| 沉积物粒度组分空间预测的神经网络残余kriging方法 |
| |
| 引用本文： | 刘付程,杨毅,张林,魏陶荣馨,王宇涵,夏量.沉积物粒度组分空间预测的神经网络残余kriging方法[J].海洋通报,2020,39(3):363-371. |
| |
| 作者姓名： | 刘付程 杨毅 张林 魏陶荣馨 王宇涵 夏量 |
| |
| 作者单位： | 江苏海洋大学测绘与海洋信息学院,江苏连云港222005;江苏海洋大学测绘与海洋信息学院,江苏连云港222005;江苏海洋大学测绘与海洋信息学院,江苏连云港222005;江苏海洋大学测绘与海洋信息学院,江苏连云港222005;江苏海洋大学测绘与海洋信息学院,江苏连云港222005;江苏海洋大学测绘与海洋信息学院,江苏连云港222005 |
| |
| 基金项目： | 国家自然科学基金 （41976187；41801316）；淮海工学院自然科学基金 （Z2014017）；江苏省大学生创新训练计划项目（SD201711641107004） |
| |
| 摘 要： | 针对近海表层沉积物粒度组分空间变异的尺度差异性,提出了基于广义回归神经网络残余kriging的沉积物粒度组分空间预测方法,并以海州湾沉积物粒度数据为基础,分析了其在沉积物粒度组分空间预测和底质类型制图中的应用效果。结果表明,广义回归神经网络残余kriging方法能够获得比普通kriging方法更高的沉积物粒度组分空间预测精度,并且其底质类型的总体空间预测精度达到85%以上,相应的Kappa系数也在0.8以上,显示底质制图的预测类型与样本的实测类型具有较高的一致性。新方法对于开展定量化的沉积物粒度组分空间预测和底质类型制图具有参考价值。
|
| 关 键 词： | 广义回归神经网络残余kriging 沉积物粒度组分 空间预测 底质制图 |
| 收稿时间： | 2019/9/15 0:00:00 |
| 修稿时间： | 2019/11/10 0:00:00 |
A generalized regression neural network residual kriging method for spatial prediction of sediment grain size compositions |
| |
| LIU Fucheng,YANG Yi,ZHANG Lin,WEI Taorongxin,WANG Yuhan,XIA Liang.A generalized regression neural network residual kriging method for spatial prediction of sediment grain size compositions[J].Marine Science Bulletin,2020,39(3):363-371. |
| |
| Authors: | LIU Fucheng YANG Yi ZHANG Lin WEI Taorongxin WANG Yuhan XIA Liang |
| |
| Institution: | School of Geomatics and Marine Information, Jiangsu Ocean University, Lianyungang 222005, China |
| |
| Abstract: | In view of spatial variability of grain size compositions in offshore surface sediments at different scales, a generalized regression neural network residual kriging (GRNNRK) method is proposed and its application effect is also analyzed and evaluated in spatial prediction of grain size compositions and sediment types based on the surface sediment grain size composition data of Haizhou Bay. The results show that GRNNRK can obtain a higher spatial prediction accuracy of sediment grain size components than the ordinary kriging method and its overall predictive mapping accuracy of sediment types reaches up to 85 % and the corresponding Kappa coefficient is also more than 0.8, indicating that the mapping types of sediments are highly consistent with their actual types. So the new method of GRNNRK has a certain degree of reference value for quantitative sediment mapping. |
| |
| Keywords: | generalized regression neural network residual kriging(GRNNRK) sediment grain size composition spatial prediction sediment type mapping |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
| 点击此处可从《海洋通报》浏览原始摘要信息 |
| 点击此处可从《海洋通报》下载免费的PDF全文 |
