A hybrid approach to identifying cloud regimes for model evaluation applied to the Southern Ocean cloud and radiation biases

Abstract:

Cloud regimes have proved useful in both observational and model evaluation studies to associate patterns of cloud-top properties with key cloud processes and dynamical contexts. Present model evaluation studies either identify cloud regimes directly from simulations, which may differ significantly from observations, or project simulated cloud on to observed cloud regimes using mean cloud-top properties. A remaining challenge for model evaluation is to compare observed and simulated cloud in a way that retains both the observed cloud structures to which we aspire, and the key modes of model bias we seek to diagnose.

We present a hybrid approach to identifying cloud regimes for model evaluation using observed and simulated cloud simultaneously. The resultant set of cloud regimes includes predominantly-observed and predominantly-simulated clouds, allowing us to distinguish cloud regimes associated with significant errors in the model from those where simulated clouds resemble observations. This approach permits a more explicit evaluation of major cloud biases while retaining observed links to cloud processes and dynamical contexts.

In this study hybrid cloud regimes are used to evaluate the Southern Ocean cloud biases in the Australian Community Climate and Earth-System Simulator (ACCESS). We identify two major sources of error: the model produces optically thin counterparts of observed low and midtopped cloud regimes, especially in the cold-air section of extra-tropical cyclones, which partially compensate for deficits of the latter; and a cloud regime associated warm conveyor belt of extratropical cyclones is not reproduced in any form in the model, contributing significantly to the total SW radiation bias. Finally the hybrid cloud regimes are used to quantify the effects of changes to the model microphysics parametrization, and the major cloud biases are compared against those in several other climate models.