Evaluation Seasonal Cycle CMIP6
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
Related diagnostics
Total Cloud Cover Seasonal Cycle
| Variables | clt |
|---|---|
| Models | IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER |
| Obs Dataset | ERA5 |
| Units | % |
| Period | 1980–2014 |
Summary high
The figure illustrates the seasonal cycle of global mean Total Cloud Cover (%) for three high-resolution models compared to ERA5 reanalysis and the CMIP6 multi-model mean. All three EERIE models exhibit a systematic positive bias of approximately 1–3% relative to ERA5 throughout the entire year.
Key Findings
- All three evaluated high-resolution models (IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER) consistently overestimate global total cloud cover compared to ERA5, with biases ranging from +1% to +3.5%.
- The CMIP6 Multi-Model Mean (MMM) tracks the ERA5 magnitude significantly better than the high-resolution models, particularly in boreal winter, though it underestimates coverage during the boreal summer minimum (Jun–Oct).
- While phasing is generally consistent (minima in Aug/Sep, maxima in Nov–Jan), IFS-FESOM2-SR displays a pronounced secondary peak in May that is less evident in the other models or observations.
Spatial Patterns
Temporally, the global cycle is characterized by a minimum in boreal late summer (August-September) and maxima in boreal winter (December-January). ICON-ESM-ER shows the largest seasonal amplitude, driven by a very high January value (~66.4%) dropping sharply by April. IFS-NEMO-ER maintains a flatter profile during the first half of the year compared to the other models.
Model Agreement
Inter-model spread is roughly 1–2%, but all sit distinctly above the observational baseline. Interestingly, the two IFS-based models (NEMO-ER and FESOM2-SR) show different seasonal shapes (e.g., the May peak in FESOM2), suggesting that ocean coupling or specific tuning choices significantly influence the atmospheric cloud state.
Physical Interpretation
The systematic high bias in cloud cover implies that the high-resolution configurations may have tuned cloud condensation or fractional cover parameterizations too aggressively, or generate excessive moisture flux. Physically, this overestimation would likely result in too much reflected shortwave radiation (cooling effect) and increased longwave trapping (warming effect) in the radiation budget. The fact that CMIP6 MMM performs better on magnitude suggests that increased resolution does not automatically resolve cloud fraction biases and may require recalibration.
Caveats
- ERA5 is a reanalysis product and relies on model physics for cloud generation; comparisons with direct satellite products (e.g., CLARA-A2 or CERES-MODIS) might yield different bias magnitudes.
- Global means obscure regional error compensation (e.g., correct total cloud cover could result from balancing too few stratocumulus and too many high clouds).
Surface Latent Heat Flux Seasonal Cycle
| Variables | hfls |
|---|---|
| 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 exhibit a systematic positive bias in global mean surface latent heat flux compared to ERA5, with IFS-NEMO-ER showing the closest agreement while ICON-ESM-ER displays significant deviations in magnitude and seasonal shape during boreal winter.
Key Findings
- Systematic positive bias: All models overestimate global mean latent heat flux throughout the year, with excesses ranging from ~1.5 to over 6 W/m² depending on the model and season.
- Model ranking: IFS-NEMO-ER is the best-performing model, nearly matching observations from September to December, though it overestimates by ~2 W/m² in the first half of the year.
- ICON winter discrepancy: ICON-ESM-ER shows the largest biases, particularly in December and January (~6-7 W/m² too high), effectively flattening the seasonal amplitude compared to the observational double-minimum structure.
- Phase agreement: All models correctly capture the timing of the annual maximum in June/July, driven largely by Northern Hemisphere summer evaporation dynamics.
Spatial Patterns
The observational seasonal cycle features a 'double-dip' minimum in March and October (~81.3 W/m²) and a maximum in July (~86.6 W/m²). While models capture the mid-year peak, they generally struggle to reproduce the depth and timing of the minima, particularly the vernal (March) minimum.
Model Agreement
The two IFS-based models (NEMO and FESOM2) show very similar seasonal shapes, with FESOM2 consistently shifted higher by ~1.5-2 W/m². ICON-ESM-ER diverges significantly in boreal winter (DJF), showing a local maximum in December where observations and IFS models show low values.
Physical Interpretation
The pervasive positive bias suggests global over-evaporation in the models, likely driven by combinations of warm sea surface temperature (SST) biases, excessive surface wind speeds, or biases in boundary layer humidity (too dry). The distinct June/July peak reflects the dominance of Northern Hemisphere land evapotranspiration and ocean evaporation in the global mean. ICON's high winter flux suggests issues with maintaining stability or excessive ventilation over wintertime oceans.
Caveats
- Global means obscure regional error compensation (e.g., excessive tropical evaporation vs. insufficient high-latitude flux).
- ERA5 is a reanalysis product and, while high quality, relies on bulk aerodynamic formulas similar to those in models, so biases may share common parameterization roots.
Surface Sensible Heat Flux Seasonal Cycle
| Variables | hfss |
|---|---|
| Models | IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER |
| Obs Dataset | ERA5 |
| Units | W/m2 |
| Period | 1980–2014 |
Summary high
This figure illustrates the global mean seasonal cycle of surface sensible heat flux, showing that all three models generally overestimate the flux compared to ERA5 reanalysis, with distinct behaviors between the IFS and ICON model families.
Key Findings
- IFS-FESOM2-SR and IFS-NEMO-ER track each other extremely closely, exhibiting a systematic positive bias of approximately 1.5–2.0 W/m² year-round compared to ERA5.
- ICON-ESM-ER displays a significantly amplified seasonal cycle (amplitude ~7.8 W/m²) compared to observations (~3.1 W/m²), matching ERA5 well in boreal winter (Dec-Jan) but overshooting drastically in boreal summer (Jun-Jul peak >22 W/m² vs ~18 W/m²).
- The peak flux occurs in boreal summer (June-July) for all datasets, but the models differ in the sharpness of this peak; ICON is very peaked, while IFS and Obs are broader.
Spatial Patterns
While spatially aggregated, the strong seasonality peaking in June/July suggests the global signal is heavily influenced by Northern Hemisphere land masses. The exaggerated summer peak in ICON implies a specific sensitivity or bias in NH summer land-atmosphere interactions.
Model Agreement
There is high agreement between the two IFS variants (FESOM2 vs NEMO), indicating that the atmospheric component (IFS) dominates the surface flux characteristics regardless of the ocean coupling. Conversely, there is significant disagreement between IFS and ICON regarding the seasonal amplitude.
Physical Interpretation
Sensible heat flux is driven by the surface-air temperature gradient and turbulence. The systematic positive bias in IFS suggests consistently stronger coupling or larger surface-air temperature deficits than ERA5. The large amplitude in ICON suggests different land surface scheme behavior, potentially partitioning more energy into sensible rather than latent heat (higher Bowen ratio) over NH land during summer, or excessive surface warming.
Caveats
- Global means can mask compensating regional biases (e.g., positive land bias vs negative ocean bias).
- ERA5 is a reanalysis product and relies on its own model physics for surface fluxes, though it is constrained by observations.
Total Precipitation Rate Seasonal Cycle
| Variables | pr |
|---|---|
| Models | IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER |
| Obs Dataset | ERA5 |
| Units | kg/m2/s |
| Period | 1980–2014 |
Summary high
The figure illustrates the seasonal cycle of global mean total precipitation rate, comparing three EERIE models and the CMIP6 multi-model mean against ERA5 reanalysis. While all models generally capture the phase of the seasonal cycle (peaking in July), there are significant differences in mean magnitude.
Key Findings
- IFS-NEMO-ER shows exceptional agreement with ERA5 observations, closely matching both the magnitude (~3.3 to 3.5 × 10⁻⁵ kg/m²/s) and phase of the seasonal cycle throughout the year.
- ICON-ESM-ER exhibits a substantial systematic wet bias, overestimating global precipitation by approximately 10% (0.3–0.4 × 10⁻⁵ kg/m²/s) compared to ERA5.
- IFS-FESOM2-SR and the CMIP6 MMM both show a moderate positive bias (~0.1 × 10⁻⁵ kg/m²/s) while correctly reproducing the seasonal timing.
- All datasets show a bimodal minimum (March and October) and a primary maximum in July (boreal summer).
Spatial Patterns
The temporal pattern is consistent across datasets: a global minimum in March, rising to a peak in July, a secondary dip in October, and a slight recovery in December/January. This likely reflects the dominance of Northern Hemisphere land-driven convection (monsoons) on the global mean budget.
Model Agreement
There is a strong hierarchy in model performance relative to ERA5: IFS-NEMO-ER (excellent) > CMIP6 MMM > IFS-FESOM2-SR > ICON-ESM-ER (poor). The two IFS-based models outperform ICON in terms of bias magnitude, likely due to sharing atmospheric physics with the ERA5 reanalysis system.
Physical Interpretation
Global mean precipitation is energetically constrained by the atmosphere's radiative cooling capability. The positive biases in ICON-ESM-ER and IFS-FESOM2-SR suggest either a too-efficient radiative cooling (requiring more latent heating to balance) or positive surface energy imbalances (e.g., warm SST biases) driving excessive evaporation. The superior performance of IFS-NEMO-ER suggests its configuration (resolution/coupling) maintains a global energy balance much closer to the reanalysis state.
Caveats
- Global means can mask significant regional compensating errors (e.g., double ITCZ vs dry zones).
- ERA5 is a reanalysis product generated by the IFS model family; therefore, IFS-based models (NEMO/FESOM) are structurally predisposed to agree better with it than ICON.
Mean Sea Level Pressure Seasonal Cycle
| Variables | psl |
|---|---|
| Models | IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER |
| Obs Dataset | ERA5 |
| Units | Pa |
| Period | 1980–2014 |
Summary high
This diagnostic evaluates the seasonal cycle of global mean Mean Sea Level Pressure (MSLP) for three high-resolution models against ERA5 reanalysis and the CMIP6 multi-model mean.
Key Findings
- IFS-FESOM2-SR demonstrates excellent agreement with ERA5, closely tracking both the amplitude and absolute magnitude of the seasonal cycle with minimal bias.
- ICON-ESM-ER exhibits a substantial systematic negative bias of approximately 80–100 Pa (~1 hPa) relative to observations throughout the year, although it correctly captures the phase.
- IFS-NEMO-ER shows a consistent positive bias of roughly 20–30 Pa but reproduces the observational phase and amplitude well.
- The CMIP6 Multi-Model Mean (MMM) overestimates the amplitude of the seasonal cycle compared to ERA5, showing higher winter peaks and lower summer troughs, whereas IFS-FESOM2-SR improves upon this.
Spatial Patterns
All models and observations show a clear seasonal cycle with a maximum in NH winter (December–February) and a minimum in NH summer (June–July). The amplitude of the cycle is approximately 30–40 Pa in observations.
Model Agreement
While all models agree on the timing (phase) of the seasonal cycle, there is significant disagreement in the global mean absolute value (intercept). IFS-FESOM2-SR agrees best with observations, while ICON-ESM-ER is an outlier with lower pressure.
Physical Interpretation
The seasonal variation in global mean MSLP is primarily driven by the seasonal cycle of atmospheric water vapor content (mass loading). The systematic offsets, particularly in ICON-ESM-ER, likely stem from differences in the initialization or conservation of total atmospheric mass (dry air mass tuning) or differences in the algorithms used to reduce surface pressure to sea level over high topography, rather than dynamic deficiencies.
Caveats
- Global mean MSLP is a derived quantity involving temperature-dependent reduction to sea level over land, which can introduce artifacts unrelated to atmospheric mass.
- Systematic offsets in global mean pressure are often a matter of model tuning/initialization rather than predictive skill.
Surface Downwelling Longwave Seasonal Cycle
| Variables | rlds |
|---|---|
| Models | IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER |
| Obs Dataset | ERA5 |
| Units | W/m2 |
| Period | 1980–2014 |
Summary high
The figure illustrates the global mean seasonal cycle of surface downwelling longwave radiation, revealing significant discrepancies in magnitude between models despite consistent phase alignment with ERA5 reanalysis.
Key Findings
- IFS-FESOM2-SR shows exceptional agreement with ERA5 observations, tracking the annual cycle almost perfectly (within ~0.5 W/m²) and outperforming the CMIP6 Multi-Model Mean.
- IFS-NEMO-ER exhibits a large systematic negative bias of approximately -6 to -7 W/m² throughout the entire year, consistently simulating values well below both observations and other models.
- ICON-ESM-ER and the CMIP6 MMM both display a positive bias, particularly in the boreal summer and autumn (July–October), overestimating radiation by roughly 1–2 W/m².
- All models correctly capture the seasonal phase, with the global minimum in January (~330 W/m²) and maximum in July (~349 W/m²), driven by Northern Hemisphere land warming.
Spatial Patterns
The seasonal cycle follows the typical global pattern dominated by the Northern Hemisphere landmass response, peaking in July. The bias offsets are largely constant throughout the seasonal cycle for IFS-NEMO-ER (negative) and IFS-FESOM2-SR (neutral), while ICON-ESM-ER's positive bias amplifies slightly during the boreal summer peak.
Model Agreement
Inter-model spread is substantial (~7–8 W/m²), which is significant for a global energy budget term. IFS-FESOM2-SR is the clear top performer relative to ERA5, while IFS-NEMO-ER is a distinct outlier on the low side.
Physical Interpretation
Surface downwelling longwave radiation is primarily a function of lower-tropospheric temperature, water vapor content, and cloudiness. The strong negative bias in IFS-NEMO-ER suggests a globally cooler lower atmosphere, reduced specific humidity, or insufficient cloud cover compared to the other configurations. Conversely, the positive bias in ICON and CMIP6 suggests slightly warmer or moister/cloudier conditions. The divergence between IFS-NEMO and IFS-FESOM is notable given they share an atmospheric core (IFS), hinting that the ocean coupling or specific resolution tuning (ER vs SR) strongly impacts the atmospheric state.
Caveats
- Global mean values can obscure compensating regional biases (e.g., opposing errors in tropics vs. poles).
- ERA5 is a reanalysis product and acts as the reference, but direct observations (e.g., surface stations) might show slight differences.
Surface Downwelling Shortwave Seasonal Cycle
| Variables | rsds |
|---|---|
| Models | IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER |
| Obs Dataset | ERA5 |
| Units | W/m2 |
| Period | 1980–2014 |
Summary high
The figure illustrates the global mean seasonal cycle of surface downwelling shortwave radiation (RSDS), driven primarily by Earth's orbital eccentricity. The high-resolution EERIE models generally exhibit lower RSDS values than the CMIP6 Multi-Model Mean (MMM), reducing the positive bias relative to ERA5 observations, though specific seasonal biases persist.
Key Findings
- The CMIP6 MMM consistently overestimates surface downwelling shortwave radiation throughout the year (by ~2–7 W/m²) compared to ERA5, indicating a systematic 'too bright' bias likely due to insufficient cloud attenuation.
- IFS-FESOM2-SR and IFS-NEMO-ER show remarkable agreement with ERA5 during the boreal summer minimum (May–August) but underestimate RSDS by ~3–4 W/m² during the boreal winter maximum (December–January).
- ICON-ESM-ER exhibits a distinct seasonal shape compared to the IFS models, overestimating RSDS in boreal spring (March–May) and matching the IFS negative bias in late year (October–December).
- Both IFS configurations track each other closely, suggesting that the atmospheric physics (OpenIFS) dominates this metric over the choice of ocean model (FESOM2 vs. NEMO).
Spatial Patterns
The temporal pattern follows the solar insolation cycle driven by Earth's eccentricity, with a global minimum in July (aphelion) and maximum in January (perihelion). While all models capture this phase, the amplitude and specific monthly deviations vary significantly.
Model Agreement
The high-resolution models generally group closer to the ERA5 observations than the CMIP6 ensemble mean. Among the high-res models, the two IFS variants are highly consistent with each other, whereas ICON diverges in the boreal spring months.
Physical Interpretation
The pervasive positive bias in CMIP6 suggests a widespread deficiency in atmospheric scattering/reflection, likely due to underestimated cloud fraction or optical depth. The IFS models correct this in the boreal summer but appear to over-correct in the boreal winter (perihelion), potentially implying excessive cloud radiative forcing during Southern Hemisphere summer. ICON's behavior suggests a different parameterization sensitivity, possibly related to seasonal cloud transitions.
Caveats
- Global means can mask significant compensating regional biases (e.g., opposite biases in the tropics vs. mid-latitudes).
- The reference dataset is ERA5 reanalysis; while robust, direct satellite observations like CERES EBAF would provide an independent validation standard.
2m Temperature Seasonal Cycle
| Variables | tas |
|---|---|
| Models | IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER |
| Obs Dataset | ERA5 |
| Units | K |
| Period | 1980–2014 |
Summary high
This figure illustrates the global-mean 2m temperature seasonal cycle, comparing three high-resolution models against ERA5 observations and the CMIP6 multi-model mean (MMM). IFS-FESOM2-SR demonstrates excellent agreement with observations, while IFS-NEMO-ER and ICON-ESM-ER exhibit varying degrees of systematic cold biases.
Key Findings
- IFS-FESOM2-SR tracks the ERA5 observational baseline almost perfectly throughout the year, with negligible bias (< 0.1 K).
- IFS-NEMO-ER shows a substantial systematic cold bias of approximately 1.2–1.5 K year-round, representing the largest deviation among the models shown.
- ICON-ESM-ER exhibits a moderate cold bias (~0.5–0.8 K) in boreal winter/spring but converges closer to observations during the boreal summer peak.
- The CMIP6 MMM slightly overestimates the seasonal maximum (July) compared to ERA5, whereas IFS-FESOM2-SR adheres more closely to the observed peak.
Spatial Patterns
While this is a global-mean diagnostic, the temporal pattern shows a clear seasonal cycle driven by Northern Hemisphere landmass dominance (peaking in July/August). All models correctly capture this phase, indicating correct fundamental physics regarding land-ocean heat capacity contrasts.
Model Agreement
There is strong disagreement in the absolute mean state (offsets up to 1.5 K), but high agreement in the phase and amplitude of the seasonal cycle. IFS-FESOM2-SR aligns with the CMIP6 MMM and observations, while the other two EERIE models are distinct outliers on the cold side.
Physical Interpretation
The global mean temperature cycle is dominated by the lower heat capacity of Northern Hemisphere land. The systematic cold biases in IFS-NEMO-ER and ICON-ESM-ER likely stem from imbalances in the global energy budget, such as excessive cloud reflectivity (SW cooling), overly extensive sea ice (albedo feedback), or cold sea surface temperature biases. The superior performance of IFS-FESOM2-SR suggests that its atmospheric-ocean coupling or tuning is better balanced for the global mean state than the ER configurations shown.
Caveats
- Global mean values can hide compensating regional errors (e.g., a warm bias in the NH cancelling a cold bias in the SH).
- The cause of the strong cold bias in IFS-NEMO-ER requires further diagnosis via radiation budget or sea ice extent figures.
10m U Wind Seasonal Cycle
| Variables | uas |
|---|---|
| Models | IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER |
| Obs Dataset | ERA5 |
| Units | m/s |
| Period | 1980–2014 |
Summary high
The figure illustrates the seasonal cycle of global-mean 10m zonal wind (U), revealing that IFS-FESOM2-SR replicates ERA5 observations with remarkable accuracy, whereas IFS-NEMO-ER and ICON-ESM-ER exhibit significant, opposing systematic biases.
Key Findings
- IFS-FESOM2-SR shows excellent agreement with ERA5, tracking the observational monthly mean almost perfectly throughout the year.
- ICON-ESM-ER displays a strong positive bias (~0.25 m/s), incorrectly simulating a net global westerly flow (positive values) in August and September.
- IFS-NEMO-ER exhibits a systematic negative bias (~0.15 m/s) relative to observations, indicating globally stronger net easterlies or weaker westerlies.
- The CMIP6 multi-model mean performs well, closely following the observational phase and magnitude, though IFS-FESOM2-SR matches the specific ERA5 trajectory even better.
Spatial Patterns
The global seasonal cycle exhibits a minimum (strongest net easterlies) in April (~-0.53 m/s) and a maximum (weakest net easterlies) in September (~-0.19 m/s). All models capture this phasing correctly, which is driven by the seasonal migration of the ITCZ and hemispheric asymmetries in the strength of trade winds versus mid-latitude westerlies.
Model Agreement
Inter-model spread is large (~0.5 m/s range between ICON and IFS-NEMO), exceeding the amplitude of the seasonal cycle itself (~0.35 m/s). While all models agree on the phase (timing of peaks/troughs), they disagree significantly on the mean state.
Physical Interpretation
Global-mean surface zonal wind is typically negative (easterly) because the surface area of tropical trade winds exceeds that of mid-latitude westerlies. ICON's positive bias suggests either underrepresented trade winds or overly zonal (strong) mid-latitude westerlies, a common issue in high-resolution models where eddy-mean flow interactions can be sensitive to drag parameterizations. The divergence between IFS-FESOM2 and IFS-NEMO—despite sharing the IFS atmospheric core—points to the influence of the underlying ocean model (FESOM2 vs NEMO) or resolution-dependent tuning (SR vs ER) on sea surface temperatures and surface drag.
Caveats
- Global mean values are residuals of large positive (westerlies) and negative (easterlies) terms; a correct global mean does not guarantee correct regional dynamics.
- Differences between SR (Standard Resolution) and ER (Eddy Rich) configurations may conflate resolution effects with ocean model differences.
10m V Wind Seasonal Cycle
| Variables | vas |
|---|---|
| Models | IFS-FESOM2-SR, IFS-NEMO-ER, ICON-ESM-ER |
| Obs Dataset | ERA5 |
| Units | m/s |
| Period | 1980–2014 |
Summary high
The figure illustrates the seasonal cycle of global mean 10m meridional (V) wind, reflecting the planetary-scale oscillation of cross-equatorial surface flow associated with the Hadley circulation and ITCZ migration. While all models capture the phase correctly, IFS-NEMO-ER demonstrates the highest fidelity to ERA5 observations, whereas other models exhibit notable amplitude biases.
Key Findings
- IFS-NEMO-ER tracks ERA5 observations remarkably well throughout the year, with only a slight positive bias in peak boreal summer months.
- IFS-FESOM2-SR exaggerates the seasonal cycle amplitude, simulating excessive southward flow in boreal winter (~-0.45 m/s vs -0.3 m/s obs) and excessive northward flow in boreal summer (~0.8 m/s vs 0.7 m/s obs).
- ICON-ESM-ER exhibits a distinct southward bias during boreal winter and spring (Jan-Apr), reaching minima of -0.53 m/s compared to -0.32 m/s in observations, though it aligns well with the boreal summer peak.
- The CMIP6 Multi-Model Mean generally lags behind the observations and high-resolution models during the boreal spring transition (Mar-Jun), underestimating the rate at which northward flow increases.
Spatial Patterns
The temporal pattern follows the solar cycle: peak northward global-mean flow occurs in July-August (boreal summer, ITCZ north), and peak southward flow occurs in January-February (boreal winter, ITCZ south). This non-zero global mean reflects hemispheric asymmetries in surface wind regimes and cross-equatorial transport.
Model Agreement
Models agree on the timing (phase) of the minima and maxima but diverge significantly on amplitude. The spread is largest in boreal winter (Jan), where ICON is ~0.2 m/s lower than IFS-NEMO/Obs, and smallest in autumn (Sep-Oct). The high-resolution IFS runs generally capture the sharpness of the seasonal transitions better than the CMIP6 ensemble mean.
Physical Interpretation
Global mean 10m V-wind serves as a proxy for the net cross-equatorial surface mass flux driven by the Hadley cells. Positive values indicate net northward flow (typical of JJA when the thermal equator is in the NH). The exaggerated amplitude in IFS-FESOM2 suggests an overly strong Hadley circulation surface branch or intensified meridional pressure gradients. ICON's negative bias in DJF implies excessively strong southward cross-equatorial flow or biases in the SH subtropical highs during austral summer.
Caveats
- Global mean V-wind is a residual of large cancelling regional flows; small mean biases may imply substantial regional circulation errors.
- Differences between IFS-NEMO and IFS-FESOM (same atmosphere) suggest ocean coupling (SST patterns/gradients) plays a critical role in driving these atmospheric circulation biases.