JPR Advance Access originally published online on June 13, 2006
Journal of Plankton Research 2006 28(9):877-878; doi:10.1093/plankt/fbl016
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
COMMENT |
Confronting complexity: reply to Le Quéré and Flynn
National Oceanography Centre, Southampton SO14 3ZH, UK
* Corresponding Author: tra{at}noc.soton.ac.uk
Received May 20, 2006; accepted in principle May 26, 2006; accepted for publication June 7, 2006; published online June 13, 2006
Communicating editor: R.P. Harris
Complexity pervades ecology. The search for unifying laws therefore entails that ‘biologists work very close to the frontier between bewilderment and understanding’ (Medawar, 1969
). One approach to developing ecological theory is to abandon simplicity at the outset and operate at the highest appropriate organizational level (Dunbar, 1980). On this basis, Le Quéré (Le Quéré, 2006) issues a plea for inclusion of plankton functional types (PFTs) in plankton models because ‘we will not understand ecology until we have built models that include the necessary processes’. From a purely a priori standpoint, my own view is quite the same: ‘The case for superseding; nutrient–phytoplankton–zooplankton–detritus (NPZD)-type models with those that include PFT is clear: biogeochemical cycling in marine systems is intimately linked to particular plankton groups if not in some instances primarily to individual plankton genera or species’ (Anderson, 2005).
The practical difficulties of representing PFTs as a series of interacting differential equations in models should not, however, be underestimated, the requirement being to demonstrate robustness of parameterization in the face of what is often inconclusive knowledge and scarcity of available data. In this context, Le Quéré claims that PFT models perform better-informed tuning than do NPZD models, which ‘rely solely on the reproduction of biogeochemical fields as a test of their performance’. NPZD modellers are effectively labelled instrumentalists whereby their models are no more than empirically adequate calculating devices employing abstract parameters with little or no basis in reality. In contrast, PFT models are elevated to a realist status, that is, having a basis in true laws of nature, parameter values having been ‘narrowed down’ by extensive laboratory work. This supposed dichotomy is mistaken. The equations of NPZD models very often have a good basis in reality, such as the use of photosynthesis–irradiance curves or Holling-type grazing responses. Yes, the parameters that go into these equations have to take account of the aggregate properties of the whole phytoplankton or zooplankton communities. But the same problem, namely that of aggregation, is fundamental to PFT models too. No amount of laboratory studies on Emiliania huxleyi will, for example, be conclusive in setting parameter values for a state variable for calcifiers, which includes numerous other species of coccolithophores, as well as foraminiferans (Anderson, 2005).
Putting aside the problem of aggregation for a moment, Flynn (Flynn, 2006) identifies difficulties at an even more fundamental level. It may not currently be possible, he suggests, to adequately parameterize a model of even a single algal species in laboratory culture, E. huxleyi surely being a case in point. Let us hope that this assessment is unduly pessimistic. Extending to the natural environment to include the maze of interactions such as competition between species and top-down control by grazers represents a formidable task. Providing examples such as the roles of allelopathy and food quality in trophic interactions, Flynn’s critique serves admirably to redouble my assertion that much of the devil is in the details of those interactions (Anderson, 2005), a worrying prospect for modellers.
The proof of the pudding, so to speak, is in the eating, and so rigorous validation lies at the heart of future progress. Recognizing the limited success of PFT models to date in this regard, Le Quéré tamely describes them as being ‘no worse’ than their NPZD counterparts. Given two empirically equivalent theories, scientists will usually deem the simpler one to be better constrained by available data and therefore more reliable. The challenge for PFT modelling is to demonstrate robustness in parameterization of the PFTs themselves, a mighty task indeed.
The strategy of model building is all about balancing complexity, accuracy and generality (Levins, 1966). Complex models tend to surrender predictive capability when applied to large or general domains rather than to specific case studies. Prediction is of course not everything. Complex models can be developed as heuristic tools, stressing unity, reality, cause and mechanism, an entirely worthwhile activity. The development of such models for specific sites of interest for which there exist comprehensive data is an obvious choice here. If, however, prediction is the desired end-point of inquiry, then one has to heed Peters’ (Peters, 1991) warning that opting for complexity ‘represents a pathology whereby the trappings of theory beguile us away from predictive power’.
The future direction of marine ecosystem modelling is unclear. Certainly, we should seek parsimony, that is, laws that are simplest in relation to the phenomena they are to explain, rather than simplicity for its own sake (Simon, 2001). However, acknowledging and confronting the unwelcome ramifications of complexity is first necessary to steer the way forward and avoid futile disappointments (Rescher, 1998). Diversity in approach, combined with rigorous model-data inter-comparison, is surely the direction to take, accompanied by an appropriate mix of open-mindedness, caution and scepticism.
 |
REFERENCES |
|---|
 TOP REFERENCES |
|---|
Anderson, T. R. (2005) Plankton functional type modelling: running before we can walk? J. Plankton Res., 27, 1073–1081.
Dunbar, M. J. (1980) The blunting of Occam’s Razor, or to hell with parsimony. Can. J. Zool., 58, 123–128.
Flynn, K. J. (2006) Reply to horizons article ‘plankton functional type modelling: running before we can walk’ Anderson 2005: II—putting trophic functionality into plankton functional types. J. Plankton Res., doi: 10.1093/plankt/fbl015.
Le Quéré, C. (2006) Reply to horizons article ‘plankton functional type modelling: running before we can walk’ Anderson 2005: I—abrupt changes in marine ecosystems? J. Plankton Res., doi: 10.1093/plankt/fbl014.
Levins, R. (1966) The strategy of model building in population biology. Am. Scientist, 54, 421–431.
Medawar, P. B. (1969) Induction and Intuition in Scientific Thought. Methuen, London, 62 pp.
Peters, R. H. (1991) A Critique for Ecology. Cambridge University Press, Cambridge, 366 pp.
Rescher, N. (1998) Complexity. A Philosophical Overview. Transaction Publishers, New Brunswick, 219 pp.
Simon, H. A. (2001) Science seeks parsimony, not simplicity: searching for pattern in phenomena. In Zellner, A., Keuzenkamp, H. A. and McAleer, M. (eds), Simplicity, Inference and Modelling. Keeping it Sophisticatedly Simple. Cambridge University Press, Cambridge, pp. 32–72.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||