gfdl's home page > about us > FY02 milestones > implement an improved FMS-based GFDL seasonal forecast system
Implement an Improved FMS-based GFDL Seasonal Forecast System
3rd Quarter GFDL Milestone - FY2002
Purpose:
The focus of GFDL's coupled model development has been to develop and implement the next generation of coupled forecast models capable of simultaneous studies of climate variability and change. The consistently most skillful component of the operational seasonal forecasts at the Climate Prediction Center is an empirical forecast tool for long-term trends. Initial results suggest that this new model is capable of simulating and forecasting trends resulting from stratospheric - tropospheric interactions, the possible interactions between natural climate variability and change resulting from anthropogenic effects and volcanic aerosols, and assessing the impacts of variability and change on hydrology, zero order terrestrial and oceanic ecosystems, and Arctic ice and snow processes. Existing seasonal forecasting models are not capable of exploring these
Efforts:
Over the past few years GFDL has been engaged in the development of state-of-the-art atmosphere, ocean, sea ice, and land models coded using modern software standards and designed for the new generation of massively parallel computers. The culmination of this focused team effort is the coupling of all of these component models under a common modeling infrastructure, GFDL's Flexible Modeling System (FMS). Public releases of the well documented FMS infrastructure and two ocean models occurred earlier this year and were reported on in earlier milestones. Public releases of other component models will follow. We anticipate that the new software infrastructure and the public releases of it and component models, will facilitate collaboration between GFDL and the broader modeling community to accelerate development of further improved climate models.
An important element for the seasonal and possible decadal forecasting is the initialization of the ocean component. This is achieved through the merging of ocean observations with model simulations using ocean data assimilation (ODA). The GFDL ODA system has also been completely redone to enhance FMS to support requirements of ODA codes on scalable computing platforms as well as the creation of standard interfaces to MOM4, data handling and quality control. Three different assimilation techniques, 3D-Variational, ensemble filtering, and 4D-Variational, are in various stages of implementation to test their utility in climate forecasting and ocean analyses.
Customers:
The end customers are individuals engaged in policy and management decisions. This model will be used for performing the standard IPCC scenarios and for policy specific scenarios exploring the co-evolution of possible technology/energy changes, with changes to climate and the impacts these would produce. An aspect of documenting the credibility of the model is in its ability to simulate present and past climate and to be used operationally for seasonal forecasting. As the model's skill and utility are demonstrated, it will be used as a tool for operational seasonal forecasting by the NWS hence providing input into management decisions.
Significance:
This GFDL coupled climate forecast system represents the next step toward production of new understandings and advances in seasonal-to-interannual forecasts. It goes beyond ENSO forecasting (current modeling capabilities) to addressing issues of the connections between climate change and variability, understanding of the origins of persistent trends which constitute a major contribution to skills in CPC forecasts, and exploring the possibility of forecasting interannual to decadal changes in midlatitude circulations (i.e. the Arctic Oscillation (AO)). ENSO only accounts for less than half the observed climate fluctuations over the U.S.; the AO accounts for about the same amount.
Success:
The primary success is that the component models have been combined into a running coupled model under FMS. Also the capability to initialize the seasonal forecasts through ocean data assimilation has also been implemented under FMS. A number of long coupled simulations with no flux adjustments have been run. The model exhibits ENSO variability, however the amplitudes of the tropical variability and midlatitude impacts are too weak. Work is underway to improve these shortcomings. Several novel approaches have been successfully incorporated in the component models. One example is that poleward of 60°N the ocean grid becomes bi-polar with the coordinate singularities over land regions, thus there is no need for high latitude filtering which greatly improves the Arctic ice simulations (see figure).
Next Steps:
Producing a coupled model that has both a good mean climate and is state-of-the-art in simulating climate variability and change is indeed daunting. An ensemble of 50-year runs of the atmospheric component forced by observed ocean temperatures is currently being run. This provides the a priori skill estimates so that the model can be used in the seasonal forecasting procedures for surface parameters at NCEP and the IRI. A twenty-year ocean reanalysis using a one degree MOM4 ocean model and the FMS version of the ocean data assimilation system will be completed this fall/winter. This data set will be used to initialize of the GFDL coupled model to ascertain its skill for ENSO forecasting starting this fall. It is also being made available to the broader U.S. ENSO forecast community for collaborative studies in improving seasonal forecasts.