Spatial Scales of Skill of Marine Biogeochemical Models
Saux Picart, Stephane1; Allen, Icarus1; Racault, Marie-Fanny1; Butenschon, Momme1; De Mora, Lee1; Yool, Andy2; Buitenhuis, Erik3; Torerdell, Ian4

Coupled hydrodynamic- ecosystem models are increasingly being used for various application including climate studies, policy making, fisheries... At global scale, models are particularly useful for climate or anthropogenic impact studies. Despite much effort directed toward the validation of these models using various data sets (from in-situ measurements to satellite remote sensing) and techniques, very little has been done to evaluate the spatial scales at which a model performs best.

  In the framework of the iMarNet project (Integrated Marine Biogeochemical Modelling Network), five different biogeochemical models (having various degree of complexity) coupled with the same hydrodynamic model have been run in the same conditions. These runs are performed at global scale and provide a unique opportunity to establish which are the strength and weakness of each of the biogeochemical components. Composites of remotely-sensed data of chlorophyll provided by the ESA - Ocean Colour Climate Change Initiative project have been used to derive diatom fraction and to compare against model outputs.

  A wavelet decomposition technique is used to estimate a skill score for a range of spatial scales and a range of quartile ranges. The skill score used here is directly comparable between models, scales and time. This allows us to determine both spatial scales of model skill and whether or not this is entirely determined by the physical configuration or if the complexity of the biogeochemical model has an influence.  Whilst all models are expected to have some degree of skill in reproducing chlorophyll pattern at the basin scale, the results of the comparison are much more uncertain at smaller scales.