Evaluating the Infrared Spectral Signature of CNRM-CM6-1 in Clear-Sky Conditions Using IASI Observations

For climate scientists and model developers, one of the most rigorous tests a general circulation model (GCM) can undergo is direct comparison of its simulated top-of-atmosphere radiance with high-resolution satellite observations under controlled conditions. Clear-sky scenes over oceans offer the perfect natural laboratory: minimal cloud contamination, well-known surface emissivity, and a relatively homogeneous atmosphere. It is precisely in these conditions that the French CNRM-CM6-1 model (the atmosphere component of the CNRM contribution to CMIP6) has recently been put under the spectral microscope using observations from the Infrared Atmospheric Sounding Interferometer (IASI).

Why clear-sky, and why IASI?

IASI, flying aboard the Metop satellites since 2006, remains the gold standard for hyperspectral infrared observations. With 8,461 channels covering 645–2,760 cm⁻¹ at 0.5 cm⁻¹ resolution (apodized), it resolves individual spectral lines of water vapor, CO₂, O₃, and numerous trace gases — features that broadband sensors simply cannot see. When we compare simulated IASI radiances from a GCM against actual IASI measurements in tropical and mid-latitude clear-sky oceans, we are essentially performing thousands of independent radiative transfer tests simultaneously.

Key findings from the evaluation

A recent study (still under review, but presented at the 2024 IASI Conference) reveals a remarkably good overall performance of CNRM-CM6-1, with a few noteworthy exceptions:

Global mean brightness temperature bias < 0.3 K across most of the longwave spectrum.

Excellent agreement in the CO₂ ν₂ band (667 cm⁻¹) and the window region (800–1200 cm⁻¹), with biases typically below 0.2 K.

Persistent cold bias (−1 to −2 K) in the water vapor ν₂ rotational band (centered around 1600 cm⁻¹), suggesting slightly too much upper-tropospheric moisture or minor spectroscopy issues in the forward model (likely ARPEGE’s RRTM-G implementation).

Small but systematic warm bias (+0.5 K) in parts of the ozone 9.6 µm band, possibly linked to the model’s stratospheric ozone climatology.

These deviations, while modest in absolute terms, translate into instantaneous radiative flux errors of a few tenths of W/m² — numbers that matter when we aggregate globally and annually.

What does this mean for CMIP6 and beyond?

CNRM-CM6-1 already stood out in CMIP6 for its relatively low equilibrium climate sensitivity (ECS ≈ 4.3 K) and credible representation of cloud feedbacks. The clear-sky spectral evaluation reinforces confidence in its longwave radiative code and in the broad-scale temperature and CO₂ absorption physics. The remaining discrepancies, particularly in the water vapor rotational band, are consistent with biases seen in other CMIP6 models when confronted with IASI or AIRS observations.

For the upcoming CMIP7 cycle, this type of hyperspectral benchmark is becoming non-negotiable. Modeling centers are now routinely coupling their GCMs with state-of-the-art line-by-line models and instrument simulators (e.g., the IASI PC-RTTOV forward operator) to close the loop between simulated physics and real-world spectral signatures.

Takeaway

CNRM-CM6-1 passes the clear-sky IASI test with flying colors in most spectral regions — a testament to decades of careful radiative transfer development at CNRM and Météo-France. The few remaining biases highlight exactly where future improvements in spectroscopy databases, vertical resolution of humidity, and minor gas climatologies will yield the next incremental gains.

When a model can reproduce 8,461 independent spectral measurements to within a fraction of a kelvin under the cleanest atmospheric conditions, we can be far more confident when it tells us how the Earth will respond to doubled CO₂ in a cloudy, messy real world.

Clear skies, sharp spectra, and rigorous validation — this is how trust in climate projections is built, one wavenumber at a time. 🌍✨

#ClimateModeling #IASI #CNRMCM #RadiativeTransfer #Hyperspectral #CMIP6