Chapter 24
ON THE USE OF DATA ASSIMILATION
IN BIOGEOCHEMICAL MODELLING
Andreas Oschlies
School of Ocean and Earth Science, Southampton Oceanography Centre, Southampton,
UK
Abstract
A main objective of applying data assimilation methods to marine ecosystem models is the optimisation of often poorly known model parameters
or even of the model’s functional form. Recent efforts in this direction
are reviewed. Results obtained so far indicate that presently available
data sets can constrain not more 10 to 15 different ecological parameters. This raises questions about the use of more complex models. On
the other hand, none of the optimised models yields a satisfactory fit to
the observations, suggesting that present ecosystem models are overly
simplistic. Implications of these apparently contradictory findings are
discussed and a data assimilative strategy for future improvement of
marine ecosystem models is suggested.
Keywords: Marine ecosystem models, data assimilation, parameter optimisation.
1.
Introduction
Interest in prognostic models of marine biogeochemical cycles arises
to a large extent from our need to better understand, quantify, and
eventually predict the ocean’s role in the global carbon cycle. This
cludes cycles of related elements, such as nitrogen, phosphorus or iron,
that can act as limiting nutrients for phytoplankton growth. Other
pects addressed by biogeochemical and ecological modelling include the
prediction of harmful algal blooms [Schofield et al., 1999], and a quantitative understanding of oceanic food webs up to fish [Loukos et al.,
2003], birds, and humans, as well as the possible impact of marine sulfur
emissions on the formation of cloud condensation nuclei [Gabric et al.,
2004]. In this chapter, I will focus on the carbon issue.
in
as
525
E.P. Chassignet Jacques Verron (eds.), Ocean Weather Forecasting, 525-547.
© 200 Springer. Printed in the Netherlands.
6
ON THE USE OF DATA ASSIMILATION
IN BIOGEOCHEMICAL MODELLING
Andreas Oschlies
School of Ocean and Earth Science, Southampton Oceanography Centre, Southampton,
UK
Abstract
A main objective of applying data assimilation methods to marine ecosystem models is the optimisation of often poorly known model parameters
or even of the model’s functional form. Recent efforts in this direction
are reviewed. Results obtained so far indicate that presently available
data sets can constrain not more 10 to 15 different ecological parameters. This raises questions about the use of more complex models. On
the other hand, none of the optimised models yields a satisfactory fit to
the observations, suggesting that present ecosystem models are overly
simplistic. Implications of these apparently contradictory findings are
discussed and a data assimilative strategy for future improvement of
marine ecosystem models is suggested.
Keywords: Marine ecosystem models, data assimilation, parameter optimisation.
1.
Introduction
Interest in prognostic models of marine biogeochemical cycles arises
to a large extent from our need to better understand, quantify, and
eventually predict the ocean’s role in the global carbon cycle. This
cludes cycles of related elements, such as nitrogen, phosphorus or iron,
that can act as limiting nutrients for phytoplankton growth. Other
pects addressed by biogeochemical and ecological modelling include the
prediction of harmful algal blooms [Schofield et al., 1999], and a quantitative understanding of oceanic food webs up to fish [Loukos et al.,
2003], birds, and humans, as well as the possible impact of marine sulfur
emissions on the formation of cloud condensation nuclei [Gabric et al.,
2004]. In this chapter, I will focus on the carbon issue.
in
as
525
E.P. Chassignet Jacques Verron (eds.), Ocean Weather Forecasting, 525-547.
© 200 Springer. Printed in the Netherlands.
6