Conditional averaging of the Cloud Radiative Effect as a higher order test of GCM radiation budgets

Global Climate Models (GCMs) are quite capable in producing temporally and spatially averaged radiative fluxes that are close to observed values. Closer examination however of clear-sky fluxes and Cloud Radiative Effects (CREs) reveal that the agreement is often the result of numerous error cancellations in the spatiotemporal and spectral domains. One manifestation of this phenomenon is canceling CRE errors among different cloud types. Recent approaches of cloud retrieval analysis from satellites allow us to determine the contribution to the total CRE of various cloud types, information that can be used as a diagnostic of the quality of cloud-radiation simulations in GCMs. In this presentation we apply such conditional averaging to CREs and cloud types provided by the International Satellite Cloud Climatology Project (ISCCP). The ISCCP D1 gridded cloud product contains the joint distribution of cloud top pressure and cloud optical depth at 280 km grid cells observed daily every 3-hours. The patterns of these joint distributions can be used to identify, via cluster analysis, distinct states of the atmosphere at the mesoscale, which ISCCP terms "weather states". The spatiotemporal distribution of distinct weather states is now available as a separate ISCCP D1 product for various geographical zones. We identify the relative contribution to the total CRE (shortwave, longwave, and net; both top of the atmosphere and surface) of these weather states separately for the extended low latitudes, northern midlatitudes and southern midlatitudes for the period 1984-2007 by conditionally averaging the CREs of the ISCCP FD data set according to weather state. Results from such a CRE breakdown that can be used as higher order GCM diagnostics include: (a) The seasonal cycle of CRE of the various weather states and the relationship between their relative strength and their frequency of occurrence; b) the identification of the most dominant weather states in terms of their relative CRE contributions; c) the range of CRE values for individual weather states and the degree to which they overlap with those of other weather states within the same geographical zone. GCMs wishing to implement stricter tests on the realism of their cloud-radiation simulations will find these results useful because the verisimilitude of their cloud types can be studied in conjunction with the fidelity of their radiative contributions.