SUMMARY

4-D electrical resistivity tomography (ERT), an important geophysical method, is widely used to observe dynamic processes within static subsurface structures. However, because data acquisition and inversion consume large amounts of time, rapid changes that occur in the medium during a single acquisition cycle are difficult to detect in a timely manner via 4-D inversion. To address this issue, a scheme is proposed in this paper for restructuring continuously measured data sets and performing GPU-parallelized inversion. In this scheme, multiple reference time points are selected in an acquisition cycle, which allows all of the acquired data to be sequentially utilized in a 4-D inversion. In addition, the response of the 4-D inversion to changes in the medium has been enhanced by increasing the weight of new data being added dynamically to the inversion process. To improve the reliability of the inversion, our scheme uses actively varied time-regularization coefficients, which are adjusted according to the range of the changes in model resistivity; this range is predicted by taking the ratio between the independent inversion of the current data set and historical 4-D inversion model. Numerical simulations and experiments show that this new 4-D inversion method is able to locate and depict rapid changes in medium resistivity with a high level of accuracy.

中文翻译：

---

基于时间分割的快速二维电阻率数据反演方法

概要

4-D电阻层析成像（ERT）是一种重要的地球物理方法，被广泛用于观察静态地下结构内的动态过程。但是，由于数据采集和反演会消耗大量时间，因此很难通过4-D反演及时检测到在单个采集周期中介质中发生的快速变化。为了解决这个问题，本文提出了一种用于重构连续测量的数据集并执行GPU并行化反演的方案。在该方案中，在采集周期中选择了多个参考时间点，这使得所有采集的数据都可以在4-D反演中顺序使用。此外，通过增加动态添加到反转过程中的新数据的权重，可以增强4-D反转对介质变化的响应。为了提高反演的可靠性，我们的方案使用主动变化的时间正则化系数，该系数根据模型电阻率变化的范围进行调整；通过采用当前数据集的独立反演与历史4维反演模型之间的比率来预测该范围。数值模拟和实验表明，这种新的4-D反演方法能够以很高的精度定位和描述介质电阻率的快速变化。根据模型电阻率的变化范围进行调整；通过采用当前数据集的独立反演与历史4维反演模型之间的比率来预测该范围。数值模拟和实验表明，这种新的4-D反演方法能够以很高的精度定位和描述介质电阻率的快速变化。根据模型电阻率的变化范围进行调整；通过采用当前数据集的独立反演与历史4维反演模型之间的比率来预测该范围。数值模拟和实验表明，这种新的4-D反演方法能够以很高的精度定位和描述介质电阻率的快速变化。