Research on conductive models of damaged soil that consider the effect of microcrack expansion (the degree of saturation and suction) is weak. By assuming an equivalent conductive path a unit series-parallel conductive model of damaged soil under environmental loads was proposed. This model shows the change in soil porosity and fractal dimension. To verify that, the soil was damaged by rainfall cycles (simulated natural drying and rainfall). Electrical measurements and X-ray microscopy tests were performed to obtain the damaged soil resistivity, porosity, and fractal dimension variation. The resistivity was calculated based on the conductive model, and the error was approximately 7.9% compared with that of the test. In addition, the soil damage variable related to soil porosity and fractal dimension was introduced, and it exhibited a logarithmic relationship with soil resistivity. Variations in soil damage during the rainfall cycles were observed. In the top layer, the soil porosity increased and the fractal dimension decreased owing to microcrack expansion, resulting in an increase in soil damage. In contrast, in the bottom layer, the soil porosity decreased and the fractal dimension increased, resulting in a decrease in soil damage due to particle migration from the top area and pore fill.

Earthquakes and rainfall both cause soil damage and strength degradation of cutting slopes, resulting in increased slope instability. However, few studies have been conducted on the failure mechanisms of cutting slopes under earthquakes and rainfall. In this study, field electrical measurements were conducted to evaluate the damage to a cutting slope hit by the Yangbi Earthquake (MS = 6.4) in Yunnan Province, China. After material segmentation using the resistivity probability density statistical method, we observed several damaged areas running along the slope depth direction, forming several potential sliding surfaces. Furthermore, considering the slope damage after the earthquake, a discrete element model of the slope was developed, and the dynamic process of the gravel-soil landslide under rainfall was analyzed. Compared with low cutting slope with thin overburden sliding along one sliding surface, the results indicate that the high cutting slope with thick overburden slides along several sliding surfaces that formed by the earthquake-step sliding mods. Slope sliding can be divided into four stages: First, the slope body at the bottom area slid and accelerated firstly, while several cracks appear on the top area due to tension (initial stage and acceleration stage). Thereafter, the upper slope body gradually slides along its respective sliding surface. The body at the bottom area of the slope was pushed by that at the upper area and slid at a high velocity along the sliding surfaces due to secondary acceleration (secondary acceleration stage). Finally, the sliding velocity of the slope gradually decreases, and an accumulation is formed, entering a stable stage (deceleration stage).

Electrical resistivity tests can potentially be applied in loess damage testing under combined freeze-thaw cycle (FTC) and earthquake conditions, which is crucial for preventing and controlling loess landslides. However, two challenges involving loess electrical resistivity measurements and damage characterization should be addressed. To achieve loess spatial resistivity measurements in extreme environments with low-uncertainty, a novel, multichannel, four-point method utilizing flexible electrodes is proposed. For loess damage characterization, a novel fusion algorithm is developed that integrates the electrical conductivity model with a data-driven process to eliminate the influence of moisture content and temperature on resistivity. To validate this approach, loess resistivity tests and damage characterizations were conducted using a combination of FTCs and earthquakes. The results indicate that the proposed method addresses the challenge of continuous measurement, ensuring that the discrepancy between the calculated and CT test results remains within an acceptable range, where the relative error ranged from 0 to 0.15. In addition, in the top and bottom areas, where considerable soil moisture exists, the calculation error associated with the previous empirical model was reduced considerably, with the relative error primarily ranging from 0.04 to 0.44.