The traditional interpretation methods of nuclear magnetic resonance (NMR) usually use ID inversion based on homogeneous half space or layered conductive earth model and the information of water amount in each layer will be derived. However, these methods usually ignore the influence of the inhomogeneous underground conductivity to the interpretation results. Meanwhile, these results are easily restricted to local water-bearing structures. For the purpose of avoiding the problems aforementioned, a novel joint interpretation method combining transient electromagnetic (TEM) and NMR data is proposed in this paper, which provides an effective and practical interpretation method for the detection of complex hydrogeological conditions. This paper proposes a novel joint interpretation method combining transient electromagnetic (TEM) and NMR data based on the forward modeling theory of NMR method in 3D conductive media. Firstly, a fast resistivity image is derived from floating plate interpretation results of TEM data. Afterwards, the resistivity imaging data will be used as the electrical model in 3D NMR inversion. The distribution of transmitting magnetic field is calculated by FE method directly. The transmitting current source is loaded by introducing a pseudo-delta source, and the influence of "weak solution" in three dimensional simulation of magnetic field is eliminated by a force introduction of divergence condition. In order to improve the illness of sensitivity matrix and eliminate the noise, a non-linear fitting objective function considering the penalty terms is proposed and the linearized inversion method is used in the inversion process. The results of homogenous half-space model show that solutions of excitation magnetic field are almost the same as FEM solutions and digital filtering solutions and the influence of "weak solution" is eliminated. Results of the distribution of the NMR kernel function in different conductive models show that: (1) The background conductive of a model affects the distribution of the kernel function, especially when it is a low conductive one, and the lower of the conductive, the smaller range of the influence of excitation at the same pulse moment will be; (2) Local low resistive bodies with reasonably scale also have a significant effect on the distribution of the kernel function, lower resistivity and shallower buried depth causing greater influence, vice versa; Several conclusions can be drawn from the inversion results: (1) smaller variation of the geoelectrical model from the real case will lead to more accurate inversion results; (2) The inversion result can deviate from reality when using a homogenous half-space as the electrical model directly while the low conductive body have a certain scale, and the accuracy of the inversion result will see a significant improvement with the introduction of the floating plate interpretation method based on TEM data; (3) Though the accuracy of the floating plate interpretation decreases with increased depth, the impact of low conductive bodies on the distribution of the kernel function also decreases with that. Therefore, the fast resistivity image result based on TEM data, in spite of being an approximate result, can still offer a reasonable electrical model for NMR inversion. This paper proposes a novel joint interpretation method combining transient electromagnetic (TEM) and NMR data based on the forward modeling theory of NMR method in 3D conductive media. Results of the 3D distribution of kernel function show that low resistive bodies have a significant effect on the distribution of the kernel function, and the media conductivity should be considered during the NMR inversion. The floating plate interpretation based on TEM data can provide a reasonable geo-electrical model, and the accuracy of the NMR inversion will be improved significantly after introducing joint interpretation method combining transient electromagnetic (TEM) and NMR data.
