雷达高度计海况偏差估计神经网络模型研究

本文基于Jason-2高度计数据,在12个不同季节的cycle数据中组合1~6个cycle的有效波高、风速和海况偏差为训练集,选取Jason-2的另外3个不同季节的cycle数据集为测试集。经检验分析,确定3个cycle对应的BP神经网络模型。将该模型应用于HY-2高度计海况偏差的估计,通过海况偏差与有效波高及风速的拟合优度、解释方差和残差对比分析,结果表明:神经网络BP模型可以有效应用于HY-2的海况偏差估计并明显优于传统海况偏差参数模型。

Study on neural network model of estimating the sea state bias for radar altimeters

In this paper, which is based on the Jason-2 altimeter data, with the data of the significant wave height (SWH), wind speed (U) and sea state bias (SSB) combination of 1-6 cycle in 12 different seasons in the cycle data as the training set, select the other 3 cycles of Jason-2 data as the test set. By the test analysis, the BP neural network model which corresponds to 3 cycles for estimating the SSB is established. The model is applied to the estimations of SSB in the HY-2 altimeter, and the performances of the model can be evaluated by the goodness of fit between U and SWH by SSB, explained variance and residual contrast analysis. It suggests that the BP neutral network model can be effectively applied to the HY-2 estimations of SSB and significantly better than the traditional parameter model of sea state bias.