Measure-Correlate-Predict techniques to improve long term numerical datasets: application to wave potential estimate

Abstract

Abstract:Many long-term numerical databases (LTND) are distributed by research centers from various countries (like Previmer, NCEP, ..). These LTND usually provide integrated wave parameters which enable the computation of operational and also extreme statistics thanks to their duration (generally between 10 and 30 years). Although these LTND are validated in several places, it is rarely the case at your location of interest and therefore data extracted from a LTND doesn't necessarily offer the best results at this point. Nevertheless, it is possible to improve the results of these LTND in a "post-processing" step, ie without re-running the numerical simulations with data assimilation. This paper will present different techniques of Measure-Correlate-Predict (MCP) to improve the results of these LTND using in-situ data. Some of these techniques are then implemented to modify a long-term wave database in northern Bay of Biscay in France. The effect of this change is evaluated in terms of significant wave heights and wave peak period, and also in term of theoretical wave energy potential calculated using the deep waters and JONSWAP approximations, by comparing the obtained results with insitu measurements. A discussion is then opened on the way of propagating these local corrections and on the way of making it on larger spatial extents and on different physical types of LTND (wave, current, wind), thanks to the Metocean Analytics software (http://www. openocean.fr/en/metocean-analytics/).