摘要
结合东北太平洋浮标资料,使用神经网络模型对WAVEWATCHⅢ海浪模式模拟的有效波高进行训练模拟,并与增加风场作为输入项的神经网络模型作了对比分析。通过分析浮标观测资料、WAVEWATCHⅢ数值模式和神经网络模拟的海浪有效波高大小,可以看出使用神经网络结合数值模式能够较好地提高有效波高的模拟精度。
Based on the NDBC buoy data in the North Pacific, artificial neural network models, with and without the wind fields as the inputs, were introduced to simulate the significant wave height (SWH) of the output of the third-generation ocean wave model. To improve the simulation precision of the SWH at a high value, a new type of neural network model was developed, which was trained at different part of swatch and was simulated synthetically. Compared with the buoy SWH, the root mean square errors (RMSE) of the ocean wave model, the neural network models without and with the QSCAT/NCEP wind fields as the inputs, and the new type of neural network model are 0.37m, 0.30m, 0.28m and 0.27m, respectively. Using the neural network models, the WAVEWATCH Ⅲ ocean wave model could simulate the SWH more accurately.
出处
《海洋预报》
2010年第2期8-14,共7页
Marine Forecasts