Figure: A schematic diagram of the
4-dimensional variational data assimilation. Depicted are the longitude-time
structures of the sea surface temperature in the equatorial Pacific
(1996-1999), for (a) the simulation, (b) the observation and (c) the
assimilation.

Data assimilation is receiving international attention
as a new research method which can integrate observational data and
a numerical model of dynamics. One of the main focuses of our data
assimilation group is 4-dimensional variational data assimilation
method (4D-VAR). Our research widely covers the development of assimilation
system, the analysis of our own reanalysis data, and so on. The reanalysis
product derived from 4D-VAR is dynamically consistent 4-dimensional
dataset and therefore it is known to be a powerful material for a
scientific research such as dynamical analysis and also be an appropriate
initial condition for a forecast. Although there are not so many institutes
performing global 4D-VAR experiment because of its complexity and
its huge computational requirement, we have realized it by using the
"Earth Simulator" as the world greatest platform and are
now producing many fruit of research. We are now working on the assimilation
research targeting on seasonal to inter-annual scale climate change.