Pyroclast Tracking Velocimetry: A particle tracking velocimetry‐based tool for the study of Strombolian explosive eruptions

Abstract

AbstractImage‐based techniques enable high‐resolution observation of the pyroclasts ejected during Strombolian explosions and drawing inferences on the dynamics of volcanic activity. However, data extraction from high‐resolution videos is time consuming and operator dependent, while automatic analysis is often challenging due to the highly variable quality of images collected in the field. Here we present a new set of algorithms to automatically analyze image sequences of explosive eruptions: the pyroclast tracking velocimetry (PyTV) toolbox. First, a significant preprocessing is used to remove the image background and to detect the pyroclasts. Then, pyroclast tracking is achieved with a new particle tracking velocimetry algorithm, featuring an original predictor of velocity based on the optical flow equation. Finally, postprocessing corrects the systematic errors of measurements. Four high‐speed videos of Strombolian explosions from Yasur and Stromboli volcanoes, representing various observation conditions, have been used to test the efficiency of the PyTV against manual analysis. In all cases, >106 pyroclasts have been successfully detected and tracked by PyTV, with a precision of 1 m/s for the velocity and 20% for the size of the pyroclast. On each video, more than 1000 tracks are several meters long, enabling us to study pyroclast properties and trajectories. Compared to manual tracking, 3 to 100 times more pyroclasts are analyzed. PyTV, by providing time‐constrained information, links physical properties and motion of individual pyroclasts. It is a powerful tool for the study of explosive volcanic activity, as well as an ideal complement for other geological and geophysical volcano observation systems.