STATISTICAL FORECASTING MODEL OF SOLITON EVENTS IN THE SULU SEA

Sulu Sea is one of the areas heavily infested by soliton events, by which offshore and subsea exploration as well as operation can be heavily affected. In spite of numerous researches being done on this issue, its occurrence continues to impact the mentioned activities. As one of the potential oil and gas regions in the Southeast Asia, a statistical forecasting model for the soliton events is necessary to reduce downtime loses and onsite casualties. Tidal wave, responsible for the soliton events in this region, is highly seasonal, hence the exponential smoothing method is opted to generate the forecast model. Current speeds at 3 different water depths at Karupang have been measured and recorded at a constant interval of 10 minutes for approximately 109 days, from 1st September 2012 to 18th December 2012. The current speed forecast is first developed using the simple seasonal exponential smoothing method, as a parsimoniousy model is desired. Level and seasonality smoothing constant α and δ of 0.50 and 0.59 are chosen by numerical methodtrial and error . The seasonality of the model, 24 hours and 15 minutes, is obtained from the non-correlated power spectrum density plot. The forecasted values are compared with the actual observed current speed, and it is shown that the percentage of accuracy and the goodness of fit of the forecast model for 1 seasonal cycle of 3 different water depths are approximately 70%. Despite literaturepopular belief , no clear correlation can be observed between high speed current events and the occurrence of solitons in this study. Nonetheless, the accuracy and robustness of this model can potentially contribute to the oil and gas industry prediction of solitons.