Diagnostic sub-projects

Widespread use of the Transpose-AMIP II data is encouraged. The data will be freely available through PCMDI for research purposes (details will be available from this site once data is available). Conditions on the submission of data from some centres mean that transpose-AMIP II data cannot be used for commercial purposes.

A call for diagnostic subprojects has been launched so that the community can see what work is proposed and to act at a catalyst for further subprojects. If you are interested in working with the transpose-AMIP II data, please send an email to keith.williams@metoffice.gov.uk with the name of the sub-project leader and paragraph outlining what you are planning to do. Submitted sub-project summaries will appear below.


Proposed subprojects
 
PI: Mark Rodwell
Work at ECMWF will involve a comparison of ‘Initial Tendencies’ and ‘Transpose-AMIP’ (forecast bias out to D+5) methodologies for assessing a model’s representation of the physics (and dynamics).
Initially, we will repeat the experiments of Rodwell and Palmer (2007) with a (possibly less extreme but nevertheless erroneous) perturbation to the convective entrainment parameter (other choices left unchanged: e.g., resolution T159). We will document and attempt to explain the evolution of the impact of such a perturbation (relative to the control model) with initial tendencies and, over subsequent forecast lead-times, as the impact is ‘felt’ further-a-field such as in the strato-cumulus regions, or in the surrounding regions where strato-cumulus ‘transitions-to’ cumulus. The same perturbed model will also be initialized from a ‘non-native’ set of analyses (as in the Transpose-AMIP methodology, but produced here using the control model) to investigate the importance of using a native analysis. One could also delve further into the data assimilation system and diagnose, for example, the variational bias corrections applied to observations within some key cloud regions. The use of additional observations (e.g. CloudSat) could also be considered. Ultimately, the two main questions to answer are (1) ‘How do errors (in, e.g. strato-cumulus cloud and their radiative effects) develop over the course of a forecast?’ and (2) ‘Which experimental designs and diagnostics (initial tendencies, D+1, …, D+5 error, atmospheric model climate) successfully identify the model error?’

PI: Keith Williams
Follow an analysis methodology similar to Williams and Brooks (2008, J. Climate) to look at the development of errors in cloud regimes (derived from clustering ISCCP simulator data), and relate these to climatological errors in the AMIP simulation.  Other cloud observations (CloudSat, CALIPSO, etc.) with the associated simulator output from the models, will be used to gain a more complete understanding of the development of biases in the regimes.