Model Intercomparison Discussion Fred Kucharski Abdus Salam ICTP
Model Intercomparison Discussion Fred Kucharski (Abdus Salam ICTP, Trieste, Italy) and Adam Scaife (Met. Office, Exeter, UK) as discussion leader
General: We have the plan to write a C 20 C model intercomparison paper. Time series of several important components of the climate variability of the 20 th century should be compared analyzed here. Proposed working title: A suitable title might be something like "Reproducing climate of the past 100 years". Another question is: "How can we use C 20 C type simulations to test models against observations? " This is a very under exploited but important area.
A. What time series did we collect (from Had. ISST forced runs? !): 1. Annual global 2 m land surface air temperature anomalies Area average of all land surface grid boxes. Weight the anomalies from 1961 -90 in each grid box with the cosine of latitude before averaging. 2. Southern Oscillation Tahiti and Darwin (or nearest grid point) monthly mean surface pressure from the model. 3. North Atlantic Oscillation Iceland Azores (or nearest grid point) monthly mean surface pressure values from the models. 4. Sahel rainfall As near as possible to 12. 5 N-17. 5 N, 15 W-37. 5 E, June-September means in mm of total rainfall averaged after weighting with the cosine of latitude of each grid box. 5. Indian Monsoon Rainfall As near as possible to 10 N-30 N, 70 E-95 E, June-September mean rain over land points, averaged after weighting with the cosine of latitude of each grid box. Some Results………
Example 5: Indian Monsoon rainfall Area average JJAS precip 70 E to 95 E, 10 N to 30 N, over land points only Data so far from: METF, UKMO, NASA, MGO-Russia, UMCP, ICTP Interannual: CC(CRU, c 20 c_multim_ensm) = 0. 27 Best: 0. 44 Worst: 0. 04 Decadal (11 -year running mean filter): CC(CRU, c 20 c_multim_ensm) = 0. 80 Best: 0. 82 Worst: 0. 56 We know that results may improve Considerably in pacemaker experiments
Regression of Decadal IMR against SST CRU C 20 C Multi model Ensemble mean As can be seen, on Decadal time scale anticorr between ENSO and IMR is reproduced. But as well Indian Ocean seem to play a role
B. what else 6. PNA index (e. g. 500 h. Pa height, [(15 -25 N, 180 -140 W)- (40 -50 N, 180 -140 W) +(45 -60 N, 125 W-105 W)-(25 -35 N, 90 -70 W)] 7. more series. . . C. Should we collect and compare fields as well? If yes which (e. g. eofs of North Atlantic 500 h. Pa height, etc)? D. What about the pace maker experiments? Would be interesting to collect time series as well from there? Results from Indian rain suggest that they improve considerable the performance of the hindcasts.
E. What kind of analysis Suggestions: 1. Compare time series with observations. 2. Distinguish as well high frequency from low frequency components. 4. Compare standard deviations, etc with observations. 5. Where possible assess potential predictability and compare with actual predictability. 6. Regress against SSTs to identify potential forcings. 7. Taylor Diagrams (= potential + actual predictability + amplitudes) F. Anything else?
- Slides: 7