Evaluation of the Short-Term Climate Variability in GEOS-5 ...€¦ · Evaluation of the Short-Term...
1
Evaluation of the Short-Term Climate Variability in GEOS-5 AOGCM Simulations Yury Vikhliaev 1,2 , Max Suarez 1 , Andrea Molod 1,4 , Bin Zhao 1,3 , Michele Rienecker 1 , Siegfried Schubert 1 1. NASA Global Modeling and Assimilation Office; 2. Universities of Space Research Association; 3. Science Applications International Corporation; 4. Earth System Science Interdisciplinary Center [email protected] 1. Introduction The GEOS-5 Atmosphere-Ocean General Circu- lation Model (AOGCM) has been developed for subseasonal-to-decadal climate prediction studies. Its main components are the GEOS-5 atmospheric model, the catchment land surface model, and MOM4, the ocean model developed by the Geo- physical Fluid Dynamics Laboratory. The ocean and atmosphere exchange fluxes of momentum, heat and fresh water through a ”skin layer” inter- face that includes parameterization of the diurnal cycle and a sea ice model, LANL CICE (fig. 1). All components are coupled together using the Earth System Modeling Framework (ESMF). The model has been tested in coupled simulations and data assimilation mode and is able to produce a stable, realistic mean climate and inter-annual climate vari- ability. An evaluation of the model performance in climate simulations is presented below. Figure 1: GEOS-5 AOGCM coupling configuration. 2. Experiment design To assess the performance of the GEOS-5 AOGCM, a coupled simulation of 90 years’ duration was analyzed. The configuration of the validation simulation is as follows: • AGCM 2.5 0 horizontal resolution with 72 vertical levels up to 0.01hPa; • OGCM 1 0 horizontal resolution, with meridional refinement to 1/3 0 in the equatorial region, and with 50 vertical levels; • AGCM component is initialized from an AMIP run; • OGCM initial conditions are taken from a test run initialized with observed temperature and salinity, and spun up for 50 years; • perpetual 1950 boundary conditions. 3. Mean climate 3.1 Comparison with AMIP simulations The atmospheric climate performance in the cou- pled model is compared with that from AMIP sim- ulations to isolate coupling issues. Both are com- pared with GMAO’s atmospheric reanalysis for the satellite era - the Modern-Era Retrospective analy- sis for Research and Applications (MERRA) Figure 2: JJA precipitation from coupled (left col- umn) and AMIP (right column) simulations. The top row shows the GEOS-5 simulation, the middle row shows the climatology from MERRA, and the bot- tom row shows the difference between the top and middle plots. The mean atmospheric climate simulated by the AOGCM is similar to that simulated by the GEOS-5 AGCM forced by the observed SST. As an example, figure 2 shows comparison of the JJA precipitation climatology in coupled and AMIP simulations. 3.2 Surface climatology The surface climate reaches equilibrium in several decades. Figure 3 shows the model SST and SSS, along with the difference from observations. The dominant errors shown in figure 3 are typical for cur- rent climate models. Figure 3: SST (top-left) and SSS (top-right) from the GEOS-5 AOGCM, and the difference from Reynolds SST (bottom-left) and Levitus SSS (bottom-right) climatologies. Figure 4: Annual Mean Equatorial SST from GEOS-5 and Reynolds. Figure 5: Equatorial annual cycle of SST from the GEOS-5 AOGCM (left) and Reynolds (right) with the annual mean removed. Figures 4 and 5 show the Pacific equatorial pro- files of SST from the validation run and Reynolds. The warm bias in the eastern equatorial Pacific and phase shift in the annual cycle of equatorial SST are typical for current coupled models. 4. Climate variability To assess the ability of the GEOS-5 AOGCM to pre- dict climate on seasonal and decadal time scales, short term variability in validation run was analyzed. In addition to observed climatologies, a comparison is made with MERRA. 4.1 El-Ni ˜ no Southern Oscillation ENSO variability is examined using the leading EOF mode of the global SST. Figure 6: Top: the leading principal components of the global SST from GEOS-5 and the Hadley Cen- tre data set. Bottom: the leading EOF patterns for GEOS-5 SST (left) and HadSST (right). The model produces a prominent ENSO signal with higher frequency and narrower pattern extended further into the western Pacific compared to ob- served ENSO (fig. 6). 4.2 Pacific North American teleconnec- tion The PNA index is defined as Z500(20 0 N,160 0 W)- Z500(45 0 N,165 0 W)+ Z500(55 0 N,115 0 W)- Z500(30 0 N,- 85 0 W), where Z500 is 500mbar geopotential height anomaly. Figure 7: Top: the PNA indexes from GEOS-5 and MERRA. Middle and bottom: composites of the GEOS-5 (left) and MERRA (right) 500mbar geopo- tential height anomaly based on the PNA indexes. The model produces a realistic PNA signal with variability on sub-seasonal to interannual time scales. The PNA pattern has positive centers in the vicinity of Hawaii and over the intermountain region of North America, and negative centers located south of the Aleutian Islands and over the south- eastern United States during the positive phase (fig. 7). 4.3 North Atlantic Oscillation Figure 8: Top: the NAO indexes from GEOS-5 and MERRA. Middle and bottom: composites of the GEOS-5 (left) and MERRA (right) SLP anomaly based on the NAO indexes. The NAO index is defined as difference of SLP anomaly at (27 0 W,34 0 N) and (32 0 W,60 0 N). The model NAO has realistic dipole pattern with the negative center over Greenland and positive cen- ter over the Atlantic ocean between 35 0 N and 40 0 N during the positive phase (fig. 8). 4.4 Pacific Decadal Oscillation The PDO index is defined as the leading princi- pal component of the North Pacific SST between (120 0 E,100 0 W) and (20 0 N,60 0 N). Figure 9: Top: the leading principal components of the North Pacific SST from GEOS-5 and the Hadley Centre data set. Middle and bottom: composites of the GEOS-5 (left) and Hadley (right) SST anomaly based on the indexes from the top panel. The PDO signal is a combination of interannual and decadal modes. The model PDO has a realistic pat- tern in the extra-tropics, but does not have a tropical imprint as in the observed PDO (fig. 9). 5. Summary The GEOS-5 AOGCM produces a stable, realis- tic mean climate and realistic representations of the major modes of short term climate variability. Some biases are currently being diagnosed, in- cluding a significant drift of the Atlantic Meridional Overturning Circulation (AMOC), which is accom- panied by cooling of the North Atlantic ocean and expansion of winter sea ice margin. The model is used by GMAO for ocean data assimilation and seasonal forecasts. An ensemble of decadal pre- diction runs have been recently undertaken by the GMAO following the experimental suite specified for the Coupled Model Intercomparison Project phase 5 (CMIP5). References www.geos5.org World Climate Research Programme conference, 24-28 October 2011, Denver, CO, USA
Transcript of Evaluation of the Short-Term Climate Variability in GEOS-5 ...€¦ · Evaluation of the Short-Term...
Evaluation of the Short-Term Climate Variability in GEOS-5AOGCM Simulations
Yury Vikhliaev1,2, Max Suarez1, Andrea Molod1,4, Bin Zhao1,3, Michele Rienecker1, Siegfried Schubert1
1. NASA Global Modeling and Assimilation Office; 2. Universities of Space Research Association; 3. Science Applications International Corporation; 4. EarthSystem Science Interdisciplinary Center
1. Introduction
The GEOS-5 Atmosphere-Ocean General Circu-lation Model (AOGCM) has been developed forsubseasonal-to-decadal climate prediction studies.Its main components are the GEOS-5 atmosphericmodel, the catchment land surface model, andMOM4, the ocean model developed by the Geo-physical Fluid Dynamics Laboratory. The oceanand atmosphere exchange fluxes of momentum,heat and fresh water through a ”skin layer” inter-face that includes parameterization of the diurnalcycle and a sea ice model, LANL CICE (fig. 1). Allcomponents are coupled together using the EarthSystem Modeling Framework (ESMF). The modelhas been tested in coupled simulations and dataassimilation mode and is able to produce a stable,realistic mean climate and inter-annual climate vari-ability. An evaluation of the model performance inclimate simulations is presented below.
Figure 1: GEOS-5 AOGCM coupling configuration.
2. Experiment design
To assess the performance of the GEOS-5AOGCM, a coupled simulation of 90 years’ durationwas analyzed. The configuration of the validationsimulation is as follows:
• AGCM 2.50 horizontal resolution with 72 verticallevels up to 0.01hPa;
• OGCM 10 horizontal resolution, with meridionalrefinement to 1/30 in the equatorial region, andwith 50 vertical levels;
• AGCM component is initialized from an AMIP run;• OGCM initial conditions are taken from a test run
initialized with observed temperature and salinity,and spun up for 50 years;
• perpetual 1950 boundary conditions.
3. Mean climate
3.1 Comparison with AMIP simulationsThe atmospheric climate performance in the cou-pled model is compared with that from AMIP sim-ulations to isolate coupling issues. Both are com-pared with GMAO’s atmospheric reanalysis for thesatellite era - the Modern-Era Retrospective analy-sis for Research and Applications (MERRA)
Figure 2: JJA precipitation from coupled (left col-umn) and AMIP (right column) simulations. The toprow shows the GEOS-5 simulation, the middle rowshows the climatology from MERRA, and the bot-tom row shows the difference between the top andmiddle plots.
The mean atmospheric climate simulated by theAOGCM is similar to that simulated by the GEOS-5AGCM forced by the observed SST. As an example,figure 2 shows comparison of the JJA precipitationclimatology in coupled and AMIP simulations.
3.2 Surface climatologyThe surface climate reaches equilibrium in severaldecades. Figure 3 shows the model SST and SSS,along with the difference from observations. Thedominant errors shown in figure 3 are typical for cur-rent climate models.
Figure 3: SST (top-left) and SSS (top-right)from the GEOS-5 AOGCM, and the differencefrom Reynolds SST (bottom-left) and Levitus SSS(bottom-right) climatologies.
Figure 4: Annual MeanEquatorial SST fromGEOS-5 and Reynolds.
Figure 5: Equatorial annual cycle of SST from theGEOS-5 AOGCM (left) and Reynolds (right) withthe annual mean removed.
Figures 4 and 5 show the Pacific equatorial pro-files of SST from the validation run and Reynolds.The warm bias in the eastern equatorial Pacific andphase shift in the annual cycle of equatorial SSTare typical for current coupled models.
4. Climate variability
To assess the ability of the GEOS-5 AOGCM to pre-dict climate on seasonal and decadal time scales,short term variability in validation run was analyzed.In addition to observed climatologies, a comparisonis made with MERRA.
4.1 El-Nino Southern OscillationENSO variability is examined using the leadingEOF mode of the global SST.
Figure 6: Top: the leading principal components ofthe global SST from GEOS-5 and the Hadley Cen-tre data set. Bottom: the leading EOF patterns forGEOS-5 SST (left) and HadSST (right).
The model produces a prominent ENSO signal withhigher frequency and narrower pattern extendedfurther into the western Pacific compared to ob-served ENSO (fig. 6).
4.2 Pacific North American teleconnec-tionThe PNA index is defined as Z500(200N,1600W)-Z500(450N,1650W)+ Z500(550N,1150W)- Z500(300N,-850W), where Z500 is 500mbar geopotential heightanomaly.
Figure 7: Top: the PNA indexes from GEOS-5 andMERRA. Middle and bottom: composites of theGEOS-5 (left) and MERRA (right) 500mbar geopo-tential height anomaly based on the PNA indexes.
The model produces a realistic PNA signal withvariability on sub-seasonal to interannual timescales. The PNA pattern has positive centers in thevicinity of Hawaii and over the intermountain regionof North America, and negative centers locatedsouth of the Aleutian Islands and over the south-eastern United States during the positive phase (fig.7).
4.3 North Atlantic Oscillation
Figure 8: Top: the NAO indexes from GEOS-5and MERRA. Middle and bottom: composites ofthe GEOS-5 (left) and MERRA (right) SLP anomalybased on the NAO indexes.
The NAO index is defined as difference of SLPanomaly at (270W,340N) and (320W,600N). Themodel NAO has realistic dipole pattern with thenegative center over Greenland and positive cen-ter over the Atlantic ocean between 350N and 400Nduring the positive phase (fig. 8).
4.4 Pacific Decadal OscillationThe PDO index is defined as the leading princi-pal component of the North Pacific SST between(1200E,1000W) and (200N,600N).
Figure 9: Top: the leading principal components ofthe North Pacific SST from GEOS-5 and the HadleyCentre data set. Middle and bottom: composites ofthe GEOS-5 (left) and Hadley (right) SST anomalybased on the indexes from the top panel.
The PDO signal is a combination of interannual anddecadal modes. The model PDO has a realistic pat-tern in the extra-tropics, but does not have a tropicalimprint as in the observed PDO (fig. 9).
5. Summary
The GEOS-5 AOGCM produces a stable, realis-tic mean climate and realistic representations ofthe major modes of short term climate variability.Some biases are currently being diagnosed, in-cluding a significant drift of the Atlantic MeridionalOverturning Circulation (AMOC), which is accom-panied by cooling of the North Atlantic ocean andexpansion of winter sea ice margin. The modelis used by GMAO for ocean data assimilation andseasonal forecasts. An ensemble of decadal pre-diction runs have been recently undertaken by theGMAO following the experimental suite specified forthe Coupled Model Intercomparison Project phase5 (CMIP5).
Referenceswww.geos5.org
World Climate Research Programme conference, 24-28 October 2011, Denver, CO, USA
