Intercomparison of the spectrally-resolved clear-sky outgoing longwave radiation estimates from multiple climate models

The outgoing longwave radiation (OLR), which results from the combination of thermal radiation emitted by the Earth surface and each layer of the atmosphere, is of critical importance for the Earth radiative budget. Climate models are generally tuned to match the space-borne reference values of the broadband OLR derived from e.g. the CERES mission. Yet a significant inter-model spread remains, originating from differences in the simulated climate system mean state and variability, especially in terms of atmospheric water vapor and temperature, surface temperature, aerosols and clouds. Spectrally-resolved OLR observations, as derived from infrared hyperspectral sounders such as IASI (and in the near future FORUM, to be launched in 2027, which will measure for the first time the far-infrared (FIR) region (100–667 cm^-1) at high spectral resolution), provide access to the spectral signature of individual climate processes and are thus valuable to identify biases in the geophysical variables of climate models. For instance, it can unveil spectral error compensations between distinct spectral ranges, beyond an apparent good match between simulated and observed broadband OLR. The present work investigates the inter-model spread of clear-sky broadband OLR of 9 CMIP6 climate models which reaches 5.6 W.m^-2. To that end, clear-sky FORUM-like spectra are simulated using the fast radiative transfer solver RTTOV and atmospheric profiles and surface properties of historical amip simulations (1979–2014 period). Spectra climatologies of the ERA5 reanalysis are also computed. Significant brightness temperature (BT) and radiance discrepancies between models arise across the OLR spectrum as a consequence of differences in simulated geophysical variables. For instance, the CO₂ band displays BT differences up to 16 K, which are directly linked to differences in the upper troposphere and lower stratosphere temperature. Differences as large as 3 K are also reported in the FIR H₂O absorption band (100–600 cm^-1) for the global annual mean BT, that can be even larger for specific latitudes. We show that the FIR H₂O region accounts for half of the inter-model broadband OLR variability and is strongly correlated to differences in mid-latitude and tropical upper-tropospheric relative humidity. This suggests that upper-tropospheric relative humidity is a key driver of the radiative budget in climate models. This work also highlights that FORUM observations shall provide a strong constrain on the climate models’ spectral signatures and thus help contribute to their improvement.