3-D Inversion of the Data-Normalized Time-Lapse Controlled-Source Audio-Frequency Magnetotelluric Method

Geophysical methods can be used to monitor changes in physical parameters in the subsurface. By collecting relevant data at different moments at the same location, changes can be visually assessed by identifying and calculating differences between inversion images. Addressing the practical requirements of the controlled-source audio-frequency magnetotellurics (CSAMT) in monitoring, it is suggested that time-lapse data can be inverted using a data-normalization method. This approach aims to ensure the efficacy of geophysical monitoring while mitigating the poor comparability of inversion results caused by separately inverting time-lapse data and eliminating the assimilation effects of time-lapse inversion on results at different moments. The process is as follows. First, data at the initial moment are used to normalize those at the subsequent moments. Then, the dynamic regularization factor is used to constrain the model at different moments, and data at any moment are inverted simultaneously. To improve the inversion efficiency, an MPI parallel algorithm is introduced to achieve dual parallelism by allocating computing tasks based on acquisition time and frequency. The model with an invariant observation system and the dynamically changing underground structure over time is adopted in the synthetic data case. The inversion results show that data-normalized time-lapse inversion can reduce the influence of noise, help better focus the location of the anomalies, achieve higher resolution for small structural changes, and significantly improve the inversion efficiency. The research results in this article lay a good foundation for the practical application of data-normalized time-lapse CSAMT.