Consistency and complementarity of different coastal ocean observations: A neural network-based analysis for the German Bight

Abstract

HF radar measurements in the German Bight and their consistency with other available observations were analyzed. First, an empirical orthogonal function (EOF) analysis of the radial component of the surface current measured by one radar was performed. Afterwards, Neural Networks (NNs) were trained to now- and forecast the first five EOFs from tide gauge measurements. The inverse problem, i.e., to forecast a sea level from these EOFs was also solved using NNs. For both problems, the influence of wind measurements on the nowcast/forecast accuracy was quantified. The forecast improves if HF radar data are used in combination with wind data. Analysis of the upscaling potential of HF radar measurements demonstrated that information from one radar station in the German Bight is representative of an area larger than the observational domain and could contribute to correcting information from biased observations or numerical models.