Validation of model simulations with respect to in situ observations by the use of probabilistic estimations

The present work addresses the problem of validation of a synthetic dataset with respect to observations. It gives an index that determines locally how much a region of the synthetic dataset fits the observations. The method uses an estimation of the probability density function computed with the probabilistic self-organizing maps. Then, an index F was defined to quantify the areas of the synthetic datasets that correspond to the observations. The method was first applied to a ldquotoyrdquo example in 2 dimensions to see its potentiality and then applied to two real datasets of optics measurements of the surface ocean. The method allowed to characterize some simulations that have not been encountered during ship campaigns.