Development of 4D Data Assimilation System and Construction of Data-base for Climate Research |
|
|
Toshiyuki Awaji (Group Leader, Data Assimilation Group, Integrated Modeling Research Program, Frontier Research System for Global Change) |
The main objective of this research is to construct an innovative fourdimensional data assimilation system capable of providing a high-quality comprehensive dataset, called "reanalysis dataset", by utilizing the maximum information content of both observing systems and full-coupled climate models. Though observations are still sparse in space and time, synthesis with the state-of-the-art general circulation models (GCMs) can produce a comprehensive 4-dimensional dataset with high accuracy and dynamical consistency, by taking the advantage that quantities of all variables are given at every grid point based on the model dynamics.
Such datasets are extremely required for more accurate prediction and analysis of global warming and hydrological cycle. This is because good reanalysis datasets work well as the initial values of prediction. Also, reanalysis datasets provide an attractive prospect for physical process studies of climate variation.
Data assimilation was developed in the context of numerical weather forecasting, and recent assimilation models are roughly classified into two categories; one for the statistical interpolation using the optimal interpolation (OI) method and the other for the dynamical interpolation using the variational method (VAR). Of these, variational assimilation models using GCMs are considered to be the most likely means of constructing dynamically consistent datasets.
However, the computational burden is quite heavy (at least, more than 100 times of simulation model's). Thus, the construction and operation of variational data assimilation systems covering the entire globe was difficult by using computational resources available so far.
The Earth Simulator (ES) could give a breakthrough for our limitations. The four-dimensional variational approach called 4D-VAR, which ensures the dynamical consistency of products not only in space but in time, can be applied to the assimilation system for climate study. It is anticipated that our advanced 4D-VAR assimilation system using both the adjoint method and the ensemble Kalman filter allows us to create a reanalysis dataset capable of improving prediction skills, dynamical analysis of climate change, and observing systems. One implication can be seen in Fig. 1.
|
Though the coupled model is not used in assimilation and forecast experiments, the forecast result using the analysis data as initial values (heavy solid curve) is better than before and the ensemble forecast result (dotted curves) provides important information on model predictability.
Our research is, in fact, a challenging task. However, we must keep in mind that in the MEXT project, an enormous research fund will be invested in the coming 5 years towards obtaining new findings regarding sustainable coexistence of human, nature and earth. This means that any research related to the MEXT Project needs to gain results that have scientific and societal impacts against the world after 5 years. Considering these facts and towards realizing the purpose of our project, a set of five sub-themes are built up, covering the construction of a value-added observational database, improvement of assimilation systems, and the building of an effective information delivery system.
Achievement of our goal requires a competitive partnership with domestic and foreign related institutions. Your criticism and encouragement are very much appreciated.
