NEPIP - Contents

Critical Parameters Must Be Measured

Successful modeling of Southern Ocean biogeochemistry requires measurements of as many critical rate processes as possible, ideally on a size-and-taxon specific basis. There is often a mismatch between the quantities measured by field programs and those required for biogeochemical models. This can be illustrated by a cursory comparison of the JGOFS core measurements versus the model parameters and variables for a simple ecosystem model. Table 1 lists the JGOFS core measurements as well as the variables in the 7-compartment ocean ecosystem model developed by Fasham et al. (1990). In many cases, the model requires data that are not measured by JGOFS such as specific growth rates. This disparity between the observational and numerical views arises for several reasons.

JGOFS Core Measurements (General Categories)	Fasham Model Variables and Parameters with an Analog to JGOFS Core	Fasham Model Variables and Parameters with No Analog to JGOFS Core
Meteorological variables	Cloudiness, PAR
Temperature		Cross-thermocline mixing rate
Salinity
Dissolved oxygen
Photic zone depth	Light attenuation due to chlorophyll
Beam transmission	Light attenuation due to water
NO₃	Nitrate	Half-sat. rate for nutrient uptake
NO₂
NO₃ + NO₂
NH₄	Ammonium	NH₄ inhibition
Urea	Labile DON	NH₄/DON uptake ratio
Phosphate
SiO₂
Alkalinity
TCO₂
pCO₂
POC
PON
DOC
Chlorophyll
Bacterial abundance	Bacterial abundance	Bacterial maximum growth rate
Cyanobacteria abundance		Bacterial specific excretion rate
Bacterial production		Bacterial half-sat. rate for uptake
Phytoplankton abundance	Phytoplankton abundance	Initial slope of the P/I curve
Primary production by ¹⁴C		Phytoplankton exudation fraction
Primary production by O₂		Phytoplankton maximum growth rate
New production by ¹⁵N		Phytoplankton specific mortality rate
Mesozooplankton abundance	Zooplankton abundance	Zooplankton maximum growth rate
Mesozooplankton grazing	Zooplankton half-sat. rate for ingestion	Zooplankton specific mortality rate
Mesozooplankton egestion	Zooplankton specific excretion rate	Zooplankton assimilation efficiency
Microzooplankton abundance		Detrital fraction of zooplankton mortality
Microzooplankton grazing		Ammonium fraction of zooplankton excretion
Sediment trap quantities	Detritus	Detrital breakdown rate
		Detrital sinking rate

Table 1 Comparison of JGOFS Core measurements with variables and parameters that are used in the ecosystem model of Fasham et al. (1990)

Time evolving models by necessity focus much attention on rates of processes (e.g., phytoplankton growth, grazing) while a large fraction of the available field data involves measurements of biological standing stocks. In part, this is due to the inherent difficulty of constraining transformation rates in the ocean, but the result is that the magnitude of key processes must be either extrapolated from laboratory work or inferred from small net changes in standing stocks. The issue of model-data comparison is further complicated by the fact that the model representation of a particular variable may be only a close cousin of the actual measured quantity. A primary example of this is the difference between observed photosynthetic assimilation rates (mg C/mg Chl/day) and modeled phytoplankton specific growth rates (mg C/mg C/day). The two quantities are linked via the phytoplankton Chl:C ratio, which is often not measured due to time and/or expense. In addition to rate measurements, models typically require the specification of a set of functional responses; for example how do phytoplankton growth and zooplankton grazing rates vary, respectively, with nutrient and prey abundance? Although culture experiments can guide the form of the model parameterizations, the appropriate parameter values for a particular region may be unknown. Deducing the functional relationships from field data may involve a range of conditions beyond that naturally observed, thus requiring small volume manipulation experiments at sea. Finally, model closure often forces modelers to include terms, such as phytoplankton mortality, that either lack a strong biological basis or are an amalgamation of a variety of processes.

In general, these terms are unconstrained by observations and are set by "tuning" the model to observations. Unfortunately, as was recently demonstrated for zooplankton mortality (Steele and Henderson, 1992), the treatment of these closure terms can dramatically alter the behavior of an ecosystem model.

If JGOFS sets as an objective the development of data assimilation models, then in addition to collecting information on key processes, it must also ensure that the sampling strategy is adequate to meet the requirements for assimilation. Lawson et al. (1995; 1996) discuss the performance of the adjoint method to recover model parameters as applied to data from the Bermuda Atlantic Time Series (BATS) station. Their conclusions bear strongly on this report. First, measurement of rate parameters is essential as these are a fundamental aspect of the dynamics of the ecosystem. Second, many of the bulk measurements collected by JGOFS (such as phytoplankton chlorophyll) are not produced directly by the numerical model. Thus information relating variables such as phytoplankton nitrogen to chlorophyll would be valuable. Third, sampling should be frequent enough to ensure that there remains adequate correlation between successive measurements, otherwise the assimilation process will fail. Lastly, the assimilation process is more robust if all variables are sampled at the appropriate critical frequency, which may differ for each variable.