1872-11-13

25. PAGE 25 1962-02-16 1962-02-17

26. PAGE 26 Take Home I. Analog-Method as statistical upscaling tool  Needs analog-fields and (historical) station data II. Reconstructed atmosph. Forcing fields  Realistic daily statistics including extremes  Good correlations for all variables on monthly scale  There is enough data also for other regions…

27. Thank you for your attention!

28. Temperature Reconstruction Needs additional care…

29. PAGE 29 Temperature Reconstruction SLP has a too weak physical link to T2m  additional predictor T2m (N=22 stations, monthly)‏  reconstruction of monthly T2m fields  projected onto daily T2m anomalies reconstructed by daily SLP  includes also T2m changes NOT affected by SLPTime-invariance:  Much too low daily auto-correlation for daily T2m  Memory effect can be added (similarity over n days)‏

30. PAGE 30 Daily T2m variability

31. PAGE 31 Daily Auto-Correlation

32. PAGE 32 Memory Effect over n Days

33. PAGE 33 Temperature Reconstruction SLP + T2m SLP only

34. PAGE 34 Analog-Method GCMsimulatedlarge-scalepatterns related to simultaneously observed local weather related to RCA simulatedlarge-scalepatterns observed station data of local weather Large-ScalePatterns Local weather observations Sample of Analogs Stat. Downscaling Stat. Upscaling PREDICTOR PREDICTAND PREDICTAND PREDICTOR

35. PAGE 35 Analog-Method @ GKSS The Analog Method as a simple statistical Downscaling Technique: Comparison with more complicated Methods  Zorita & v. Storch (1999), J. of Clim. 30: 133-144  AM most simple method  serves best Improved field reconstruction with the analog method: searching the CCA space  Fernández & Sáenz (2003), Clim. Res. 24: 199-213  CCA better correlation, EOF/Analog better variance

36. PAGE 36 Analog-Method @ GKSS Influence of similarity measures on the performance of the analog method downscaling daily precipitation  Matulla et al. (2008), Clim. Dyn. 30: 133-144  Most simple method (Euclidian distance) serves best Reconstructing highly resolved atmospheric forcing fields using analog method for statistical upscaling  Schenk & Zorita (2010), in prep.  AM can be used also as simple non-linear upscaling tool