基于支持向量机由海洋地震勘探数据对海洋水体温度的预测  

Prediction of Ocean Water Temperature from Marine Seismic Data Based on Support Mector Machine

作　　者：彭阳阳 张进[1,2,3] Peng Yangyang;Zhang Jin(College of Marine Geosciences,Ocean University of China,Qingdao 266100,China;Function Laboratory of Marine Geo-Resource Evaluation and Exploration Technology,Qingdao 266237,China;The Key Laboratory of Submarine Geosciences and Prospecting Techniques,Ministry of Education,Ocean University of China,Qingdao 266100,China)

机构地区：[1]中国海洋大学海洋地球科学学院,山东青岛266100 [2]海洋矿产资源评价与探测技术功能实验室,山东青岛266237 [3]中国海洋大学海底科学与探测技术教育部重点实验室,山东青岛266100

出　　处：《中国海洋大学学报（自然科学版）》2022年第9期103-109,共7页Periodical of Ocean University of China

基　　金：国家自然科学基金项目(91958206)资助。

摘　　要：当前海洋水体的温度预测多采用CTD、XBT等仪器单点测定,所测得的海洋水体温度虽然垂向分辨率高,但横向连续性差。本文提出了一种利用支持向量回归机结合地震属性预测海水水体温度参数的方法,该方法在少量CTD数据约束下能够反演出纵横向高分辨率的水体温度结构。本文提取海水反射地震数据的13种属性值进行主成分分析,然后利用支持向量回归机(ε-SVR)结合CTD数据进行模型训练并对水体温度剖面进行预测,所预测的渤海夏季海水温度剖面特征与文献记载一致。At present, the temperature prediction of ocean water is mostly determined by CTD, XBT at a single point. The measured ocean water temperature has a high vertical resolution, but has a low horizontal resolution. In this paper, we propose a method to predict seawater temperature parameters by using support vector regression machine combined with seismic attributes. This method can invert the high-resolution water temperature structure in vertical and horizontal directions under the constraints of a small amount of CTD data. This paper extracts 13 attribute of seismic reflection data of seawater for principal component analysis, and then uses support vector regression(ε-SVR) combined with CTD data to conduct model training and predict the water temperature profile. The predicted characteristics of the summer sea temperature profile of Bohai Sea are consistent with those recorded in the literature.

关 键 词：地震属性 支持向量机 主成分分析 水体温度预测 

分 类 号：P714.8[天文地球—海洋科学]