Synthesis and Modeling Working Group Summary

Models have several uses, not all of which are obvious. A principal use of models is to sharpen hypotheses by checking the logical consistency of postulated mechanisms. Models can aid in analyzing data, both in real time and in retrospective analyses, using data assimilation and other techniques. Models can be used to generate global syntheses and to extrapolate (project) into the future (or past), when conditions may differ substantially from those of the present. To be believable, these models must be based on underlying mechanisms that are thought to determine the key interactions. Models are also useful for evaluating variability on a variety of space and time scales.

Sociological Issues

A central organizing principle of OCTET should be close collaboration between modelers and non-modelers, and this collaboration should extend (in any particular study) from the formative stage, through the period of data collection and analysis, and into the synthesis and modeling stage. The justification for this close collaboration is that an essential output from OCTET will be the formulation of algorithms that allow better prediction of the effects of climate change on ocean processes, and of the feedbacks of these changes to climate. This goal is likely to be met to a greater degree if data collection is designed with algorithm development in mind, so that data is collected in a manner that insures its usefulness in modeling. Conversely, close collaboration will insure that any models that are developed will be based on the best current understanding of the ocean carbon system as described by non-modelers.

To insure this collaborative atmosphere, misperceptions about the place of modeling in the scientific endeavor must be overcome. "Modelers" create and use analytic and numerical mathematical models to understand the implications of mechanisms and empirical generalizations, the goal being both improved understanding and improved skill at numerical prediction. "Non-modelers" also use models, but because these tend to be verbal, pictorial, or statistical (e.g., regression analysis), they are often not recognized as such. The distinction between modelers and non-modelers is therefore one of technique, not of goal. Bringing all available techniques to bear in all phases of OCTET studies has the potential to increase substantially the rate of progress in our field

Strategy: Three specific suggestions were made concerning mechanisms for bringing modelers and non-modelers together.

1. Joint participation by both modelers and non-modelers on OCTET proposals should be encouraged. This mechanism was begun in JGOFS SMP and is also strongly encouraged in NSF Biocomplexity proposals, and seems to be leading to a better working relationship between modelers and non-modelers.
   
   
2. Participation of modelers in data-gathering exercises, such as cruises, should be encouraged. The inclusion of modelers who use real-time satellite data may directly enhance sampling efficiencies. However, it should also be valuable to include more process-oriented modelers, so that their insights and confusions in using data to construct better algorithmic representations of reality can be confronted in real time.
   
   
3. Courses could be created to introduce non-modelers to modeling; this would introduce non-modelers to the requirements and thought processes of modeling, in much the same way that including modelers in data-collection would improve their intuitive appreciation of data.
   

Modeling Issues

The current generation of numerical ocean biogeochemical models are based almost entirely on the aggregated N-P-Z (nutrient, phytoplankton, zooplankton) box model framework dating back at least several decades in aquatic ecology. Model advances over the time period of OCTET can be expected through the extension and sophistication of these techniques (e.g., multi-nutrient limitation; plankton functional groups; more explicit dissolved organic matter/microbial interactions; eddy resolution; data assimilation). Progress may also depend on the implementation of more novel, ecologically based approaches.

The modeling challenges for OCTET will involve issues including (but not limited to):

• Development of improved algorithms for large-scale simulations should be a stated goal, and the success of OCTET should be measured by how much algorithms are improved. This requirement alone will encourage the collaboration of modelers and non-modelers.
  
  
• Species compositions may change in response to climate change, altering stoichiometric ratios among biogenic elements; in turn, changes in nutrient abundances will feed back to community structure. Existing food web models have for the most part been developed to represent changes on shorter time scales; a gap exists between these models and models that possess the taxonomic complexity and numerical efficiency to be used on longer time scales, where species replacements may be of paramount importance. Novel approaches are needed to formulate efficient biogeochemical models that incorporate functional diversity (diatoms, coccolithophorids, Phaeocystis, photosynthetic prokaryotes, N-fixers). A key question is the level at which functional taxonomic groups will need to be resolved.
  
  
• Physical models must be developed with explicit consideration of their use in biogeochemical simulations. Physical models that seem to work well for temperature and salinity may nonetheless not behave well with biogeochemical tracers. Related problems concern the adequacy of present forcing fields (e.g., winds) and with air/sea and sea/ice couplings.
  
  
• A persistent problem facing modelers is one of scale: how can results of small process-oriented studies be extrapolated to larger scales?
  1. Can satellite imagery serve as the basis for this extrapolation?
  2. How must mechanistic parameterizations that make sense on small space and time scales be changed for use in GCMs with large grid boxes, limited vertical resolution, and uncertainty in parameterizations of small-scale physical forcings, such as turbulence?
  3. The ocean is not one-dimensional: how do we analyze data at depth when its source is not the surface directly overhead? At what spatial scales does this matter for large-scale simulation?
     
     
• The objective fusion of data and models remains a key issue in oceanography, particularly for biological and biogeochemical models where large-scale data assimilation is still in its infancy. With the expected long-term availability of satellite ocean color imagery and the rapid development of autonomous in-situ samplers, sufficient data may be available to generate reasonable ocean biogeochemical state estimates, at least for key surface ocean properties (e.g., autotrophic biomass, productivity, new production, sea surface pCO₂). Many technical and scientific issues exist, however, including:
  1. What are the time/space scales of biological variability?
  2. What are the tradeoffs between measurements of extensive (e.g., satellite chlorophyll) and intensive (e.g., size class structure; grazing rates) properties?
  3. How best can one define the dynamic relationships among the ecosystem variables such that assimilation of one observable quantity (e.g., chlorophyll) projects onto other, unobserved ecosystem compartments (e.g., bacterial and zooplankton biomass)?
     
     
• The biogeochemical modeling treatment of the subsurface water column has until now been primarily empirical in nature (e.g., the Martin et al. particle flux curves). This situation should improve with an increased understanding of the mechanisms involved and the use of inverse modeling.