Fully remote assessment of rockfall incidents based on crowdsourced imagery

Abstract

This paper presents a fully remote approach for the assessment of rockfall incidents that is based on leveraging data that become available online with the goal to develop three dimensional (3D) models, document in detail the rockfall trajectory immediately following the incident and conduct rockfall analyses fully remotely. Such an approach can reduce the effort necessary to collect data and learn from incidents. The approach is well suited following natural disasters, where a wealth of field performance data may become available online through social media platforms and local news media. The steps to implement this approach involve: datamining the internet for crowdsourced data and particularly Unmanned Aerial Vehicle (UAV) footage of the incident, reconstructing the site morphology in the three-dimensional space by applying the Structure-from-Motion method, extracting insights from the crowdsourced data and conducting three-dimensional rockfall trajectory back-analysis. We demonstrate the approach through two incidents that occurred in Greece, where different amounts of crowdsourced data became available. We evaluate the proposed approach, discuss its limitations and benefits, and provide insights based on these two incidents. This paper shows that in both cases, the proposed approach enabled the rapid extraction of critical, perishable insights such as block detachment positions, block size, and fragment distribution. Also, the proposed approach allowed for the collection of all the input necessary to conduct detailed three-dimensional trajectory analyses. This supports the creation of high-precision inventories of both past and future incidents. Implementing this approach can enhance risk assessment accuracy, and inform mitigation strategies. The proposed approach allows the evaluation of geohazards globally fully remotely and possibly without the need for on-site visits.