There has been an increasing demand in recent years for detailed and accurate landslide maps and inventories in disaster-prone areas of subtropical and temperate zones, particularly in Asia. Most standard mapping methods require detailed fieldwork to be conducted by knowledgeable, skilled professionals. When predicting landslides, it is important to understand past landslide cases and prepare for situations in which the same phenomena occur. Developing automatic analysis methods using deep learning can contribute to the sophistication and cost of screening. This case study analyzed the potential of using deep learning convolutional neural networks（CNNs）for landslide detection with digital elevation models（DEMs）before the slide for deep-seated landslides（DLs）that occurred during Typhoon Talas. Here, we created 36,985 pieces of learning data from topographic information, which were then applied to a CNN using a multi-modal learning model. Eight types of influence factor images were created using the DEM as the learning data. The learning outcome achieved an accuracy of ＞0.856 for a 50 × 50-pixel window size CNN. This indicates that the decrease in the number of influence factor image types influenced the outcome. This study uses data from a limited range of DL sites in a topography specific to the accretionary zone. Although this CNN model is still in the initial stages of development, it accumulated many collapse cases and could contribute to disaster location screening, risk assessment, and hazard mapping during disasters.

In tunneling projects, various geological conditions, such as different rock types in the tunnel, are encountered on a daily basis. Visual observation is usually conducted by site engineers, but the precision of rock type determination is highly dependent on the experience of the site engineers. Therefore, we developed a rock type judgement system that determines the rock type from photographs of the rock using artificial intelligence （AI）. Visible light images taken from various angles for 29 types of rock obtained from 2,656 specimens were prepared for training and 7,000 images were randomly extracted for each rock type for learning by AI. The AlexNet was used to construct the AI algorithm. The percentage of correct answers in the learning model was 72.1％ on average. Using the learning model, we established a rock type judgement system that responds the rock type of captured images by sending them to the online server. To improve the accuracy of the system, we added a filter function to automatically narrow the list of rock types.

Recently, many IoT sensors are available at low cost. It is possible to acquire various data by combining appropriate sensors according to the items we want to measure and using control devices equipped data recording device such as Arduino and Raspberry Pi. On the other hand, there are high-precision measuring instruments for indoor experiments in the laboratory. By combining these with IoT technology, it is possible to improve the accuracy of field observation. Here, we introduce examples of field observation at a historical site that have continued since 2018 by performing such a combination. The historical site located in Saitama Prefecture are cave tombs dug in soft tuffaceous rock. On the surface of the slope facing southwest, many open fractures are formed by the intrusion of tree roots, then creates a rock block separated by fractures. There is a risk of rockfall. Therefore, the movement of the rock block is decided to monitor by real-time monitoring of the change in the fracture aperture. The purpose of the observation is to confirm whether the dangerous situation is getting worse. In this report, we introduce the know-how on the measurement technique leading to data acquisition and the knowledge obtained from the acquired data by real-time monitoring.