线性回归分析技术推估海底地形

Predicting submarine topography by linear regression analysis

为了填补船测海深数据空白,给出了海底地形起伏与重力异常和重力异常垂直梯度之间的导纳函数关系。据此,以测高重力异常、重力异常垂直梯度作为输入数据,采用线性回归分析技术,在西南太平洋相关海域开展了海底地形反演试验。结果表明,通过不同方法获取的比例因子与海底地形呈现一定的内在联系,地形平坦海域,比例因子较小;海山分布较多的地形起伏较大的海域,比例因子相对较大,反映了重力数据与海底地形较强的相关性。同时,采用线性回归方法构建的海底地形模型检核精度最高,相较于传统方法获取的海底地形模型,精度最高提升了46%左右,与ETOPO1海深模型和DTU10海深模型相比较,模型精度最大提高了近一倍有余。另外,不同方法对于不同的海底地形具有各自不同的优势,靠近海山区域,采用线性回归技术反演的海深结果优于传统方法;在海山部分,传统方法反演精度又好于线性回归技术。不同数据源反演海底地形的统计结果表明,以重力异常垂直梯度构建的海底地形模型的检核精度优于以重力异常作为输入数据构建的海底地形模型。

In order to fill the blank of ship-measured bathymetrical data, the admittance function between bathymetry and gravity anomalies/vertical gravity gradient anomalies was derived. Based on the gravity anomalies/vertical gravity gradient anomalies data, the linear regression analysis technique was used to carry out the bathymetry inversion test in the relevant sea area of the Southwest Pacific. The test results show that the scale factors obtained by different methods have certain internal relations with the submarine topography. The scale factor in flat terrains is smaller, and is larger in up and down terrain, which reflects the strong correlation between gravity data and sea depth. The accuracy of the sea depth model constructed by linear regression analysis technique is the highest, which is about 46% higher than that of the traditional method, and is more than doubled compared with those of the ETOPO1 model and the DTU10 model. In addition, different methods have different advantages for different submarine topography. Near the seamount area, the accuracy of the model constructed by linear regression analysis technique is better than that of the traditional method. On the seamount, the accuracy of the model constructed by traditional method is better than that by the linear regression analysis technique. The statistical results of bathymetry inversion from different sources show that the accuracy of the sea depth model constructed by vertical gravity gradient anomalies is superior to that of the model constructed by using gravity anomalies as input data.