311
The large variety of sequential statistical methods developed in the
context of scientific or operational applications is an indication that a
given assimilation technique cannot be considered as a “plug-and-play”
system, capable of solving universal problems. To design the best possible assimilation system, it is necessary to clearly define the purpose of
data assimilation (forecast initialization, reanalysis, model error detection, etc.), the physical characteristics of the processes of interest, the
sampling properties of the observation systems, and the limitations of the
assimilation techniques. Moving towards hybrid sequential/variational
methods is probably an astute way of taking the best from both approaches. Due to the complexity of models and algorithms, the success
of an assimilation system also depends on the community of people working within a common framework. In the future, the sharing of generic
assimilation tools between operational and research teams should accelerate the progress of the methods and provide feedback from more
intensive utilizations.
A number of important issues for ocean data assimilation have yet to
be fully resolved, such as the incorporation of inequality constraints in
statistical estimation algorithms. Such constraints are needed, e.g. in
the context of tracer data assimilation to maintain the positiveness of
tracer concentrations, or for the proper handling of the physical properties of the water column such as static stability. Another important
challenge will be to further develop assimilation systems for coupled
physical-biological models, with the aim of demonstrating the capacity
to estimate and forecast marine ecosystems and biogeochemical variables
in the ocean on a routine basis. As a first step, the incorporation of
biogeochemical models into the physical assimilative systems developed
within GODAE will provide new “bio-metrics” to evaluate the representation of the physical processes of critical importance to biology. A
further challenge will be to implement suitable methods to assimilate
ocean colour data in the coupled assimilative systems.
Acknowledg ments
The author gratefully acknowledges his colleagues from LEGI and
MERCATOR Ocean for numerous discussions and valuable suggestions.
He is supported by the Centre National de la Recherche Scientifique.
Computations were carried out at the IDRIS/CNRS computing centre.
The contribution of students from the GODAE school who commented
an earlier version of these lecture notes has been much appreciated.
This work has been partly supported by the MERSEA project of the
European Commission under Contract SIP3-CT-2003-502885.
OCEAN DATA ASSIMILATION
e
The large variety of sequential statistical methods developed in the
context of scientific or operational applications is an indication that a
given assimilation technique cannot be considered as a “plug-and-play”
system, capable of solving universal problems. To design the best possible assimilation system, it is necessary to clearly define the purpose of
data assimilation (forecast initialization, reanalysis, model error detection, etc.), the physical characteristics of the processes of interest, the
sampling properties of the observation systems, and the limitations of the
assimilation techniques. Moving towards hybrid sequential/variational
methods is probably an astute way of taking the best from both approaches. Due to the complexity of models and algorithms, the success
of an assimilation system also depends on the community of people working within a common framework. In the future, the sharing of generic
assimilation tools between operational and research teams should accelerate the progress of the methods and provide feedback from more
intensive utilizations.
A number of important issues for ocean data assimilation have yet to
be fully resolved, such as the incorporation of inequality constraints in
statistical estimation algorithms. Such constraints are needed, e.g. in
the context of tracer data assimilation to maintain the positiveness of
tracer concentrations, or for the proper handling of the physical properties of the water column such as static stability. Another important
challenge will be to further develop assimilation systems for coupled
physical-biological models, with the aim of demonstrating the capacity
to estimate and forecast marine ecosystems and biogeochemical variables
in the ocean on a routine basis. As a first step, the incorporation of
biogeochemical models into the physical assimilative systems developed
within GODAE will provide new “bio-metrics” to evaluate the representation of the physical processes of critical importance to biology. A
further challenge will be to implement suitable methods to assimilate
ocean colour data in the coupled assimilative systems.
Acknowledg ments
The author gratefully acknowledges his colleagues from LEGI and
MERCATOR Ocean for numerous discussions and valuable suggestions.
He is supported by the Centre National de la Recherche Scientifique.
Computations were carried out at the IDRIS/CNRS computing centre.
The contribution of students from the GODAE school who commented
an earlier version of these lecture notes has been much appreciated.
This work has been partly supported by the MERSEA project of the
European Commission under Contract SIP3-CT-2003-502885.
OCEAN DATA ASSIMILATION
e