Evaluation Global Mean Time Series CMIP6

10 figures · 1980–2014

CMIP6 Multi-Model Mean Context

Comparison with CMIP6 ensemble mean from 11 members.

Contributing models: ACCESS-ESM1-5, AWI-CM-1-1-MR, CNRM-CM6-1, CNRM-ESM2-1, EC-Earth3, FGOALS-g3, GISS-E2-1-G, INM-CM5-0, IPSL-CM6A-LR, MPI-ESM1-2-LR, MRI-ESM2-0

Synthesis

IFS-FESOM2-SR demonstrates superior skill by balancing radiative closure and dynamic circulation, while the high-resolution ensemble generally corrects CMIP6 shortwave excesses through increased cloud opacity.
The high-resolution EERIE models exhibit a distinct systematic shift in radiative balance compared to the standard CMIP6 ensemble, driven primarily by differences in cloud parameterization. All three high-resolution configurations simulate 1.5–2.5% higher global total cloud cover than ERA5 and the CMIP6 Multi-Model Mean. Physically, this increased cloudiness acts as a necessary correction to the known 'too bright' bias of CMIP6; while CMIP6 overestimates surface downwelling shortwave radiation by ~4 W/m², the EERIE models reduce this to near-zero (IFS variants) or +1.5 W/m² (ICON-ESM-ER). However, the coupled response to this radiative tuning varies: IFS-FESOM2-SR achieves an optimal equilibrium, tracking ERA5 surface temperatures and circulation metrics with remarkable skill, whereas IFS-NEMO-ER falls into a 'cold/dry' state (1.5 K cold bias, 7 W/m² deficit in downwelling longwave) reinforced by excessively strong trade winds (negative 10m U-wind bias). Hydrologically, the models diverge significantly. ICON-ESM-ER drives a hyper-active global water cycle, acting as a positive outlier in precipitation, latent heat flux, and seasonal sensible heat amplitude, suggesting overly aggressive land-atmosphere coupling or turbulent exchange. Conversely, the IFS-based models align closely with CMIP6 hydrological baselines, though all models overestimate latent heat flux relative to the early ERA5 record. Crucially, the comparison highlights non-physical artifacts in the observational baseline; ERA5 displays spurious trends in latent heat and step-changes in precipitation (circa 1998) that the free-running models—which correctly capture physical transients like the 1991 Pinatubo eruption—do not replicate.

Related diagnostics

global_mean_sea_surface_temperature_bias cloud_radiative_effect_maps precipitation_climatology_maps zonal_mean_temperature_and_wind

Total Cloud Cover Global Mean Time Series

Total Cloud Cover Global Mean Time Series
Variables clt
Models IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER
Obs Dataset ERA5
Units %
Period 1980–2014

Summary high

All three high-resolution EERIE models (IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER) consistently overestimate global mean total cloud cover (TCC) by approximately 1.5–2.5% throughout the 1980–2014 period compared to ERA5 reanalysis and the CMIP6 multi-model mean.

Key Findings

  • IFS-NEMO-ER exhibits the largest positive bias, with global mean TCC fluctuating around 65.0–65.5%, approximately 2.5% higher than ERA5.
  • IFS-FESOM2-SR shows the lowest bias among the high-resolution models (mean ~64.0–64.5%), yet remains roughly 1.5% above the observational baseline.
  • The CMIP6 Multi-Model Mean (MMM) tracks the ERA5 reanalysis very closely, sitting slightly below it (~62.3%), highlighting a distinct behavior in the high-resolution runs which are offset significantly higher.
  • Interannual variability is captured reasonably well by the models, with IFS-FESOM2-SR showing a dip around 1999–2000 that matches the observational record.

Spatial Patterns

While this is a global mean time series, the temporal pattern shows a systematic, time-invariant positive offset for all high-resolution models relative to the baseline. No significant long-term secular trend is observed in the models or observations, although ERA5 shows a slight upward variation post-2000.

Model Agreement

The high-resolution models are in qualitative agreement regarding the sign of the bias (positive), but diverge in magnitude. IFS-NEMO-ER and ICON-ESM-ER cluster closer together (higher cloud cover), while IFS-FESOM2-SR is distinctively closer to observations. There is a clear separation between the conventional resolution CMIP6 MMM (lower cloud cover) and the high-resolution EERIE protocols (higher cloud cover).

Physical Interpretation

The systematic high bias in total cloud cover suggests that the cloud fraction parameterizations (likely involving critical relative humidity thresholds or cloud overlap assumptions) in these high-resolution configurations produce more extensive cloudiness than the standard resolution CMIP6 ensemble or the reanalysis. Physically, this overestimation of cloud cover would likely result in excessive reflection of shortwave radiation (too strong SW Cloud Radiative Effect) unless the clouds are optically thinner than observed. The difference between IFS-based free-running models (IFS-NEMO/FESOM) and the IFS-based reanalysis (ERA5) highlights the impact of data assimilation in constraining moisture and cloud fields in the latter.

Caveats

  • The 'Observation' used is ERA5 reanalysis, which itself relies on model parameterizations for cloud cover rather than direct satellite retrievals (like ISCCP or CLARA-A2), implying model-dependence in the reference.
  • Global means can mask significant regional compensating errors (e.g., lacking stratocumulus decks but excessive trade cumulus).

Surface Latent Heat Flux Global Mean Time Series

Surface Latent Heat Flux Global Mean Time Series
Variables hfls
Models IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER
Obs Dataset ERA5
Units W/m2
Period 1980–2014

Summary high

This time series compares global-mean surface latent heat flux (1980–2014) from three high-resolution coupled models against ERA5 reanalysis. The models consistently simulate higher latent heat fluxes (greater evaporation) than ERA5, though a strong increasing trend in the reanalysis significantly narrows this gap over time.

Key Findings

  • Systematic positive bias: All models overestimate global mean latent heat flux compared to ERA5, particularly in the 1980s and early 1990s (bias ranging from +4 to +7 W/m²).
  • Trend discrepancy: ERA5 exhibits a strong, non-linear increase in latent heat flux starting around the late 1990s (~5 W/m² rise), while the free-running models show much flatter, more stable time series with only slight positive trends.
  • Model ranking: ICON-ESM-ER (green) has the highest magnitude (~87 W/m²), followed by IFS-FESOM2-SR (blue), with IFS-NEMO-ER (orange) being the closest to observations (~84 W/m²), eventually overlapping with ERA5 in the post-2010 period.

Spatial Patterns

While spatial patterns are not shown, the temporal pattern reveals a distinct regime shift in the reanalysis (ERA5) around 1998 that is absent in the coupled models.

Model Agreement

The models show high inter-model consistency in terms of interannual variability (e.g., dips around 1992, likely Pinatubo) and lack of strong trend, but differ in absolute mean offset. They diverge significantly from the observational trend.

Physical Interpretation

The large upward trend in ERA5 latent heat flux is likely driven by inhomogeneities in the observing system (e.g., changes in satellite assimilation) rather than a purely physical acceleration of the hydrological cycle. Consequently, the apparent 'improvement' in model agreement in the 2000s may be coincidental. The persistent positive bias in models suggests a more vigorous global hydrological cycle (stronger evaporation) in the simulations compared to the reanalysis baseline.

Caveats

  • The strong trend in ERA5 is likely partly artifactual, complicating the assessment of true model bias trends.
  • Global mean values mask potential compensating regional biases (e.g., over western boundary currents vs. trade wind regions).

Surface Sensible Heat Flux Global Mean Time Series

Surface Sensible Heat Flux Global Mean Time Series
Variables hfss
Models IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER
Obs Dataset ERA5
Units W/m2
Period 1980–2014

Summary high

All three high-resolution models overestimate global mean surface sensible heat flux by approximately 1.5–2.0 W/m² compared to ERA5 reanalysis. While the IFS-based models exhibit seasonal cycle amplitudes comparable to ERA5, ICON-ESM-ER displays a significantly exaggerated seasonal variability.

Key Findings

  • Systematic positive bias: Models cluster around 18.5 W/m², while ERA5 averages ~16.5–17.0 W/m².
  • Seasonal cycle amplitude: ICON-ESM-ER has a drastically larger seasonal range (~8 W/m²) compared to IFS models (~4 W/m²) and ERA5 (~3.5 W/m²).
  • Model grouping: IFS-FESOM2-SR and IFS-NEMO-ER are nearly identical in mean and variability, indicating the atmospheric/land component (IFS) dominates this flux independent of the ocean model.
  • Trend discrepancy: ERA5 shows a gradual increasing trend and potential volcanic signals (e.g., post-1991) that are not clearly replicated in the relatively flat model time series.

Spatial Patterns

The analysis is global-mean only, but the strong seasonality implies the signal is dominated by Northern Hemisphere land masses. The exaggerated seasonality in ICON suggests regional issues likely over land (e.g., excessive summer heating or drying).

Model Agreement

High agreement between the two IFS variants suggests robust behavior across ocean couplings. Low agreement between ICON and the other datasets regarding seasonal amplitude. Poor agreement between all models and ERA5 regarding absolute magnitude.

Physical Interpretation

The positive bias suggests the models partition too much surface available energy into sensible heat rather than latent heat (higher Bowen ratio) compared to ERA5. ICON's extreme seasonality likely points to its land surface scheme (JSBACH) or coupling processes, possibly drying out too much in summer or having excessive turbulent exchange coefficients. The similarity of IFS-NEMO and IFS-FESOM confirms that global sensible heat is largely determined by atmospheric/land parameterizations.

Caveats

  • The 'Obs' reference is ERA5 reanalysis, where sensible heat flux is a model-derived forecast product rather than a direct observation, subject to its own biases.
  • Global means can obscure compensating regional errors.

Total Precipitation Rate Global Mean Time Series

Total Precipitation Rate Global Mean Time Series
Variables pr
Models IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER
Obs Dataset ERA5
Units kg/m2/s
Period 1980–2014

Summary high

Time series analysis of global mean total precipitation rate (1980–2014) comparing three high-resolution coupled models against the CMIP6 multi-model mean and ERA5 reanalysis.

Key Findings

  • ICON-ESM-ER exhibits a distinct positive ('wet') bias, with a global mean rate (~3.7e-5 kg/m²/s) significantly higher than ERA5, the IFS models, and the CMIP6 ensemble.
  • IFS-FESOM2-SR and IFS-NEMO-ER align well with the CMIP6 Multi-Model Mean (MMM), with IFS-NEMO-ER being slightly drier and closest to ERA5 absolute magnitudes during the 1980s.
  • A coherent dip in precipitation is visible across all models and observations around 1992, capturing the suppression of the hydrological cycle following the Mt. Pinatubo eruption.
  • ERA5 reanalysis shows a suspect step-increase in global mean precipitation around the late 1990s, likely an artifact of observing system changes (e.g., ATOVS assimilation) rather than a physical trend.

Spatial Patterns

While this is a global mean time series, the temporal evolution reveals a distinct discontinuity in the ERA5 observational record around 1998-2000 that is not reproduced by the free-running models, suggesting non-physical shifts in the reanalysis quality over time.

Model Agreement

The two IFS-based models (NEMO and FESOM2) show high agreement with each other and the CMIP6 MMM, clustering around 3.35-3.45e-5 kg/m²/s. ICON-ESM-ER is an outlier with a systematic high bias. All models agree on the timing of interannual variability driven by volcanic forcing (Pinatubo).

Physical Interpretation

Global mean precipitation is energetically constrained by atmospheric radiative cooling (latent heating balances cooling). ICON's wet bias suggests it generates stronger atmospheric radiative cooling than the IFS models. The 1992 dip illustrates the thermodynamic constraint where volcanic cooling leads to a temporary reduction in the global hydrological cycle.

Caveats

  • The step-change in ERA5 around 1998 makes it a poor baseline for assessing long-term trends in model precipitation.
  • Global mean precipitation in reanalysis is a forecast product dependent on the model physics and assimilation increments, not a direct observation, and often suffers from spin-up issues.

Mean Sea Level Pressure Global Mean Time Series

Mean Sea Level Pressure Global Mean Time Series
Variables psl
Models IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER
Obs Dataset ERA5
Units Pa
Period 1980–2014

Summary high

Time series analysis of global mean Mean Sea Level Pressure (MSLP) from 1980 to 2014, comparing three high-resolution models (IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER) against ERA5 reanalysis and the CMIP6 multi-model mean.

Key Findings

  • ICON-ESM-ER exhibits a significant systematic negative bias, with a global mean approximately 60-70 Pa lower than observations and the CMIP6 ensemble.
  • IFS-NEMO-ER shows a systematic positive bias of roughly 15-20 Pa relative to the CMIP6 mean.
  • IFS-FESOM2-SR aligns closely with the CMIP6 multi-model mean (~101135 Pa) and observations, showing the smallest absolute bias among the evaluated models.
  • ERA5 observations display a noticeable upward trend over the 35-year period that is not reproduced by the free-running climate models, which generally maintain stable long-term means.

Spatial Patterns

While spatial patterns are not shown, the temporal domain reveals distinct seasonal cycles in global mean MSLP. ICON-ESM-ER displays a noticeably larger seasonal amplitude compared to the IFS models and observations.

Model Agreement

There is significant disagreement in the absolute global mean value, representing offsets in total atmospheric mass or diagnostic calculation methods. IFS-FESOM2-SR is the only model that closely matches the CMIP6 MMM baseline. All models agree on a relatively flat long-term trend, diverging from the drift seen in ERA5.

Physical Interpretation

Global mean MSLP is a proxy for the total mass of the atmosphere. The constant offsets between models likely reflect differences in the initial total dry air mass specification or different algorithms for reducing surface pressure to sea level (particularly over high topography). The seasonal cycles likely result from global fluctuations in water vapor mass and temperature-dependent reduction errors. The upward trend in ERA5 is likely an artifact of the changing observing system (data assimilation increments) rather than a physical mass increase, which mass-conserving climate models would not capture.

Caveats

  • The trend in the observational reference (ERA5) should be interpreted with caution as reanalyses are not strictly mass-conserving and are subject to observing system changes.
  • Differences in global mean MSLP are often calibration issues (initial mass) rather than dynamic flaws.

Surface Downwelling Longwave Global Mean Time Series

Surface Downwelling Longwave Global Mean Time Series
Variables rlds
Models IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER
Obs Dataset ERA5
Units W/m2
Period 1980–2014

Summary high

This time series analysis evaluates global mean surface downwelling longwave radiation (1980–2014) for three high-resolution coupled models (IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER) and the CMIP6 multi-model mean against ERA5 reanalysis.

Key Findings

  • IFS-NEMO-ER exhibits a distinct, systematic negative bias of approximately 7–8 W/m² compared to ERA5 throughout the simulation period.
  • ICON-ESM-ER and the CMIP6 MMM show a positive bias of roughly 2–3 W/m², indicating a tendency to simulate higher downwelling longwave flux than observed.
  • IFS-FESOM2-SR demonstrates the closest agreement with ERA5, particularly from 1990 onwards, where the bias is negligible.
  • All models successfully capture the transient reduction in radiative flux (~1–2 W/m²) around 1992 following the Mt. Pinatubo eruption.

Spatial Patterns

Not applicable (global mean time series). Temporally, all datasets show a robust seasonal cycle and a long-term increasing trend consistent with planetary warming and increasing atmospheric emissivity.

Model Agreement

While models agree on the timing of variability (seasonal cycles and volcanic responses) and the long-term upward trend, there is significant disagreement in the absolute mean state, with a spread of over 10 W/m² between the coolest (IFS-NEMO-ER) and warmest (ICON-ESM-ER) models.

Physical Interpretation

Downwelling longwave radiation is driven by lower tropospheric temperature, water vapor, and cloud cover. The strong negative bias in IFS-NEMO-ER suggests a generally colder or drier atmosphere (or insufficient cloud fraction/optical depth) relative to reanalysis. Conversely, the positive bias in ICON-ESM-ER implies a warmer, moister, or cloudier atmosphere. The 1992 dip reflects the global cooling response to stratospheric aerosols from Pinatubo, reducing atmospheric thermal emission.

Caveats

  • IFS-FESOM2-SR shows a noticeable erratic drop and recovery around 1984, likely indicative of model spin-up or initialization shock.
  • The observational reference is ERA5 reanalysis, which is a model-constrained estimate rather than direct radiometric observation.

Surface Downwelling Shortwave Global Mean Time Series

Surface Downwelling Shortwave Global Mean Time Series
Variables rsds
Models IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER
Obs Dataset ERA5
Units W/m2
Period 1980–2014

Summary high

Time series of global mean surface downwelling shortwave radiation (1980–2014) comparing three high-resolution EERIE models and the CMIP6 multi-model mean against ERA5 reanalysis.

Key Findings

  • IFS-FESOM2-SR and IFS-NEMO-ER track ERA5 observations closely but exhibit a slight negative bias of approximately -1 W/m².
  • ICON-ESM-ER shows a positive bias of ~1.5 W/m² relative to ERA5, though it is significantly closer to observations than the CMIP6 Multi-Model Mean, which has a large positive bias of ~4 W/m².
  • All models and observations capture a distinct reduction in surface solar radiation (~2-3 W/m²) around 1991–1993, corresponding to the Mt. Pinatubo eruption.

Spatial Patterns

The time series is dominated by a strong seasonal cycle (amplitude ~20 W/m²) and the episodic dimming event in the early 1990s; no strong long-term trend is evident in the global mean over this period.

Model Agreement

The two IFS-based models (with different ocean components, NEMO vs. FESOM) are nearly identical, indicating that the atmospheric component (IFS) primarily determines surface shortwave biases. ICON stands apart with a higher mean value closer to, but lower than, the CMIP6 ensemble.

Physical Interpretation

The 1991/1992 dip is physically consistent with the 'dimming' effect of stratospheric sulfate aerosols from the Mt. Pinatubo eruption reflecting incoming solar radiation. The general positive bias in CMIP6 (and to a lesser extent ICON) compared to ERA5/IFS suggests differences in cloud cover parameterization or aerosol optical depth, with CMIP6 likely allowing too much shortwave radiation to reach the surface.

Caveats

  • The observational reference is ERA5 reanalysis, which is model-derived; comparisons with direct satellite products like CERES (available post-2000) might yield different absolute bias magnitudes.
  • The CMIP6 MMM represents a diverse ensemble, so its large bias reflects a mean state that individual models might deviate from significantly.

2m Temperature Global Mean Time Series

2m Temperature Global Mean Time Series
Variables tas
Models IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER
Obs Dataset ERA5
Units K
Period 1980–2014

Summary high

This time series compares global mean 2 m temperature from 1980–2014 across three high-resolution models, the CMIP6 multi-model mean, and ERA5 observations. While IFS-FESOM2-SR closely tracks observations, the other two models exhibit systematic cold biases of varying magnitudes.

Key Findings

  • IFS-FESOM2-SR shows excellent agreement with ERA5 and the CMIP6 MMM, effectively capturing both the absolute magnitude (~287.5 K) and the long-term warming trend.
  • IFS-NEMO-ER exhibits a severe cold bias of approximately 1.5 K relative to observations, with a mean state around 286 K.
  • ICON-ESM-ER shows a moderate systematic cold bias of roughly 0.5 K throughout the simulation period.
  • All models capture the prominent cooling event following the 1991 Mt. Pinatubo eruption, indicating correct implementation of volcanic forcing.

Spatial Patterns

A clear long-term warming trend is visible in all datasets. IFS-NEMO-ER displays a noticeably amplified seasonal cycle compared to observations and other models; this amplification is asymmetric, characterized by significantly deeper annual minima (winter cooling) while maxima are closer to the other models.

Model Agreement

IFS-FESOM2-SR is the best-performing model, indistinguishable from the observational baseline in the annual mean. CMIP6 MMM also tracks observations very well. IFS-NEMO-ER diverges significantly as an outlier with the strongest bias.

Physical Interpretation

The persistent cold biases in IFS-NEMO-ER and ICON-ESM-ER suggest systemic energy balance issues, likely related to cloud radiative forcing (excessive reflection) or surface albedo feedbacks (e.g., excessive sea ice extent). The exaggerated seasonal cycle in IFS-NEMO-ER, driven by deep minima, points to excessive radiative cooling or stable boundary layer issues over land/ice during winter months.

Caveats

  • Global mean metrics can mask significant regional compensating errors (e.g., warm tropics vs. cold poles).
  • The cause of the strong seasonal amplification in IFS-NEMO-ER requires spatial investigation to pinpoint (e.g., NH land vs. SH sea ice).

10m U Wind Global Mean Time Series

10m U Wind Global Mean Time Series
Variables uas
Models IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER
Obs Dataset ERA5
Units m/s
Period 1980–2014

Summary high

This time series illustrates the global mean 10m zonal (U) wind from 1980 to 2015, comparing three high-resolution coupled models (IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER) against ERA5 reanalysis and the CMIP6 multi-model mean.

Key Findings

  • The observational global mean (ERA5) and CMIP6 MMM remain stable around -0.35 to -0.40 m/s, indicating a net easterly component dominates the global surface wind field.
  • IFS-FESOM2-SR shows excellent agreement with the observational climatology, exhibiting minimal bias throughout the simulation period.
  • IFS-NEMO-ER exhibits a systematic negative bias (~ -0.15 m/s relative to ERA5), implying globally stronger easterlies or weaker westerlies.
  • ICON-ESM-ER shows a substantial positive bias (~ +0.25 m/s relative to ERA5), averaging around -0.1 m/s, which indicates weaker global easterlies or stronger westerlies compared to observations.

Spatial Patterns

While purely temporal, the figure implies spatial imbalances. The persistent negative global mean reflects the dominance of the vast tropical trade wind belts (easterlies) over mid-latitude westerlies in area-weighted averages. The stability of the mean over 35 years suggests these large-scale circulation components remain structurally consistent in the models, despite mean-state biases.

Model Agreement

There is significant divergence in the mean state. While IFS-FESOM2-SR aligns closely with ERA5 and CMIP6, the other two models bracket the observations with biases of opposite signs. The inter-model spread (~0.4 m/s) is comparable to the magnitude of the signal itself.

Physical Interpretation

The global mean 10m U wind represents a residual balance between tropical easterlies and mid-latitude westerlies. ICON-ESM-ER's positive bias suggests either an underestimation of trade wind strength or an equatorward shift/intensification of mid-latitude westerlies. Conversely, IFS-NEMO-ER's negative bias points to overly vigorous trade winds, often associated with a 'cold tongue' bias or strong Walker circulation. IFS-FESOM2-SR successfully captures the observed balance.

Caveats

  • As these are free-running coupled simulations, the phase of interannual variability (thin lines) is not expected to match observations (e.g., ENSO timing), only the statistical magnitude.
  • A correct global mean does not guarantee correct regional distribution; compensating errors between the NH and SH or between tropics and extratropics could mask spatial biases.

10m V Wind Global Mean Time Series

10m V Wind Global Mean Time Series
Variables vas
Models IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER
Obs Dataset ERA5
Units m/s
Period 1980–2014

Summary high

This time series shows the global mean 10m meridional (V) wind from 1980 to 2014, comparing three high-resolution models against ERA5 reanalysis and the CMIP6 multi-model mean.

Key Findings

  • The global mean annual V-wind is consistently positive (~0.18 m/s in ERA5), indicating a net northward surface flow component when averaged globally.
  • IFS-NEMO-ER slightly overestimates the annual mean (~0.21 m/s), while IFS-FESOM2-SR (~0.16 m/s) and ICON-ESM-ER (~0.10 m/s) underestimate it relative to ERA5.
  • ICON-ESM-ER shows the largest negative bias, falling below even the CMIP6 multi-model mean (~0.13 m/s), driven by deeper negative excursions in its seasonal cycle.
  • The seasonal cycle amplitude is large (ranging from approx. -0.5 to +0.8 m/s), dwarfing the annual mean values.

Spatial Patterns

The data exhibits a strong, consistent seasonal cycle with sharp peaks and troughs, reflecting the seasonal migration of the ITCZ and trade winds. No significant long-term trend is observed in the global mean over the 1980-2014 period.

Model Agreement

IFS-NEMO-ER provides the closest match to the observational magnitude (ERA5), albeit slightly high. IFS-FESOM2-SR is also close but slightly low. ICON-ESM-ER is a distinct outlier with a systematic low bias compared to both the IFS models and the CMIP6 ensemble mean.

Physical Interpretation

Global mean V-wind is a residual of large, opposing hemispheric flows (Hadley cells, monsoons). The positive mean likely reflects asymmetries in land-sea distribution and cross-equatorial flow (e.g., strong boreal summer monsoons). ICON's low bias suggests either a southward shift in the mean ITCZ position or imbalances in the strength of northerly vs. southerly trade winds compared to reality.

Caveats

  • Global mean V-wind is a small residual of large cancelling terms; small regional biases can significantly shift the global mean.
  • ERA5 reanalysis is used as ground truth, which itself relies on model physics where observations are sparse.