T.D. Reid and N.M.J. Crout (UK)
Antarctica, lake ice, model selection criteria
Model selection criteria are proposed as useful tools for assessing the optimum complexity and predictive power of a model. However there are few examples of these techniques being applied to real-data situations. In this study a number of the criteria were calculated for a complex thermodynamic model and simple empirical models of freshwater lake ice, in conjunction with data from a year-long study on Crooked Lake, Antarctica. The simple models captured the system behaviour remarkably well, although the selection criteria suggested the physics based approach was more generalizable. In addition, the criteria gave convincing evidence for ice modellers that using variable physical ice characteristics (albedo, extinction coefficient and geometric surface roughness) improves the accuracy and generalizability of a physics based ice model.
Important Links:
Go Back
