A strong constraint on radiative forcing of well-mixed greenhouse gases
Summary
This Nature paper presents the first global, decadal line-by-line calculation of longwave (LW) instantaneous radiative forcing (IRF) for well-mixed greenhouse gases (WMGHGs) under realistic, all-sky conditions. Using NOAA/GFDL’s GPU-optimised GRTcode together with ERA5 reanalysis and satellite OLR (CERES EBAF), the authors derive a simple but powerful regression linking local outgoing longwave radiation (OLR) to LW IRF. The regression is validated against full line-by-line calculations and satellite observations, producing a tight observational constraint on present-day LW IRF (tropopause) of 3.69 ± 0.07 W m−2 by 2024 (relative to 1850). The method also isolates and reduces inter-model spread in CO2-induced effective radiative forcing (ERF) in Earth system models (ESMs), by diagnosing and correcting radiation-parameterisation biases.
Key Points
- GRTcode–ERA5 line-by-line calculations (2001–2024) show LW IRF increased from 2.66 to 3.70 W m−2 (tropopause) and TOA IRF from 1.85 to 2.49 W m−2.
- A near-linear, geographically consistent relationship exists between local OLR and LW IRF for all main WMGHGs (CO2, CH4, N2O, CFCs, HFCs), holding under clear-sky and all-sky conditions.
- The OLR–regression model, trained on clear-sky line-by-line results, accurately reconstructs all-sky LW IRF and reproduces global spatial patterns with ~2% uncertainty (95% CI ≈ ±0.064 W m−2 from regression).
- Applying the regression to CERES OLR yields an observationally constrained LW IRF increase from 2.65 to 3.69 W m−2 (2001–2024) with combined 95% CI ≈ ±0.073 W m−2.
- For 4×CO2 experiments, LW IRF explains most of the inter-model spread in ERF; correcting model-specific IRF biases reduces ERF spread by ~50–60%.
- The approach offers a computationally efficient alternative to full line-by-line diagnostics for benchmarking model radiation schemes using model or satellite OLR.
Content summary
The authors used GRTcode, a GPU-optimised line-by-line radiative transfer model, with ERA5 monthly fields and CMIP7/NOAA greenhouse-gas time series to compute gridded OLR and LW IRF from 2001–2024. They performed sensitivity experiments holding individual gases at pre-industrial levels to obtain IRF for each species. Clear-sky benchmarks agreed with previous line-by-line studies; all-sky values were then linked to OLR via a simple linear regression derived from 2010 clear-sky data and extended across concentration ranges.
The regression method reproduces spatial patterns and zonal means of LW IRF for each gas, and when driven by CERES satellite OLR gives an observationally constrained estimate of present-day forcing. The authors applied the framework to ESM diagnostics (double-call radiation and offline RTE-RRTMGP runs) to show that much of the ERF spread across models stems from differences in LW IRF produced by radiation parameterisations. By diagnosing and bias-correcting model IRF using each model’s OLR, the multi-model spread in 4×CO2 ERF is substantially reduced.
Context and relevance
Radiative forcing uncertainty is a major contributor to uncertainty in climate projections and equilibrium climate sensitivity. Prior benchmarks largely focussed on clear-sky, sampled profiles and left cloud/state dependencies poorly constrained. This study links spectrally resolved physics with long-term satellite OLR records to constrain LW IRF in realistic all-sky conditions and to provide a practical tool for model evaluation. It therefore ties directly into efforts to reduce uncertainty in climate projections and to improve confidence in multi-model assessments such as IPCC reports.
Author style
Punchy: the paper lays out a compact, physically motivated regression that both explains and constrains a leading source of model spread. If you care about reducing uncertainty in modelled CO2 forcing or want a practical diagnostic for radiation schemes, the details here matter.
Why should I read this
Short version: they found a neat trick — OLR tells you almost everything you need to know about longwave greenhouse-gas forcing in the real, cloudy world. If you squint at model biases or care about making climate projections less fuzzy, this saves you time and huge compute costs. Worth a read if you build, test or interpret climate models (or just want better numbers for CO2 forcing).