Automatic photomonitoring analysis for spatiotemporal evaluation of rockfall failure hazard

The assessment of rockfall failure hazard in both spatial and temporal components is a critical task for risk assessment since it must rely on a complete and detailed inventory. Recently, rockfall hazard evaluation has often been carried out leveraging Terrestrial Laser Scanning (TLS) and drone-based photogrammetry for accurate rockfall detection. However, these techniques are time-consuming and costly, thus affecting the possibility of obtaining high temporal-resolution data. In this framework, Photomonitoring is emerging as a new technology that uses digital images captured by optical sensors to automatically measure changes and displacements, reducing human effort and economic expenses while increasing the data acquisition frequency. In this study, we installed a remote 4G camera at the Poggio Baldi natural laboratory, monitoring the rock slope by regularly capturing photographs for 396 consecutive days. The automatic image processing pipeline, integrated with a 3D model of the slope, resulted in a multi-temporal rockfall inventory enriched by several attributes, such as the date and location of occurrence, magnitude estimation, and involved lithology. By exploiting these features, the spatiotemporal distribution of rockfalls was investigated using descriptive statistics, geospatial analysis, failure probability, and cumulative frequency-magnitude relationship. Accordingly, the analysis of the slope activity rate combined with cumulative frequency-magnitude relationships outlined the local influence of lithology and pelitic-arenaceous ratio in rockfall failure hazard.