交叉梯度理论及其在地球物理联合反演中的应用  

The theory of cross-gradient and its application in geophysical joint inversion

作　　者：王俊[1,2] 孟小红[1,2] 陈召曦[1,2] 李芳[1] 

机构地区：[1]中国地质大学(北京)地球物理与信息技术学院,北京100083 [2]地下信息探测技术与仪器教育部重点实验室(中国地质大学,北京),北京100083

出　　处：《地球物理学进展》2013年第4期2094-2103,共10页Progress in Geophysics

基　　金：国土资源部深部探测技术与实验研究专项(SinoProbe)(201011039);国家自然科学基金(41074095)联合资助

摘　　要：地球物理联合反演中,建立合适的目标函数将不同物性参数联合起来是非常重要的一步工作,然而并不是所有的物性参数之间都存在较为明显的显式关系.本文系统介绍并分析了地球物理联合反演中用于耦合不同物性参数的新方法—交叉梯度法.该方法充分利用交叉梯度函数的性质,从结构上对不同物性参数进行耦合.在联合反演算法的目标函数中引入交叉梯度项,无需不同物性间的其他耦合关系即可实现广义意义下的联合反演.理论密度数值模型验证了交叉梯度函数的相关性质,相关应用实例也表明了该方法在联合反演中应用效果良好,具有一定的实际应用价值.During the process of geophysical joint inversion, one of the key issues is how to select an appropriate analytic expression to link all the used geophysical parameters together in a property way. Only in this way, can we make full use of different parameters to define the objective geologic body more precisely including its shape, position and so on. However, it is not always an easy job to find the coupling relationship between the two or more used parameters. Generally speaking, the relationships among different parameters are not obvious. In this paper, we do some intensive researches on a new method named cross-gradient which can be used in geophysical joint inversion to link all the parameters. This technique makes the best use of the properties of cross gradient function to establish relationships among different parameters in structure. In the algorithm of the inversion, we set up our objection function based on cross-gradient function. This way we can finish the joint inversion without searching for other obvious links among the geophysical parameters. After the general introduction of the cross-gradient function, we form a density model to test the properties of this function mentioned above. Some application examples also indicate that this method has good performances in geophysical joint inversion and has great application value in daily work,

关 键 词：交叉梯度 耦合关系 离散化 地球物理 联合反演 

分 类 号：P631[天文地球—地质矿产勘探]