Interactive comment on “ Evaluation of a present-day climate simulation with a new coupled atmosphere-ocean model GENMOM ”

We have used both the spirit of the reviewer’s comments and our own critical review to revise the manuscript as a whole so that it flows better and reads a bit more logically. The overarching recommendations from both reviewers are to reduce the number of figures and to focus more on evaluation of GENMOM and less on comparison with other AOGCMs. We agree with these recommendations and we have accommodated them by: 1) including only one figure in which we compare GENMOM surface temperature and precipitation with NCEP and three other AOGCMs (as opposed to eight).


Introduction
We present a new non-flux corrected coupled atmosphere-ocean general circulation model (AOGCM), GENMOM, which combines GENESIS version 3 (Global ENvironmental and Ecological Simulation of Interactive Systems) and MOM2 (Modular Ocean Model version 2) general circulation models.Both models have been used widely in climate studies that demonstrate their overall ability to produce climate simulations that are in agreement both with observations and with similar models.GENESIS version 1 was developed starting in 1989 at the National Center for Atmospheric Research (NCAR) with a focus on linking terrestrial physical and biophysical processes with the atmosphere to provide a model that could be applied to investigate paleoclimate and possible future climates under global warming.Introduction

Conclusions References
Tables Figures

Back Close
Full GENESIS version 1 was released in 1991 (Thompson and Pollard, 1995) and included a land-surface transfer model (LSX) and an atmospheric general circulation model derived from NCAR CCM1.GENESIS version 2 was released in 1995 and included many improvements ranging from new prognostic cloud amounts, the use of hybrid vertical coordinates, the inclusion of gravity wave drag, and to improvements in LSX (Thompson and Pollard, 1997;Pollard and Thompson, 1997).
GENESIS version 3 expands on version 2 by including the NCAR CCM3 radiation code and the ocean can optionally be represented by the MOM2 ocean general circulation model in addition to fixed sea surface temperatures or the slab ocean.MOM also has a long history of use and development spanning back to the early 1990s and is used as the ocean component in many other AOGCMs (Pacanowski, 1996).Our current version of GENMOM uses T31 (∼3.75 • × 3.75 • latitude and longitude) horizontal resolution for both the atmosphere and ocean to balance computational requirements needed for long simulations with the ability to simulate important features of the general circulation.
We evaluate a simulation of modern climate against observations and other coupled AOGCMS.The evaluation demonstrates that GENMOM produces a realistic simulation of modern climate that is on par with that of the models used in the Intergovernmental Panel of Climate Change (IPCC) Forth Assessment Report (AR4).For additional evaluation, we compare GENMOM to eight models evaluated in the World Climate Research Programme's (WCRP's) Coupled Model Intercomparison Project phase 3 (CMIP3, Meehl et al., 2007a), a multi-model dataset that was subsequently used in the IPCC AR4 (described in Table 1, Randall et al., 2007).A full description of GENESIS and MOM2 as well as their coupling is provided in Sect.2, atmospheric and oceanic results from a modern climate simulation are presented in Sect.
The nominal GENESIS resolution is spectral T31 (3.75 • × 3.75 • ) with 18 vertical sigma coordinate levels, 4 of which are above the tropopause.Spectral transform dynamics are used for mass, heat and momentum (Williamson et al., 1987).A semi-Lagrangian transport in grid space is used for water vapor (Williamson and Rasch, 1989).Convection in the free atmosphere and in the planetary boundary layer is treated using an explicit sub-grid buoyant plume model similar to, but simpler than, Kreitzberg and Perkey (1976) and Anthes (1977, Sect. 4).Stratus, convective and anvil cirrus clouds are predicted using prognostic 3-D water cloud amounts, (Smith, 1990;Senior and Mitchell, 1993) and clouds are advected by semi-Lagrangian transport and mixed vertically by convective plumes and background diffusion.
The land-surface transfer model, LSX, accounts for the physical effects of vegetation (Pollard and Thompson, 1995).Up to two vegetation layers (trees and grass) can be specified at each grid point, and the radiative and turbulent fluxes through these layers to the soil or snow surface are calculated.A six-layer soil model extends from the surface to 4.25 m depth, with layer thicknesses increasing from 5 cm at the top to 2.5 m Figures

Back Close
Full freezing and thawing of soil ice.A three-layer snow model, which includes fractional area cover when the snow is thin, is used for snow cover on soil, ice-sheet and sea-ice surfaces.A three-layer sea-ice model accounts for local melting, freezing, fractional sea-ice cover (Semtner, 1976;Harvey, 1988), and includes dynamics associated with wind and ocean current using the cavitating-fluid model of Flato and Hibler (1992).

MOM2 description
MOM2 was developed by the Geophysical Fluid Dynamics Laboratory (GFDL) in the early 1990s, but builds off previous work that began back in 1969 (Pacanowski, 1996).MOM2 is a finite difference implementation of the primitive equations of ocean circulation based on the Navier-Stokes equations with the Boussinesq, hydrostatic, and rigid lid approximations (Bryan, 1969).The Boussinesq approximation invokes constant density with depth, with the exception of terms that contain gravity, thereby reducing computational complexity.The hydrostatic approximation assumes that vertical pressure gradients are density driven.A nonlinear equation of state couples temperature and salinity to fluid velocity.An insulated lateral boundary is used such that no temperature or salinity flux is exchanged between ocean and land cells.Unlike the sigma levels used for atmospheric altitude in GENESIS, MOM uses a fixed z-axis for depth, which simplifies the equations used in the finite difference representation.Our version of MOM2 uses 20 unevenly spaced vertical levels that become progressively thicker with depth, so that the uppermost ocean layers are well resolved.The topmost level is 25 m thick, while the bottommost level is ∼660 m thick.A horizontal resolution of 3.75

GENMOM coupling
To simplify the coupling between the atmosphere and ocean, both the GCMs are implemented on essentially the same T31 grid.In MOM2, the latitudinal grid spacing is not exactly T31, but is adjusted with a cosine-stretching factor (Pacanowski, 1996) to closely approximate T31.GENESIS has a 30-min timestep, and MOM2 has a 6 h timestep for scalar fields.The two models interact in an essentially synchronous manner, communicating every 6 h.6 h averages of the surface fluxes of heat, water and momentum are passed from GENESIS to MOM2, and MOM2 is run through one 6 h scalar timestep.The updated SSTs are passed back to GENESIS and used to run it for the next 6 h.Sea ice is treated within the LSX module of GENESIS, and under sea ice, fluxes between the sea-ice base and the uppermost ocean layer are passed to MOM2.Continental freshwater river runoff is globally averaged and spread over the world ocean.

Simulation of the present climate
We analyze the annual and seasonal climatologies of the last 30 years of a 700-year control simulation produced by GENMOM.Where possible, we compare the GENMOM results to ensembles of the AOGCMs used in the IPCC AR4 (Randall et al., 2007).We use the last 30 years of the Climate of the 20th Century experiment from eight selected IPCC AR4 models (Table 1) to provide context for evaluating the performance of GENMOM.For the present-day GENMOM simulation, atmospheric CO

Validation datasets and input files
In contrast to the IPCC AR4, wherein a variety of observed datasets are used to evaluate model performance, whenever possible we rely solely on the NOAA NCEP Reanalysis 2 data set (NCEP2, Kanamitsu, et al., 2002) to maintain consistency between and among variable fields.We use a standard climatology period of 1981-2005 for the NCEP2 data unless otherwise specified.Observed SST data are derived from the NOAA Optimum Interpolation Sea Surface Temperature V2 (OI SST, Reynolds, et al., 2002), which is a 1 • ×1 • gridded dataset based on combining in situ measurements and satellite observations.We use a climatology period of 1982-2005 for OI SST, because 1982 is the first full year for which the data are available.Global subsurface ocean temperatures were obtained from The World Ocean Atlas 2005 (WOA05, Locarnini et al., 2006), which is also a 1   oceanography reanalysis product forced with the first NCEP Reanalysis data (K öhl and Stammer, 2008;NCEP1;Kalnay et al., 1996).GENMOM input files for topography, bathymetry, and land-ocean mask were derived by interpolating the ICE-4G model (Peltier, 2002) reconstruction from 1 • × 1 • to T31

Atmospheric fields
Overall, the distribution of the zonally averaged profile of air temperature simulated by GENMOM is in agreement with the NCEP2 profile (Fig. 1); however, some deficiencies deserve additional attention.GENMOM simulates a cold bias raging up to 6 • C north of 30 • N in the Northern Hemisphere (NH) during both winter and summer, whereas a cold bias south of 60 • S in The Southern Hemisphere (SH) is present during austral summer.A cold bias is present in the simulated temperature in the uppermost atmosphere above the wintertime pole.Seasonally, GENMOM simulates the meridional shift of peak insolation and warmest surface temperatures well when compared to observations.The modeled tropical warm region is slightly more compressed meridionally than the NCEP2 data.
The summer and winter patterns and magnitudes of the annually averaged planetary jet stream structure are well captured by GENMOM (Fig. 2).In both winter hemispheres the core of the jetstream (at ∼200 hPa) and related upper level winds (500 hPa) are slightly enhanced relative to the NCEP2 data.These minor mismatches notwithstanding, the overall structure of the simulated jetstream suggests that GEN-MOM produces a realistic mean planetary-scale wind structure that is essential to the related positioning of the stormtracks.
GENMOM simulates the seasonally persistent positions of planetary ridges and troughs and thus the upper atmospheric flow and 500 hPa geopotential (Fig. 3a-d).
During boreal winter, the ridge over western North America is shifted eastward in GENMOM relative to observations and the associated trough to the east over northern Canada and the North Atlantic is similarly slightly displaced and more zonal relative to that of the NCEP2 data (Fig. 3b).The 500 hPa heights over North America and Eurasia are lower than those of the NCEP2 data resulting in slightly reduced wind velocities, particularly over eastern North America and the North Atlantic.In the SH, austral summer 500 hPa heights are well simulated but wind velocities associated with the westerlies are somewhat reduced due to the lower pressure gradient over Introduction

Conclusions References
Tables Figures

Back Close
Full the Southern Ocean and Antarctica and the lack of actual topographic forcing due to smoothing in the model.During boreal summer, the ridge over western North America is correctly placed in GENMOM, but the amplitude of the ridge is greater than observed and the related downstream trough is slightly deeper than that of the NCEP2 data (Fig. 3c).Heights in the region extending east of the Mediterranean and across India and China appear modestly lower than observed; however, part of the apparent discrepancy stems from values that are just above or just below the color breaks in the plotting scales.
In the SH, the comparison for the austral winter is similar to that of the austral summer.Spatial patterns of winter and summer mean sea level pressure (MSLP) are captured by GENMOM; however, regional differences exist (Fig. 3e-h).During boreal winter GENMOM simulates lower-than-observed MSLP in the Aleutian and Icelandic lows.As a result, wind velocities are enhanced over North America.In the SH, surface pressure and winds are comparable with those of the NCEP2 data except along SH westerlies where MSLP is higher and wind velocities are lower due to the reduced pressure gradient.
The boreal summer simulation of MSLP and wind velocities is quite good; the subtropical highs in the NH are well placed and the associated wind velocities are comparable to NCEP2 (Fig. 3h).MSLP and wind velocities in the tropics and the SH are also well simulated by GENMOM.During austral summer, the SH high-pressure anticyclones are somewhat weaker than observed.The simulated south Pacific high is weak and so does not produce anticyclonic flow, which contributes to a weakened South Pacific Gyre.The SH westerly winds are simulated to be too weak, presumably due to coarse resolution topography, which will influence ocean overturning.GENMOM simulates stronger-than-observed westerly winds across southern Europe, which is caused by the overactive Icelandic low.This is likely due to a cold temperature bias in the Norwegian Sea, which fails to isolate the low pressure center.
The vertical profile of atmospheric specific humidity simulated by GENMOM is in agreement with the NCEP2 data (Fig. 4).A dry bias, evident over the tropics, and Introduction

Conclusions References
Tables Figures

Back Close
Full a wet bias, evident in the SH below 700 hPa, are associated with the warm bias in atmospheric temperature caused by a warm Southern Ocean.A wet bias over the NH below 700 hPa is not associated with a warm bias in atmospheric temperature; rather, it is caused by weak convection from the surface to ∼400 hPa between 70 • N-80 • N.

Modeled surface temperatures
The simulated global mean-annual 2 m air temperature is 278.3K, in good agreement with the NCEP2 value of 278.9 K (Fig. 5).Over land the simulated temperature is 1.3 K colder than observed and over the oceans simulated temperature is 0.6 K warmer than observed.GENMOM simulates the meridional temperature gradient well.Major topographic features resolved by the model such as the Rocky Mountains, the Andes and the Himalayas, have regional temperatures that match well to observation.We note that the high latitude temperature anomalies (Fig. 6) are partially attributed to a mismatch between the ICE-4G derived land mask and that of NOAA OI SST V2 interpolated to T31.Where a mismatch occurs, large anomalies are created due to comparing an SST grid cell to a 2 m air temperature grid cell.
Exceptions to the agreement between simulated and observed 2 m temperatures are primarily found over the oceans (Figs. 5 and 6).The Southern Ocean warm bias may be caused by weaker-than-observed westerly winds across the Southern Ocean resulting in weak ocean overturning.The cold bias over the Norwegian Sea is caused by too much simulated sea-ice, as a result of too weak meridional overturning.Because the cold water tongue associated with the California Current is ∼300 km wide, it is not adequately resolved at T31, and leads to a warm bias along the Pacific coast of North America.The northward branch of the South Pacific Gyre also is not well resolved in addition to weak westerly trade winds, resulting in a weak cold water Humboldt Current along the western coast of South America.The weakened circulation results in a warm SST bias off the coast of Chile.
Our GENMOM simulation has many features in common with the IPCC AR4 models, including: (1) a cold bias over northern Europe, (2) a warm SST bias in the waters Introduction

Conclusions References
Tables Figures

Back Close
Full west of South America, (3) a warm bias in the Southern Ocean and (4) cold biases over the Himalayas and Greenland (Fig. 6).In contrast to many of the IPCC AR4 models, GENMOM simulates the annual surface temperature over much of Antarctica with anomalies <2 • C.
GENMOM captures the global patterns of the seasonal cycle of temperature but overestimates the amplitude over Greenland, South America, southeast United States and Australia and underestimates the amplitude over northern Africa, the western United States and much of Europe and Asia (Fig. 7).The model also simulates greater variability over some of the oceans, particularly in the mid latitudes.Similar to Fig. 6, grid cells where the land-ocean distribution does not match have large seasonal cycle amplitude anomalies.

Precipitation
GENMOM simulates global mean-annual precipitation reasonably well relative to both the reanalysis data and the IPCC models (Figs. 8,9).Similar to other AOGCMs, GEN-MOM produces a split Intertropical Convergence Zone (ITCZ) in the tropical Pacific.
During DJF, the southern branch of the ITCZ simulated by GENMOM extends too far to the east.In JJA, the northern branch of the ITCZ simulated by GENMOM is compressed and extends too far to the north relative to observations.Lin (2007) found that many IPCC AR4 models produce a double ITCZ which is caused by: ( 1 Globally, GENMOM underestimates seasonal DJF precipitation in the Indian Ocean and JJA precipitation over the southern Asian landmass and Atlantic precipitation.The zonally averaged annual precipitation clearly illustrates the double ITCZ in GENMOM (Fig. 10); the double peak in total precipitation is evident at 10 • N and 10 • S rather than the observed single and stronger peak between 10 • N and 0 • N. Outside of the tropics however, the modeled precipitation compares well with reanalysis and the selected IPCC AR4 model simulations.

Oceanic fields
The overall patterns of surface and subsurface ocean temperatures simulated by GEN-MOM compare well to observations; however, anomalies reveal biases exceeding 2 • C (Fig. 11).A cold bias is simulated over much of the surface and a warm bias is simulated in around the thermocline in the tropics.A warm bias in the near-surface of the Southern Ocean is consistent with the surface temperature anomalies shown in Fig. 6.The Southern Ocean warm bias is likely caused by weak ocean overturning.The warm bias in the tropical ocean mid-depths is attributed to weakened simulated upwelling and the use of a relatively high vertical diffusion coefficient (0.35 cm 2 s −1 ) that is prescribed to maintain reasonable ocean overturning; too much heat is diffused from the surface to the mid-depths.GENMOM captures the observed zonal distribution of salinity well for the Atlantic Ocean and Indian + Pacific Oceans with a few region specific discrepancies (Fig. 12).
Relative to the WOA05 data, in the Atlantic, GENMOM simulates lower salinity waters at high latitudes and higher salinity waters in the northern mid latitudes; the maximum centered on 30 • N exceeds observations and the maximum at 30 ocean orography, and weaker-than-observed Atlantic Meridional Overturning Circulation (AMOC).GENMOM does a good job at simulating the difference in salinity between the Atlantic Ocean and Indian + Pacific Oceans.Salinity in the Indian + Pacific Oceans matches well to observations with much of the zonal bias being less than ±0.2 PSS.We compare simulated global and basin ocean overturning for the full 300-year GEN-MOM with observations (Fig. 13).The last 30 years of the simulation coincidently displayed one of the weakest periods of overturning in the 300-yr simulation, so we use the full 300-year record as more representative in that multidecadal variability is smoothed out in the average.Globally, GENMOM simulates an overturning similar in pattern to that of the GECCO data.The most notable shortcoming in the GENMOM simulation is that the strength and depth of the Deacon Cell, which is characterized as deep clockwise meridional circulation in the Southern Ocean driven by windstress, are poorly captured.Wind velocities across the Southern Ocean are weaker-than-observed (Fig. 3) thereby failing to produce sufficient windstress to drive deep overturning (Toggweiler and Samuels, 1995;Sijp and England, 2009).The weak westerly winds are likely due to the coarse meridional resolution (Held and Phillipps, 1993;Tibaldi et al., 1990) and may also contribute to weak AMOC biases in non-flux corrected models with coarse atmospheric resolution, as found in earlier studies (Bryan et al., 2006;Schmittner et al., 2010).GENMOM's failure to simulate the Deacon Cell contributes to the warm Southern Ocean temperature bias by not upwelling deep cold water.
The simulated AMOC is similar in pattern to that of the GECCO data but it is somewhat weaker in strength.The maximum AMOC strength over the 300-year simulation is 13.3 ± 0.8 Sv, which is lower than the observed 16-18 Sv range.The models used in the IPCC AR4 generally fall between 12-20 Sv (Meehl et al., 2007b;Schmittner et al., 2005).The combined Indian Ocean and Pacific Ocean overturning matches well with observations with the exception that GENMOM simulates deeper-than-observed clockwise overturning in the northern tropics, which may imply the vertical diffusion coefficient is too high.Introduction

Conclusions References
Tables Figures

Back Close
Full GENMOM produces ocean surface currents that match generally well to observations on an annual average (Fig. 14).The major Atlantic surface currents are well simulated in GENMOM, with the exception of the Gulf Stream, which is too weak.The Antarctic Circumpolar Current flowing through the Drake Passage is well resolved, as is the continuing flow to the South Atlantic Current.In the Pacific, the equatorial currents are well simulated, but the North Equatorial Counter Current is not present and the North Equatorial Current weaker than that of the observations.The Kuroshio Current is well placed but also slightly weaker than that of the observations.The California Current is noticeably absent from the GENMOM simulated surface currents.Both the Humboldt Current and Antarctic Circumpolar Current have weaker-than-observed strengths.The strength of the Antarctic Circumpolar Current through the Drake Passage is simulated to be 35% weaker than the 119 Sv found in the GECCO reanalysis.In the Indian Ocean, GENMOM simulates the Indonesian Throughflow realistically, matching observations well.The Indonesian Throughflow throughput is found to be 12.7 ± 0.8 Sv, which compares well to the observed estimates of 9.3 ± 2.5 Sv (Gordon et al., 1999) and 13.2 ± 1.8 Sv (Lumpkin and Speer, 2007).Surface currents in the northern Indian Ocean are modulated by the monsoon, where currents flow westward during winter and eastward during summer.The annually averaged surface currents in Fig. 14 show westward flow dominating in GENMOM whereas eastward flow dominates in the observations.The incorrect direction of surface currents is most noticeable in GENMOM during winter and spring.Most of the deficiencies in simulated surface currents are attributed to the coarse resolution of the model and related inability to resolve subgrid components of the current.
GENMOM simulates winter sea-ice extent and concentration well in the NH for both DJF and JJA (Fig. 15).Sea-ice extends too far into the Norwegian Sea during both winter and summer and too far into Hudson Bay during winter.The excessive sea-ice in the Norwegian Sea is likely due to a weak AMOC, which is not transporting enough warm mid-latitude water north.Antarctica shows deficient sea-ice during both seasons, which may be explained by the warm temperature bias in the Southern Ocean.Introduction

Conclusions References
Tables Figures

Back Close
Full attributed to weak global overturning and the use of a high value of the vertical diffusion coefficient, which was needed to maintain realistic global ocean overturning.Salinity is generally well simulated, but with a fresh bias in the North Atlantic caused by underrepresentation of narrow channels (i.e., the Norwegian Sea) at T31 model resolution and a 1+ PSS salinity bias in the northern mid latitudes originating in the Gulf of Mexico.Ocean overturning is simulated with the correct spatial pattern, but is generally weaker-than-observed.We attribute the weak meridional ocean overturning to (i) weak and northwardly displaced westerly winds in the SH due to coarse topography and (ii) a narrow and shallow Drake Passage also due to coarse orography.Most ocean surface currents are well simulated by GENMOM, with the exception of narrow currents such as the Gulf Stream and the Kuroshio Current that are weakerthan-observed again due to the coarse T31 resolution.Northern Hemisphere Sea-ice is well simulated with the exception of excess sea-ice in the Norwegian Sea.However, the SH sea-ice extent is too small compared to observations.Both NH and SH deficiencies are linked to weak ocean overturning.
The evaluation performed here has shown that GENMOM produces a realistic climatology that is comparable to the models used in the IPCC AR4.GENMOM shares similar deficiencies with other models such as a double ITCZ, failure to resolve features due to resolution limitations (the California and Humboldt Currents), weak ocean overturning and having a general global cold bias.Despite these deficiencies, GEN-MOM produces biases that are within the range seen in the IPCC models.The overall climatology simulated by GENMOM is generally similar to that of previous GENESIS versions.The addition of a coupled ocean model, however, allows GENMOM to be used in studying phenomena such as ENSO that require dynamic ocean-atmosphere interaction.

GMDD Figures Back Close
Full Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 2 concentration is prescribed at a constant 355 ppmV, near the mean value for our climatology period of 1981-2005.GENMOM was initialized with a latitudinal-dependent temperature profile while salinity was uniformly prescribed at 35 ppt.Analysis of ocean temperatures indicates that spin up of the model was suitably achieved after 400 years.Over the last century of the simulation the deep ocean (>1000 m) warmed by ∼0.002 • C/decade whereas the mid layer (200 m-1000 m) warmed by ∼0.003 • C/decade and the surface layer was free of drift.Discussion Paper | Discussion Paper | Discussion Paper | resolution.Ice-sheet cover and thickness is prescribed by interpolating the ICE-4G model reconstruction to T31.To maintain numerical stability, over the northernmost Arctic Ocean cells in MOM2 we smooth the bathymetry field derived from ICE-4G with a 9-cell moving window.At T31 horizontal resolution the Bering Strait is closed.Modern values for the distribution of vegetation(Dorman and Sellers, 1989), soil texture(Webb     et al., 1993)  and freshwater lakes(Cogley, 1991) are prescribed.The use of ICE-4G orography to derive global topography, bathymetry, and ice-sheet extent is based on our goal of streamlining the configuration of GENMOM for paleoclimate applications.Figures Back Close Full Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | ) excessive tropical precipitation, (2) high sensitivity of modeled precipitation and surface air humidity to SST, (3) a lack of sensitivity of cloud amount to precipitation, and (4) a lack of sensitivity of stratus cloud formation to SST.GENMOM produces a cold SST bias in the Pacific Basin along with a confined cold tongue, both of which Lin (2007) noted as factors that result in a double ITCZ.Consistent with Lin (2007), GENMOM does not produce a significant double ITCZ when coupled to a slab ocean due to weakened ocean-atmosphere feedbacks.The double ITCZ problem can potentially be resolved by improving these ocean-atmosphere feedbacks.Discussion Paper | Discussion Paper | Discussion Paper | • S underestimates observations.A 1+ PSS salinity bias in the northern mid latitudes between 400-1000 m is attributed to a build up of salinity in the Gulf of Mexico caused by weaker-thanobserved circulation associated with the coarse resolution of ocean orography.Similarly, the low salinity bias north of 60 • N is associated with reduced northward penetration of the North Atlantic Drift into the Arctic, again due to the coarse resolution of Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |Antarctic winter sea-ice fails to reach the full extent seen in the observed dataset due to the warm SST bias in the Southern Ocean.4 ConclusionsWe present the first formal evaluation of the new AOGCM GENMOM, which is a nonflux corrected model comprised of GENESIS 3 atmospheric model, MOM2 ocean model and LSX.The main changes in GENESIS version 3 are (i) solar and thermal infrared radiation are calculated using the NCAR CCM3 radiation code, and (ii) the ocean is represented by the MOM2 ocean general circulation model.The spectral resolution of T31 for both atmosphere and ocean is used during this evaluation.The simulated global 2 m air temperature is 0.6 • C warmer over oceans and 1.3 • C colder over land.The jet stream structure and major planetary features of sea level pressure are well captured by the model.GENMOM produces a realistic mean planetary-scale wind structure that is needed to produce the correct position of stormtracks.The 500 hPa ridges and troughs are well simulated, as are the seasonal surface pressure cyclones and anticyclones.The annual surface temperature gradient and spatial distribution compare well both to observations and to the IPCC AR4 models.Cold SST anomalies in the Norwegian Sea are explained by excessive sea-ice in both winter and summer, which is in turn caused by weak Atlantic Ocean overturning.A warm bias in the Southern Ocean is attributed to a weak ocean overturning resulting in a poor simulation of the Deacon Cell, which suppresses associated cold water upwelling in the Southern Ocean.GENMOM fails to resolve adequately the South Pacific Gyre, which results in a warm SST bias in the eastern Pacific Ocean and weak anticyclonic atmospheric circulation around the gyre.GENMOM simulates a double ITCZ when coupled with the OGCM, which is not present when GENESIS is coupled to a slab ocean.The global ocean temperature is generally well simulated, with the exception of a warm bias between 200-1000 m in the tropics and mid-latitudes.The warm bias is Discussion Paper | Discussion Paper | Discussion Paper |
• × 3.75 • is used to match the atmospheric T31 resolution.The hybrid mixing