Tsuyoshi Wakamatsu
Institute of Ocean Sciences, DFO Canada
A data assimilation system is underdevelopment at the Institute of Ocean
Sciences/DFO. The goals of the system are (i) to create the best estimate
of the Pacific Ocean state during last decade, (ii) to create optimal
initial value for the seasonal prediction study for the northeast Pacific
region. For the purpose of (i), we employed a smoothing technique, a
variation of the representer method developed by Andrew Bennett and
colleagues at the Oregon State University. Since the smoothing technique
enables us to estimate a temporally continuous flow field, it has
advantages over sequential data assimilation when water mass analysis is
required. The other advantage of this technique is that the optimal
solution is represented by a summation of representer vectors
corresponding to the contribution of each datum to the optimal solution.
We can evaluate which dynamical processes transfer observed information to
surrounding points, assuming that the large scale structure in a given
background flow field is close enough to the optimal solution. A Pacific
ocean box model with simple components is constructed as a testbed for the
data assimilation system. In this seminar, the following topics are
presented (i) details of the representer method and its system (ii)
observational system for the Pacific Ocean, and (iii) character of adjoint
fields computed for data located at Kuroshio extension, Alaskan gyre,
central equator and western US coast.