基于交叉梯度耦合的大地电磁与地震走时资料三维联合反演

为了有效解决目前大地电磁和地震走时资料单方法反演结果一致性不好的问题,同时克服基于岩石不同物性参数间关系耦合约束联合反演的局限性,本文研究了基于交叉梯度耦合约束的大地电磁与地震走时资料的三维联合反演算法.以较为成熟的天然地震走时资料三维正反演和大地电磁三维正反演算法为基础,实现了具有共同的反演网格,以交叉梯度结构耦合约束,并能同时获得电阻率和速度模型的三维联合反演算法.分别利用单棱柱体模型和双棱柱体模型合成数据进行了联合反演试算.结果表明:无论是单棱柱体模型还是双棱柱体模型,联合反演结果比单独反演对异常体的空间形态都有更好的恢复,其中单棱柱体模型反演的异常体电阻率更接近于真实电阻率,双棱柱体模型的联合反演结果不仅消除了围岩的部分电阻率假异常,而且增强了对异常体深部速度结构特征的恢复程度.联合反演还能同时改善电阻率和速度反向变化异常体的单独反演结果,进一步证明交叉梯度耦合不依赖于岩石物性关系,而强调地下结构的相似性,具有更普遍的适用性.

Three-dimensional joint inversion of magnetotelluric and seismic travel time data with cross-gradient constraints

To effectively overcome the weak consistency of single inversion of magnetotelluric (or teleseismic travel time data), and also to solve the limitations of petrophysical approach, we study a three-dimensional joint inversion algorithm of magnetotelluric and seismic travel time data with cross-gradient constraints. Based on well-developed three-dimensional teleseismic and magnetotelluric forward and inversion algorithms, a three-dimensional joint inversion algorithm having common inversion grids, with cross-gradient as structural constraint, and with both resistivity and velocity models able to be simultaneously obtained has been achieved. With synthetic data from a single-prism model and two double-prism models, trial computation of joint inversion has been carried out, and the results indicate that, compared with single inversion, joint inversion has better effect in recovering both single prism's and double-prism's spatial morphology. The resistivity value of the single-prism is much closer to the true model in joint inversion, while by double-prism models, joint inversion not only removes false resistivity anomaly of wall rock to a certain degree, but also has large improvement on increasing the resolution of P-wave deep structure of anomalous body. Moreover, joint inversion improve the separate inversion results when resistivity and velocity is incompatible, indicating the cross-gradient is not be restricted on the relationship of different petrophysical parameters, but rather focuses on structural similarity and thus has wide applicability.