Abstract. Most radiation schemes in weather and climate models use the
“correlated k distribution” (CKD) method to treat gas absorption,
which approximates a broadband spectral integration by N
pseudo-monochromatic calculations. Larger N means more accuracy and
a wider range of gas concentrations can be simulated but at greater
computational cost. Unfortunately, the tools to perform this
efficiency–accuracy trade-off (e.g. to generate separate CKD models
for applications such as short-range weather forecasting to climate
modelling) are unavailable to the vast majority of users of radiation
schemes. This paper describes the experimental protocol for the
Correlated K-Distribution Model Intercomparison Project (CKDMIP),
whose purpose is to use benchmark line-by-line calculations: (1) to
evaluate the accuracy of existing CKD models, (2) to explore how
accuracy varies with N for CKD models submitted by CKDMIP
participants, (3) to understand how different choices in the way that CKD
models are generated affect their accuracy for the same N, and (4) to generate freely available datasets and software facilitating the
development of new gas-optics tools. The datasets consist of the
high-resolution longwave and shortwave absorption spectra of nine
gases for a range of atmospheric conditions, realistic and idealized.
Thirty-four concentration scenarios for the well-mixed greenhouse
gases are proposed to test CKD models from palaeo- to future-climate
conditions. We demonstrate the strengths of the protocol in this
paper by using it to evaluate the widely used Rapid Radiative Transfer
Model for General Circulation Models (RRTMG).