Biogeographic validation of a global ocean biogeochemical model

Currently biogeochemical models of the global ocean focus on simulating the coupling between prevalent physical conditions and the biogeochemical processes with the underlying assumption that coherent biological properties are a direct (or modulated) response to physics. This is one possible biogeographic characterisation of the pelagic environment, since biogeochemistry represents only one aspect of marine ecosystems. Several models are currently capable of simulating the chlorophyll distribution observed from space, though an objective validation with respect to relevant ecosystem properties is still lacking.
In this paper we analyse the results of one of the most comprehensive models of ocean biogeochemistry with an emphasis on biogeographic validation sensu Longhurst (Ecological Geography of the Sea, 2007, 2nd edition, Academic Press). A set of multivariate statistical tools, Multi Dimensional Scaling (MDS) and Principal Components Analysis (PCA), are used to verify the existence of pre-defined biogeographic provinces and their statistical significance. The MDS ordination indicates that the given provinces are recognizable in the model on the basis of the selected variables. Analysis of Similarity (ANOSIM) shows that the provinces are statistically separable and they can be more easily distinguished in terms of their environmental features rather than their biology. The underlying relationships between the physical and biological properties are investigated through correlation analyses, thus providing some insights on how the model reproduces features characteristic of the various regions.
Satellite chlorophyll data have been used to demonstrate external validation at the biogeographic level. The a priori provinces as characterised by chlorophyll values cannot be statistically separated in either the data or the model. It is likely this is related to the arbitrary choice of province boundaries, which are not necessarily the same as those derivable from non-interpolated SeaWiFS data.
The PCA comparison of modelled and observed chlorophyll demonstrated some objective skill in the model as it generally captures the dominant mode of the data, although severe mismatch was identified in certain regions by visual comparison (Indian and Southern Oceans). The model also overestimated seasonal variability compared to the data. The method shows promise for helping overcome problems with model verification due to undersampling of most ocean biogeochemical variables.