Parameterization of water and ice cloud near-infrared single-scattering co-albedo in broadband radiation schemes

The parameterization of cloud shortwave absorption poses a difficult proble m in broadband radiation schemes that treat the near-IR region as a single interval. This problem arises because the spectral variation of the single- scattering co-albedo 1 - omega of cloud droplets and ice crystals is enormo us in the near-IR region, and because the cloud particle absorption is over lapped by sharply varying water vapor absorption. In this paper, several pa rameterization methods of cloud near-IR (0.68-4.00 mu m) 1 - omega are inte rcompared using a large set of atmospheric columns generated by a GCM. The methods include 1) linear averaging of 1 - omega, weighting with the TOA so lar flux; 2) "thick averaging" by Edwards and Slingo; 3) Fouquart's formula , which presents water cloud near-IR 1 - omega as a function of optical thi ckness; and 1) the "correlated omega" technique by Espinoza and Harshvardha n. An extension of the correlated omega technique to ice clouds is suggeste d. In addition, a new "adaptive omega" broadband parameterization technique is developed and tested. In this method, the near-IR 1 - omega of a cloud layer is parameterized in terms of the cloud properties (phase, optical thi ckness, and effective particle size) and the properties of the overlying at mosphere (slant vapor path and clouds). Two slightly different versions of the method are considered. The results of the intercomparison indicate that the adaptive omega method yields higher accuracy than the other broadband techniques tested. Linear a veraging is by far the least accurate method; in particular it is shown tha t linear averaging of near-IR 1 - omega can lead to substantially overestim ated absorption in ice clouds also. However, when the near-IR region is sub divided into three bands, the combination of thick averaging for water clou ds and linear averaging for ice clouds provides results superior to those o f all the broadband methods.