Vertical decorrelation length (Lcf) as used to determine overlap of cloudy layers in GCMs was obtained from CloudSat/CALIPSO measurements, made between 2007 and 2010, and analyzed in terms of monthly means. Global distributions of Lcf were produced for several cross-sectional lengths. Results show that: Lcf over the tropical convective regions typically exceeds 2 km and shift meridionally with season; the smallest Lcf (b1 km) tends to occur in regions dominated by marine stratiform clouds; Lcf for mid-to-high latitude continents of the Northern Hemisphere (NH) ranges from 5–6 km during winter to 2–3 km during summer; and there are marked differences between continental and oceanic values of Lcf in the mid-latitudes of the NH. These monthly-gridded, observationally-based values of Lcf data were then used by the Monte Carlo Independent Column Approximation (McICA) radiation routines within the Beijing Climate Center's GCM (BCC_AGCM2.0.1). Additionally, the GCM was run with two other descriptions of Lcf: one varied with latitude only, and the other was simply 2 km everywhere all the time. It is shown that using the observationally-based Lcf in the GCM led to local and seasonal changes in total cloud fraction and shortwave (longwave) cloud radiative effects that serve mostly to reduce model biases. This indicates that usage of Lcf that vary according to location and time has the potential to improve climate simulations.