Monday, 10 January 2005: 4:30 PM
Representing land in a global climate model: simplicity vs. fidelity
The components of the climate system, including the continental lithosphere and biosphere, are complex. Therefore, a detailed land model has potential for higher fidelity (i.e., accuracy in its representation of land) than a simple model. Hence, the modeler is faced with a tradeoff between simplicity and potential fidelity. According to one caricaturing viewpoint, all modelers can be divided into two camps. On one side are those who argue for inclusion of everything but (and somtimes including) the kitchen sink. On the other side are those who are ignorant of the importance of myriad processes for the situation at hand. In reality, few modelers fit either of these descriptions, though many show tendencies in one (or even both) of these directions. A more nuanced viewpoint holds that the degree and nature of detail included in a model should depend on the problem to which the model will be applied. However, if a model is to be used for a wide range of problems, by this argument, it should at least be capable of including much detail. A desire to ensure flexibility in future model applications, therefore, can lead us back to the 'kitchen sink' model. Trends in the power of computing resources tend to ratchet us in the same direction, sometimes with good reason. Social pressures that arise in the peer-review process have a similar effect. Furthermore, increased detail may increase potential fidelity without increasing actual fidelity; the latter may even be degraded. Known constraints on system behavior are more readily applied to simple models than to some detailed models. Added detail adds risk of errors in coding and implementation. Model detail also has the potential to generate a false sense of confidence in the model. On the interpretive front, model detail can sometimes cloud our view of essential physical mechanisms, with the trees preventing us from seeing the forest. In view of the contemporary tendency for modeling to drift in the direction of increasing detail, the interest of intellectual discourse benefits when a competent scientist has the audacity to stick out his or her neck and "assume a spherical cow."