weather and climate studies apply only to plane-par<strong>al</strong>lel (horizont<strong>al</strong>ly homogeneous) atmospheres; they cannot be applied directly to partly cloudy atmospheres. Effi- cient scattering <strong>al</strong>gorithms for application to a horizon- t<strong>al</strong>ly nonhomogeneous atmosphere are not y<strong>et</strong> available. In a GCM, it is common that c<strong>al</strong>culations of LW ra- diative terms take >30% of the tot<strong>al</strong> computing time. To include c<strong>al</strong>culations of scattering of LW radiation in a partly cloudy atmosphere, it will require either smear- ing of a partly cloudy layer and reducing it to an equiv- <strong>al</strong>ent homogeneous layer or dividing the atmosphere into homogenous sections. The former approach will degrade the accuracy of radiation c<strong>al</strong>culations, while the latter approach will greatly enhance the computing time. Therefore, it is highly desirable to avoid explicit c<strong>al</strong>- culations of multiple-scattering in the therm<strong>al</strong> infrared (IR) in climate studies and, at the same time, r<strong>et</strong>ain the accuracy of flux c<strong>al</strong>culations. In this study, we develop a simple param<strong>et</strong>erization for the scattering of LW ra- diation by clouds, which can be implemented in long- term GCM climate simulations without requiring ex- plicit c<strong>al</strong>culations of multiple scattering. In addition to enhancing the accuracy, this param<strong>et</strong>erization practi- c<strong>al</strong>ly requires no extra computing time as compared to a pure absorbing/emitting atmosphere. The radiative transfer c<strong>al</strong>culations in this study are one-dimension<strong>al</strong>, only in the vertic<strong>al</strong> direction. We address neither the emission/scattering on the sides of clouds nor the effect of horizont<strong>al</strong> inhomogeneity in clouds. These problems are very complicated (e.g., Harshvardhan and Weinman 1982; Killen and Ellingson 1994; Cah<strong>al</strong>an <strong>et</strong> <strong>al</strong>. 1994) and are beyond the scope of this study. 2. Cloud optic<strong>al</strong> properties and the radiative transfer model The effect of clouds on LW scattering is studied here for spheric<strong>al</strong> liquid water dropl<strong>et</strong>s and randomly ori- ented hexagon<strong>al</strong> ice cryst<strong>al</strong>s. The important param<strong>et</strong>ers of cloud dropl<strong>et</strong>s in radiative transfer is the extinction coefficient, single-scattering <strong>al</strong>bedo, and scattering phase function (or asymm<strong>et</strong>ry factor). For water drop- l<strong>et</strong>s, these param<strong>et</strong>ers are computed using the Mie scat- tering <strong>al</strong>gorithm assuming a modified gamma function for the size distribution. For ice cryst<strong>al</strong>s, they are com- puted using the m<strong>et</strong>hod of Fu <strong>et</strong> <strong>al</strong>. (1998), which em- ploys a linear combination of single-scattering proper- ties derived from the Mie theory, the anom<strong>al</strong>ous dif- fraction theory, and the geom<strong>et</strong>ric optics m<strong>et</strong>hod. A tot<strong>al</strong> of 28 cirrus particle size distributions from aircraft mea- surements are used. Figures 1 and 2 show the distri- butions of the extinction coefficient, asymm<strong>et</strong>ry factor, and single scattering co-<strong>al</strong>bedo in the LW spectr<strong>al</strong> re- gion. Results shown in Fig. 1 are for various particle size distributions with the mass-weighted effective mean particle radius for water cloud, rw, equ<strong>al</strong> to 4, 8, and 16 ,um, and the results shown in Fig. 2 are for two cirrus cloud samples with the geom<strong>et</strong>ric mean particle size, rr - 0.6 7 -: 0.5 E .- g 0.4 $ 0 0.3 % ‘3 is 0.2 -t Ii 0.1 0.8 0.6 1 .o 0.8 0.6 0.4 0.2 0.0 1 . w-n ,,,,,,,,,,I,,,,,,,,,,,,,,! 0 500 1000 1500 2000 2500 3000 Wavenumber ( cm”) FIG. 1. Spectr<strong>al</strong> distributions of (a) the extinction coefficient, (b) asymm<strong>et</strong>ry factor, and (c) single scattering co-<strong>al</strong>bedo of liquid water cloud dropl<strong>et</strong>s. Here, T,+ is the mass-weighted effective mean particle radius. as defined in Fu (1996), of 50 and 95 pm. Gener<strong>al</strong>ly, the extinction coefficient decreases with increasing Y, (Figs. la and 2a), whereas the asymm<strong>et</strong>ry factor increases with increasing re (Figs. lb and 2b), where re denotes T%, for water dropl<strong>et</strong>s and T, for ice cryst<strong>al</strong>s. For the single scattering co-<strong>al</strong>bedo (Figs. lc and 2c), it decreases with increasing is the wavenumber. Y, for v > 1000 cm-‘, where v
JANUARY <strong>1999</strong> CHOU ET AL. 161 h -.k m 0.08 5 ‘B E 0.06 8 0 5 ‘3 0.04 2 ‘3 lz 0.02 1.2 ""["","',,'.","',,",, 0.8 0.6 0.8 0.6 0.4 FIG. 2. Same the gener<strong>al</strong>ized respectively. = 95pm Iil,,IC ,