Use of International Hydrographic Organization Tidal Data for Improved Tidal Prediction

Abstract

Tides are the rise and fall of water level caused by gravitational forces exerted by the sun, moon and earth. Understanding sea level variation and its impact currents is very important especially in coastal regions. With knowledge of the tide-generating force and boundary conditions, hydrodynamic models can be used to predict or model tides in coastal regions. However, these models are not sufficiently accurate, and in-situ tide gauge data may be used to improve them in coastal regions. The International Hydrographic Organization (IHO) tidal data bank consists of over 4000 tide gauge stations scattered all around the globe, most of which are in coastal regions. These tide gauge data are very valuable for tidal predictions. One drawback of the IHO data is that a considerable number of stations are located in rivers or near man-made structures or small-scale, complex topographic features. Another drawback is the unknown accuracy of the IHO data. To avoid these drawbacks, quality control has been done in the present study. Each IHO tide gauge station has been categorized according to its proximity to rivers, lagoons, man-made harbors, and other factors that may influence tidal elevation. Quantitative metrics such as water depth, distance to the continental shelf break, and horizontal length scale of station site morphology have been computed. Comparisons among IHO data, the output of O.S.U. Tidal Inversion Software (OTIS), and other data sources, such as Global Sea-Level Observing System (GLOSS) data, have been done to test the quality and accuracy of IHO data. Moreover, the characteristics of stations with large error have also been examined. The good comparison of IHO with duplicate GLOSS stations shows that, as far as can be determined, IHO data are reliable and ought to be used in improving coastal tide models. The non-Gaussian character of the errors suggests that further improvements in tidal modeling will require advances in data assimilation which are robust to non-Gaussian data error.