Author(s): S B Charhate; M C Deo
Linked Author(s): M.C. Deo
Keywords: Wave heights; Spatial mapping; Artificial neural network; Genetic programming
Abstract: The measurement of significant wave heights through wave rider buoys is costly and as a result it becomes difficult to collect such data for a very long period of time and also over all stations of interest. Many times therefore the buoys are deployed for a reasonable period of time, say a few years, and at a few locations of interest and thereafter removed to use them somewhere else. For more important locations however the data collection may continue for a long time. A need therefore exists to come up with methods to generate wave data at a given location where they are unavailable from the information at stations where they are measured for relatively long durations. The current study deals with this issue and involves retrieving values of monthly maximum significant wave heights through spatial correlations. The monthly waves can provide useful input in applications such as planning of shipping routes and derivation of long term design waves. The study has specialty that it is based on the data driven methods of artificial neural network (ANN) and genetic programming (GP), which are purely non-linear soft computing approaches. The regions where data belong pertains to Gulf of Maine and the Gulf of Mexico in which the National Data Buoy Center (NDBC) had deployed wave rider buoys. Records of hourly significant wave heights over a period of 10 to 27 years at select locations were considered. From the entire data base the monthly maximum significant wave heights were extracted. The ANN and GP models were built to obtain wave heights at a target location on the basis of wave observations at nearby source locations. The model performance was studied with respect to different error measures. Both GP and ANN worked satisfactorily, with GP doing marginally better.
Year: 2010