LAVA BLOCK SIZE MEASUREMENTS USING 3D POINT CLOUDS DERIVED FROM 2D PHOTOS AND STRUCTURE-FROM-MOTION

The crust of lava flows is formed as the lava surface cools and solidifies, which may be folded or heavily fragmented during emplacement. Crust fragmentation is a balance between applied stresses and yield strength with initial block sizes determined by cooling, effusion rates and crustal thickness. Block sizes have been used as indicators of emplacement conditions; e.g. silicic lava block sizes often decrease with distance from the vent due to mechanical interactions and thermal cooling, but in basaltic lava blocks may increase in size due to agglutination or cooling-induced thickening of the crust.

Manual measurement of lava block sizes are tedious and time consuming and satellite and aerial remote sensing techniques (e.g. lidar) are typically not capable of determining block sizes less than a meter in diameter. Structure-from-Motion (SfM) allows block size data collection in the field using only a digital camera. SfM uses 2D photos to reconstruct a 3D point cloud of the area of interest. SfM points clouds, at close range (meters), produce point clouds of similar resolution to ground-based lidar scans, but with less equipment to haul around over rough terrain.

In this study, we have determined characteristic block sizes from several lava flows in the central Oregon Cascades. VisualSfM was used in initial photo processing to create dense 3D point clouds. CloudCompare was then used in processing the point clouds to remove extraneous points, rescale the data to known parameters (using eight inch cube targets in the field), and to mesh the data into gridded 3D surfaces.

The point clouds were segmented into individual blocks in Matlab using a watershed analysis. The watershed segmentation allowed for statistical analyses of block dimensions and orientations. The watershed segmentation is resolution dependent and prone to over- or under-segmentation and must be tuned to each specific dataset and analyzed for accuracy. However, preliminary results indicate the SfM-watershed analyses produce results comparable (mean block sizes typically within a few centimeters) to manual field measurements. SfM point clouds and watershed segmentation allows for unbiased, efficient, and more quantitative analyses than manual field measurements and can be used to analyze both effusive and explosively derived block sizes.