Landslides are one of the major causes of casualties and property damage, and they occur mostly in mountainous areas worldwide, which are characterized by human habitation and engineering activity (Parise and Wasowski, 1999; Raetzo et al., 2002; Staub, 2001). Rainfall conditions often lead to rainfall-induced landslides (Dai & Lee, 2002). For example, rainfall-induced landslides occurred in Taipei, Taiwan Province of China, where the daily rainfall exceeded 370 mm on November 1, 2000 (Wen et al., 2018). The rainwater entered the fractured zone of the volcanic bedrock, where the rock flaked off, and caused the movement of an unstable rock mass constituting water-absorbing tuffaceous conglomerate and weathered volcanic rock. Another example of a rainfall-induced landslide is the deep-seated creeping landslide caused by continuous heavy rains on Mt. Wanitsuka, Kyushu, Japan, on September 7, 2005 (Wen et al., 2018). The landslide mass was composed of soil and broken gravel, detached from the basal sandstone strata. Another rainfall-induced landslide, named the Shibangou landslide, occurred in Sichuan Province, China, on September 18, 2011(Wen et al., 2018). After two days of continuous heavy rainfall, the strength of the soft rock strata overlying the interbedded sandstone and mudstone base rock reduced due to water immersion, initiating the landslide.
After studying the geological conditions of rainfall-induced landslides in Chongqing, China, Wen et al. (2018) determined that the probability of landslides induced on soft rock slopes reaches 93.7% and only 1.5% on hard rock slopes because of their high rock strength and weathering resistance. Although the landslide probability on a hard rock slope is low, over 31 landslides have been induced on the sandstone, limestone, and granite slopes (Wen et al., 2018). Thus, it is necessary to monitor and predict the initiation of a landslide before it causes substantial damage. Geophysical methods can provide a preliminary assessment of a landslide and are non-destructive to the landslide mass compared to traditional methods such as drilling, probing, and trenching. The main geophysical methods used to prospect for landslides include seismic methods (Bruno & Martillier, 2000; Feng et al., 2003; Jin & Du, 2004; Jongmans & Garambois, 2007; Lin et al., 2020; Malehmir et al., 2013), high-density resistivity methods (Aleksandar, et al. 2012; Epada et al., 2012; Guo et al., 2004; He et al., 2016; Jiang et al., 2008; Kong et al., 2008; Roux et al., 2019; Wang et al., 2019), ground-penetrating radar (Jiang et al., 2000; Zheng et al., 2006), and their integrated application (Liu, 2019).
In addition to the above-mentioned prospecting methods, we herein discuss another geophysical method, the controlled source audio-frequency magnetotellurics (CSAMT) method, which can be used to predict landslides induced on both soft and hard rock slopes. As a frequency-sounding method using an artificial source, the frequencies of CSAMT range from 0.25 to 8192 Hz: the lower the frequency, the greater the investigation depth. Until now, CSAMT has been applied effectively to the exploration of ore (Basokur et al., 1982), water resources (Wannamaker, 1997), and engineering geology (Di et al., 2002).
In our previous study, we examined potential landslides in the Jinsha River, southwest China, using the seismic reflection method (Wang et al., 2006). We visualized the geometric structure of the sliding surface. When the layer properties are absent, we complemented the seismic reflection method using the CSAMT method (Gong et al., 2005). Wen et al. (2018) pointed out that the sliding surface is always located at the bottom of a water-bearing soft rock layer or at the fracture of a hard rock, where the resistivity is lower than that in the surrounding rock after heavy precipitation. CSAMT is one of the most suitable methods for recognizing water-bearing strata, but it is rarely applied for landslide exploration. Therefore, in this study, we focus on the forward modeling of CSAMT to investigate the response of the sliding surface and provide a theoretical basis for the application of CSAMT in landslide exploration.
Although sliding surfaces normally have a tilted 2D structure with a certain width and length, the dependent surrounding terrain has a 3D shape. Consequently, 3D modeling should be applied to comprehensively study the response of sliding surfaces with overprinted landforms. The finite element method (FEM) is known to be more precise for the numerical modeling of complex structures and is used in this study. FEM has been previously used to model the elastic waves in natural earthquakes (Wang et al., 1999). Herein, we implemented FEM to model scatter wave in CSAMT with a ridge topography. In this paper, we first introduced the CSAMT modeling method with the FEM technique, then tested the coded program, provided the modeling results of a landslide model, and analyzed the electromagnetic field and apparent resistivity characteristics, followed by discussing the CSAMT relative survey parameters. We also assessed the prospecting ability of CSAMT, revealed the technical and physical restrictions of CSAMT for landslide exploration, and provided suggestions for future work.