Determining the three-dimensional structure of a volcanic plume using Unoccupied Aerial System (UAS) imagery

Aboud Albadra,Kieran Wood,Lucy Berthoud,Andrew Calway,Matthew Watson,Helen Thomas,Thomas Richardson,Emma Liu,Gustavo Chigna

doi:10.1016/j.jvolgeores.2019.106731

Abstract

This study presents a photogrammetric method for 3D reconstruction of a volcanic plume outline to retrieve its spatial properties. A dataset of sequential multi-view images was collected, using a drone-mounted camera, for a small-scale volcanic plume emitted from Volcán Pacaya, Guatemala. A ‘Space Carving’ algorithm has been applied to estimate the plume shape, top height, volume and drift direction. The complete method workflow is presented herein, including data capture, camera projection, image segmentation, and model reconstruction. The process applied is considered the simplest approach to reconstruct a 3D plume model from sequential imagery, whilst accounting for scene evolution within a probabilistic framework. The algorithm is sensitive to the method of image segmentation, scene resolution and number of images used, with unquantifiable uncertainty in the estimated plume volume due to the lack of ground-truth data. This proof-of-concept investigation confirms that 3D quantification of volcanic plume geometry can be achieved using UAS-based photogrammetry and shows promising results for a new method of measuring volcanic source parameters to validate and adjust dispersion models of volcanic plumes.

Highlights

IntroductionDispersion models can provide 3D forecasts of volcanic cloud trajectories (Stohl et al, 2011) that are ‘fully volumetric’ (i.e. the inner regions of plumes are modelled) and up to hemispheric in scale
Volcanic emissions are mainly monitored using a combination of satellite-based observations (Thomas and Watson, 2010) and dispersion modelling (Peterson et al, 2015)
A 3D volume estimation algorithm has been developed based on an image dataset capturing a volcanic gas plume emitted from Pacaya volcano, Guatemala

Summary

Introduction

Dispersion models can provide 3D forecasts of volcanic cloud trajectories (Stohl et al, 2011) that are ‘fully volumetric’ (i.e. the inner regions of plumes are modelled) and up to hemispheric in scale. These model algorithms are initialized based on plume source parameters (Mastin et al, 2009). Poor constraints on these initial parameters can result in large uncertainties in the spatial distribution of volcanic plumes and their physical properties (Devenish et al, 2012) (Mastin et al, 2009) (Beckett et al, 2014). There is a need, to collect plume source parameters (which may be at the meter scale) in order to initialize the plume dispersion models (which may be km in scale)

Objectives

Results

Conclusion