Our research has brought together several methods for better diagnosing and understanding low-latitude cloud feedbacks on current climate and climate change based on a variety of models.
A large part of our CPT research has focussed on developing metrics for evaluating changes in tropical clouds in climate models. We initially have focussed on the method of Bony et al (2004) which utilizes monthly mean mid-tropospheric vertical velocity to stratify the tropics into dynamic regimes. The different dynamic regimes are associated with different cloud types: ascent with deep cumulus convection, weak subsidence with trade cumulus clouds, and strong subsidence with stratocumulus clouds. We have applied this method to a suite of climate perturbation experiments using models from each of the three participating CPT modeling centers, the NCAR CAM 3.0, GFDL AM 2.12b, and GMAO NSIPP-2.
The monthly climatologies produced by the models are compared to NCEP
and ECMWF ERA40 reanalysis climatologies, and to SSM/I, TRMM, ISCCP, and ERBE
satellite data, similarly sorted using the Bony method. There are considerable
differences between model and observation and between models. As an example of the
differences between models, the figure below shows the Bony method applied to the models'
liquid+ice condensate climatology.
In terms of climate response, the inter-model differences show a large degree of cancellation, partly due to the opposing effects of middle and high cloud on the shortwave and longwave radiation budget. Therefore the net tropical climate sensitivies of the models are largely governed by their changes in low-cloud fraction and low cloud thickness in the tropics, which vary considerably between models. Some results of this work are documented in Wyant et al (2005) currently in review.
Because of the apparent importance of low clouds in modulating climate in the tropics and elsewhere, and their relative insensitivity to mid-tropospheric vertical velocity, we are also investigating other thermodynamic factors that control low-level cloudiness, such as lower-tropospheric stability. We have also studied the behavior of low cloud parameterizations in the climate models using model output from individual model colums with high time resolution.
We are also working with Marat Khairoutdinov at Colorado State University to evaluate and understand the cloud sensitivity of 'superparameterization runs' using NCAR CAM model with nested 2-D cloud-resolving models (CAM SP). This model has demonstrated more realistic behavior than other GCMs as measured by the annual mean tropical precipitation, the diurnal timing of precipitation, and the representation of the Madden-Julian oscillation (Khairoutdinov et al 2005). We are currently analyzing climate experiments with fixed climatological SST with uniform +2K climate perturbations. These experiments show key differences between conventional GCM's and superparameterization in their representations of clouds and the responses of clouds to climate changes.