Photogrammetry of Fissure 8, Mount Kilauea, Hawai’i

“Estimating Volume Changes in Volcanic Features with UAV Photogrammetry”

A three-dimensional model of new lava extruded within Fissure 8 between Aug 27th and Sept 2nd, calculated and visualized with Cut Fill in ArcScene.

Duration: Oct 2018 - Dec 2018
Affiliation: Yale School of the Environment
What: Course project
My Role: Sole analyst and author

The Task

After decades of dormancy, Mount Kilauea began actively erupting on May 3, 2018. Fissure #8, emerging within a populous subdivision, was the most active and most closely monitored fissure during this eruption.

For my “Photogrammetry with Drones” course project, I worked with then-unreleased UAV imagery of the fissure (thanks to Angie Diefenbach, USGS / VDAP) to try and reproduce findings by USGS analysts.

The Objectives:

Create a 3D mesh and digital elevation model (DEM) of Fissure 8’s spatter cone and lava channel based on two UAV photo surveys, taken 7 days apart (late August & early September)
Georeference one of the 3D models, which lacked GPS coordinates
Calculate the difference in fissure volume between the two dates
Compare the height of the UAV-derived fissure models to a LIDAR-derived surface surveyed earlier that summer
Visualize cross-sections of the fissure structure and compare across dates

My Approach:

First, I generated a point cloud of the feature using AgiSoft Photoscan, which required extensive manual editing to remove erroneous points caused by steam. After creating a 3D mesh of the fissure, I then used the UAV imagery and DEM of the August model to georeference the September model within AgiSoft.

Photoscan’s volume calculation tool did not suit my purposes, since I did not have a reference plane against which to compare my spatter cone model. Instead, I exported my DEMs to ArcScene 10.7 to calculate the volume difference between the August and September models.

Finally, I plotted various cross-sections of the fissure using the Stack Profile 3D Analyst tool in ArcMap, to compare the area’s terrain in July, August, and September simultaneously.

Two of ten elevation cross section plots of Fissure 8. While LIDAR (yellow) of the area differed from our DEMs, a clear filling of the spatter cone is noticeable between August (purple) and September (blue).

Outcome highlights

A lava lake within the Fissure 8 cone appeared sometime between Aug 27 and Sept 2, containing an estimated 13,066 m³ of lava (+/- 28 m³). Assuming that lava infill began on Sept 1–as noted in field reports–the estimated rate of lava fill is 363 m³ per hour.
The area of the fissure crater floor in the August 3D model matches USGS field observations almost exactly (65 x 15 m). Meaning: this photogrammetric model is a good representation of the real thing, and provides a functionally useful model for earth scientists.
Large elevation discrepancies between the July LIDAR surface and the 3D models made the two impossible to compare quantitatively. However, the offset that I found matched the offset found by the Hawai’i Volcano Observatory (A. Diefenbach, personal communication).
My cross-sections identified two zones of significant terrain change between the LIDAR and UAV DEMs; I attributed these changes to lava channel shallowing caused by slow-moving lava hardening into rock.

Data source:
Patrick, M.R., Dietterich, H., Lyons, J., Diefenbach, A., Parcheta, C., Anderson, K., Namiki, A., Sumita, I., Shiro, B., and Kauahikaua, J., 2019, Cyclic lava effusion during the 2018 eruption of Kīlauea Volcano: data release: U.S. Geological Survey data release, https://doi.org/10.5066/P9PJZ17R.

« Back to Projects