A RESIDUAL ANALYSIS METHOD FOR QUANTITATIVE COMPARISON OF SPATIAL DATA FROM THE ICELAND-FAEROES FRONT

A spatial statistical method has been developed from the well-known Kr iging technique in geostatistics, as a way of providing quantitative c omparison between a pair of spatial data sets, and a measure for such a comparison. This residual analysis method is applied to oceanographi c data in order to compare Iceland-Faeroes Front (IFF) model predictio ns against appropriate field observations, with an aim to assess the I FF model performance and its prediction accuracy. The method is also u sed to evaluate the model-generated dynamical variability within the m odel predictions, as well as the natural variability within the fronta l observations. From the results, it has been found that the IFF model is highly robust, and gives better predictions at depth 150-350 m tha n elsewhere. Within such a depth (i.e. 250-350 m), there is also evide nce to suggest that the main frontal region is most active above the I F Ridge. The natural variability obtained from the observations appear s to be comparable to the model-generated dynamical variability after 20 days of integration, indicating a certain degree of accuracy in the model predictions. The method reported in this paper could also be ex tended for further use in model data assimilation. Thus, the work not only demonstrates how spatial statistics can be applied to oceanograph ic data, but also opens up new statistical tools for data handling in ocean modelling. (C) 1997 Elsevier Science Ltd.