| Summary: | 博士 === 國立臺灣大學 === 工程科學及海洋工程學研究所 === 106 === A wave ensemble forecast system is being developed based on the NOAA WAVEWATCH III (NWW3) two nesting multi-grid model over Taiwan area. The ensemble system consisted of 20 ensemble members and was set with spatial resolutions of 0.25 and 0.1. The wind forcing is coming from the WRF-based ensemble forecast system (WEPS) 10m wind fields of Central Weather Bureau (CWB) with spatial resolutions of 45km and 15km. The cycle initial condition of each wave ensemble member from the previous run of the same ensemble member is applied to generate a history perturbation of swell. The objectives of this work are to verify the impact of different wind forcing formulas, to find the better composition of ensemble members, and to evaluate the forecast capacity of resulting ensemble forecast system. We first proposed the combination of using two built-in wind forcing formulas to form twenty ensemble members (each for ten members), which can reserve the advantages of different formulas under various wind fields (monsoon and typhoon period), increase the average ensemble spread and decrease the difference between the root mean square error and average ensemble spread based on the truth value at open seas. With Reliability diagram, Brier Skill Score and Relative Operating Characteristic analyses of assessing the quality of forecasts, the ensemble system has good forecast capacity and discriminate between the events and non-events. It also has better forecast skill than the operational deterministic forecast, and can be comparable with NCEP global ensemble system. Consequently, the wave ensemble forecast system is approved to have the skill in terms of probability forecast at open sea and some coast areas around Taiwan.
Nevertheless the overestimation near some coast areas could be improved by increasing the grid resolution and resolving nearshore wave simulation to reduce RMSE. For the underestimation of SPRD we intend to add perturbation at low frequency swell as initial condition to increase SPRD in the near future.
Application of ensemble forecast includes the establistment of operational wave ensemble forecast system and probability forecast on decision making of marine installation. The operational wave ensemble forecast system performs 4 times daily and 72 hours forecast for each time. Products of ensemble forecast system involve point output and gridded output. The point output utilizes boxplot to show ensembles. The gridded output contains ensemble members, ensemble means and spread, 10% exceeding probabilities, probability and spaghetti diagrams at different thresholds every three hours. Ensemble forecast combined with monte-carlo method could provide the probability of operation under the thresholds of wave height, wind speed and duration of operation for decision making.
|
|---|
