There is close similarity between the models used for work on medium and extended range forecasting and the general circulation models (GCMs) used for climate research. Both modelling communities rely heavily for model validation on the extensive climate datasets prepared by national and international agencies under the aegis of the WMO/ICSU. However there are few systematic methods to identify the sources of problems in long runs of a general circulation model, because almost all model errors are fully developed and fully interactive. It is much easier to diagnose errors (say in a parameterization scheme) when they grow in an otherwise accurate series of forecasts that start from accurate analyses. This diagnostic approach, which stems from the operational data assimilation systems and from verifications of operational forecasts, has provided weather prediction modellers with powerful tools to identify missing processes in the model and to refine the representations of well-known processes, The approach is currently being adapted to climate community needsthrough initiatives such as the adjoint-AMIP or CAPT initiatives, where climate models are run in forecast mode from sequences of reanalyses and verified against operational and field experiment data.
Four-dimensional variational assimilation systems provide accurate analyses and reanalyses of all available in situ and satellite data, and result in forecasts of remarkable quality. The accuracy of the operational analyses, and of extended reanalyses, is of tremendous assistance in the diagnosis of both NWP and GCM model problems.
The variety of data resources available for the development of forecast models may be contrasted with the much smaller data resources available for direct validation of simulations of any climate other than the present climate. One can have confidence in simulated climate scenarios only if one has confidence in the physical formulations and feed-back loops of the GCMs. A strong case could be made that every GCM should be equipped with a data assimilation system, so that one can diagnose its performance with field experiment data and in medium- and extended-range forecasts. Such diagnosis is bound to provide penetrating insights on how to improve the physical formulations of the GCMs. Indeed one may question the value for climate change scenarios of a model which performs poorly in medium or extended range forecasts. The CAPT approach may be viewed as a first step in that direction.
Striking convergences in the development of forecast and GCM methodologies are occurring in two areas. One is in shared development of ensemble methodologies to indicate the predictability of medium-range / seasonal forecasts on the one hand, and to assess the degree of consensus between many different GCMS on projections of future climate for a particular period. Consensus among models may indicate reliability, provided the models do not share common weaknesses, such as a poor representation of blocking phenomena.
The second area of convergence is the extension of current forecast capabilities to include atmospheric composition as well as atmospheric dynamics and thermodynamics. Here the forecast community is learning a lot from the GCM and CTM communities, as well as facing a challenging array of problems in the assimilation of observations of greenhouse gases, reactive gases and aerosols.