DATA ASSIMILATION IN BIOGEOCHEMICAL MODELLING
529
solved dead organic matter which does not sink but moves passively
with the water. Besides the different transport properties, the distinction among dissolved and particulate organic matter is also useful in
terms of elemental ratios: while the elemental composition of particulate organic matter is, on average, found to be close to the Redfield
ratio [Redfield, 1934; Redfield et al., 1963], dissolved organic matter
ten contains several times more carbon than the Redfield ratio would
predict [Williams, 1995; K¨ ahler and Koeve, 2001].
Using mass conservation as underlying concept, the compartments
simulate stocks of atoms of the relevant element, and fluxes such as primary production, grazing, or mortality all describe the transfer of atoms
among the different compartments. Often only a single element (usually one associated with a potentially limiting nutrient, e.g., nitrogen
for nitrate, phosphorus for phophate) is modelled explictly, and its concentration in each of the compartments becomes a prognostic variable.
Concentrations and fluxes of other elements (in particular, carbon) are
usually diagnosed via the Redfield ratio. While this seems to be consistent with the analysis of average inorganic remineralisation products
[e.g., Anderson and Sarmiento, 1995], more detailed investigations
veal local and temporal systematic deviations [K¨ ortzinger et al., 2001;
Sterner and Elser, 2002; Klausmeier et al., 2004]. A few recent models
have therefore begun to explicitly resolve the cycling of different elements
[Moore et al., 2002].
For each marine ecosystem model, the particular choice of its compartments and of the parameterisation of fluxes between the compartments
contains subjective elements, which may for example be influenced by
operational measurement protocols, historical paradigms, or taxonomic
nomenclature. Such an approach is, of course, valid and probably necessary in a field in which a strong theoretical framework is not yet available
(and in which key species may not even be discovered yet). Progress will
be made by trying to construct models that can explain the observations
and at the same time tell a plausible story, and by more or less steadily
changing the story as new observations add new information. In this
process it is important to keep in mind that the underlying rules that
make up a particular ecosystem model are generally assumed rather than
demonstrated and hence are subject to change.
After these cautionary remarks about the theoretical foundations of
marine ecoystem models, it is time to point out that these very models
may greatly help to improve our understanding of marine ecosystems by
allowing us to test the assumed hypothetical rules against observations
in a quantitative way. In the following I will try to present my subjective
view of how this can be achieved in practice.
dis
of
re
529
solved dead organic matter which does not sink but moves passively
with the water. Besides the different transport properties, the distinction among dissolved and particulate organic matter is also useful in
terms of elemental ratios: while the elemental composition of particulate organic matter is, on average, found to be close to the Redfield
ratio [Redfield, 1934; Redfield et al., 1963], dissolved organic matter
ten contains several times more carbon than the Redfield ratio would
predict [Williams, 1995; K¨ ahler and Koeve, 2001].
Using mass conservation as underlying concept, the compartments
simulate stocks of atoms of the relevant element, and fluxes such as primary production, grazing, or mortality all describe the transfer of atoms
among the different compartments. Often only a single element (usually one associated with a potentially limiting nutrient, e.g., nitrogen
for nitrate, phosphorus for phophate) is modelled explictly, and its concentration in each of the compartments becomes a prognostic variable.
Concentrations and fluxes of other elements (in particular, carbon) are
usually diagnosed via the Redfield ratio. While this seems to be consistent with the analysis of average inorganic remineralisation products
[e.g., Anderson and Sarmiento, 1995], more detailed investigations
veal local and temporal systematic deviations [K¨ ortzinger et al., 2001;
Sterner and Elser, 2002; Klausmeier et al., 2004]. A few recent models
have therefore begun to explicitly resolve the cycling of different elements
[Moore et al., 2002].
For each marine ecosystem model, the particular choice of its compartments and of the parameterisation of fluxes between the compartments
contains subjective elements, which may for example be influenced by
operational measurement protocols, historical paradigms, or taxonomic
nomenclature. Such an approach is, of course, valid and probably necessary in a field in which a strong theoretical framework is not yet available
(and in which key species may not even be discovered yet). Progress will
be made by trying to construct models that can explain the observations
and at the same time tell a plausible story, and by more or less steadily
changing the story as new observations add new information. In this
process it is important to keep in mind that the underlying rules that
make up a particular ecosystem model are generally assumed rather than
demonstrated and hence are subject to change.
After these cautionary remarks about the theoretical foundations of
marine ecoystem models, it is time to point out that these very models
may greatly help to improve our understanding of marine ecosystems by
allowing us to test the assumed hypothetical rules against observations
in a quantitative way. In the following I will try to present my subjective
view of how this can be achieved in practice.
dis
of
re