Exploring the Possibility to Forecast Annual Mean Temperature with IPCC and AMIP Runs Peitao Peng Arun Kumar CPC/NCEP/NWS/NOAA Acknowledgements: Bhaskar Jha and Yun Fan
Background: Warming Temperatures Annual Mean/Global Mean Land Temperature (Hoerling et al, 2008)
Background: Warming Temperatures - GISS - GHCN (NCDC) - Reynolds merged land/sea - Reanalysis - CRU Hoerling et al., 2008: Reanalysis of Historical Climate Data for Key Atmospheric Features: Implications for Attribution of Causes of Observed Change. CCSP SAP1.3
Background: Warming Temperatures
Background: How is Warming Trend Affecting S/I Predictions? ANBtotal A N B total OBS Forecast Contingency Table for a performance evaluation of CPC 1-month lead seasonal temp forecast ; 168 seasons over period and 232 grid points over CONUS give cases. About 50% are above normal in OBS, 60% of them are correctly forecasted About 70% skill are from warm cases.
Background: How is Warming Trend Affecting S/I Predictions? Cai et al. 2009: The role of long-term trend in seasonal predictions: Implications of global warming in the NCEP CFS. Weather and Forecasting Fixed GHG content of 1988 Cold Bias due to the underestimated GHG
Objectives Use available model simulations to understand the predictability and prediction skill of land temperatures; Examine the predictability and prediction skill of trends in land temperature;
Procedures of the Study 1.Examine the performance of IPCC and AMIP data in simulating land temperature variations, in particular the warming trend; 2.Make empirical hindcasts with IPCC and AMIP data; 3.Compare the model data based empirical hindcasts with that based on analysis data.
Data Sets –AR4 CMIP3 Climate of 20 th Century (with the observed evolution of external forcing; ) Special Emission Scenario (SRES) A1B ( ) 48 simulations from 22 different models –AMIP Forced with SSTs Ensemble of simulations Three AGCMs ( ) –OBS CPC Merged GHCN-CAMS analysis (Fan and Huug) (1949-current) Data of their common period ( ) are used for our analysis
Empirical Methods in Hindcast: Persistent: one year persistent extension of bias corrected model data; Sloped Extension: make a linear regression over the latest n year model data, then do a sloped extension to the target year. Flat Extension (OCN): take the latest n (n>1) year average as the forecast for the target year Note: 1.Flat extension is only applied to analysis data and the result will be used as a bench mark to evaluate other methods applied to model data. 2.Optimal n corresponds to the highest forecast skill over the verification period
Annual and Global Mean Land Temperatures: OBS vs Models Total Quantities Anomalies w.r.t climate
Simulation Skill of CIMP3 and AMIP for Land Temp Indices ACC RMS AMIP is superior to CMIP3 in simulation
Annual and Global Mean Land Temperature Forecasts with CMIP3 CMIP3 based forecast has the almost the same skill as the analysis based OCN
Annual and Global Mean Land Temperature Forecast with AMIP The skill of AMIP based forecast is lower than the OCN
Forecast Skill Comparison ACC RMS
Simulation Skill: CIMP vs AMIP CMIP3AMIP
Forecast Skill: CMIP3 Persistent vs OCN CMIP3 PersistentOCN
Forecast Skill: Sloped Extension of CMIP3 vs OCN Sloped Ext of CMIP3OCN
Summary and Future Work Warming trend is an important source of CPC S/I forecast skill in last couple of decades; CMIP3 well catches the warming trend of land temperature in observations; Empirical extension of CMIP3 data has a potential to provide compatible or better forecast than the flat extension of observational data (OCN) for annual mean temperature. AMIP runs are generally better than CMIP3 runs in simulation, but their empirical extension is not as good as that of CMIP3 in forecast due to bigger “noise”. The analysis will be extended to seasonal mean and longer lead forecast, and a further study will be toward a statistical-dynamical tool to project warming trend and other LF components onto S/I forecast.
Optimal Window length
Background: How is Warming Trend Affecting SI Predictions? Official CPC Sfc. Temp Forecasts