Collabora've REAnalysis Technical Environment Intercomparison Project CREATE- IP

Transcription

1 Collabora've REAnalysis Technical Environment Intercomparison Project CREATE- IP Gerald L. Po,er1 Tsengdar Lee2 Laura Carriere1 1NASA Goddard Space Flight Center 2NASA Headquarters 1

2 Why do climate model intercomparison? History first to understand differences process studies Feedback and sensiivity Early studies to understand why models had different sensiivity to increasing CO 2 Isolated effect of clouds models sill disagree (25 years later) Why models behave the way they do (devise numerical 11 years sill large range experiments) 2 Courtesy Bob Cess

3 Climate Model Intercomparison 2000 s Volume of data PB TB GB MB 1980 s 1990 s Document model differences AMIP FANGIO Climate predicion CMIP CFMIP Understand feedback mechanisms Understand feedback mechanisms Complexity 3

4 Model&Understanding Cloud&Process Process&Observations Diagnostic&Studies Models Field&Experiments&(FIRE,&GATE) Model (SPOOKIE,&RCE) GCM/CRM/LES Satellites&(&AHtrain,&EarthCARE) Hypothesis Development New&Process Models/&GCM Parameterization Forcing&Scenarios Control,&IPCC&RCPs LGM,&SST,&4X&CO2 Model Model&Understanding Climate&Model Test&Hypothesis: Hypothesis Diagnostic&Studies Predictions Models&vs&Obs Testing CMIP,&PMIP,&CFMIP (Modern&and&Paleo) Prediction Uncertainty& Estimate&(IPCC) Courtesy: Bruce Welicki NASA/LARC 4

5 Recent obs4mips workshop Considerable discussion for expansion Higher frequency Expanded list Golden Year focus Proxy data sets off line simulators Possible more reanalysis variables Co- located data sets Relaxed model equivalent data rules Address process diagnosics In- situ data Include more proxies in the CMIP output experiments Condensed from a list courtesy Robert Ferraro and others 5

6 Reanalysis* for climate model evaluaion and comparison ana4mips IniIated to fill the gaps where direct observaions are not possible. Select variables from reanalyses to correspond with CMIP5 similar to obs4mips Limited to those variables that have model evaluaion potenial (U,V,T,Q,PS) *Reanalysis is a scienific method for developing a comprehensive record of how weather and climate are changing over Ime. In it, observaions and a numerical model that simulates one or more aspects of the Earth system are combined objecively to generate a synthesized esimate of the state of the system. 6

7 CREATE- IP CollaboraIve REanalysis Technical Environment Intercomparison Project Expanded set of variables, tendencies, observaions High frequency data O- A, O- F to address model and observaion biases Gridded observaions used in assimilaion Enablement of a computaional environment Designed to serve as a launch point for reanalysis intercomparison 7

8 Reanalysis intercomparison Courtesy: CoordinaIng Earth observaion data validaion for RE- analysis for CLIMATe ServiceS (Core Climax) Procedure for comparing reanalyses and comparing reanalyses to assimilated observaions and CDRs. 17 July

9 a very high latent heat flux, indicating major problems reanalyses, we also briefly examine the corresponding FIG. 10. The background values of radiation or energy flows (Trenberth et al. 2009) are based The background values of radiaion or energy flows (Trenberth et al. 2009) are on observations for Superposed, with the key (lower left), are values from the various based on ofor bservaions uperposed, with tishe key leo), acoded; re reanalyses the for period exceptsfor ERA-40, which for the(lower 1990s (color 22 values the the various reanalyses or the period xcept for radiation, ERA- 40, and W m f).rom Above graphic, values are fgiven for2albedo (%), ASR,enet TOA OLR; the boxthe labeled SFC near the bottom net flux absorbed atvthe surface. thefor which is for 1990s (color coded; W gives m2). the Above the graphic, alues are gfor iven s the latter value is 0.6 W m. albedo (%), ASR, net TOA radiaion, and OLR; the box labeled SFC near the bo,om gives the net flux absorbed at the surface. For the 1990s the la,er value is 0.6 W m2. From Trenberth et al J. Climate DOI: /2011JCLI

10 June- July- August PrecipitaIon Anomaly for 2010 (based on ) 2010 Russian Heat wave GPCP observaions CFSR ECMWF MERRA 20CR* *anomaly from the reanalysis 10

11 Examples of reanalysis differences 100 Which Reanalysis Depicts the Effects of the 2010 Russian Heat Wave? 2010 Russian Heat wave how well do the reanalyses capture the intense surface drying over Western Russia? CFSR LE CFSR SH ERA-Interim LE ERA-Interim SH Latent Heat Sensible Heat Wm 2 40 Surface drying/heaing in 2 reanalyses Global Precipitation Anomaly MERRA ECMWF 20 Jan Apr Jul Oct Jan Apr Jul Oct Jan Apr Jul Oct Jan Time Global precipitaion anomaly in 2 reanalyses AMSU assimilated into MERRA Year AMSU had a significant impact on MERRA but the ECMWF choose not to assimilate that data because of the impact on precipitaion 11

12 Current Data Holdings Name Source Time Range Assimila'on Resolu'on available from CREATE- IP Dataset Output Times and Time Averaging NASA Modern- era Analysis for Research and (MERRA) NASA GMAO present 3D- VAR, with incremental update 2/3 lon x1/2 degrees; 42 pressure levels Monthly average ECMWF Interim (ERA- Interim) ECMWF present 4D- VAR 0.75x0.75 degrees; 37 pressure levels Monthly average NCEP Climate Forecast System Reanalysis (CFSR) NCEP present 3D- VAR 0.5x0.5 degrees; 22 pressure levels Monthly average Japanese 25- year Reanalysis (JRA- 25) Japan Meteorological Agency (JMA) and Central Research InsItute for Electric Power Industry (CRIEPI) present 3D- VAR 1.25x1.25; 12 pressure levels Monthly average NOAA- CIRES 20th Century Reanalysis (20CR) NOAA/ESRL PSD Ensemble Kalman Filter 2.0x2.0 degrees; 19 pressure levels Monthly average

13 Conclusions Reanalysis in ESGF in similar format Used by modeling and diagnosic community ana4mips Variables not directly observable winds etc. New more comprehensive reanalysis intercomparison- CREATE- IP Help evaluate observaions, perform process studies, help with model biases, idenify causes of climate change in the past, help understand uncertainty 13

14 Problem of evaluaing and using reanalysis requires new thinking Data analysis in the tradiional way won t work Library paradigm must change for large amount of data expected in CREATE- IP project. Dan Duffy s talk Wednesday 14