Introduction
Unmanned, semi-autonomous robots represent the present and future of… everything? Home delivery, search and rescue, infrastructure monitoring, and data collection for scientific inquiry are facilitated by autonomous robotic devices including UAVs, unmanned aerial vehicles. As prolific collectors and users of near-surface geoscience data, Bureau researchers have conspicuous need for near-surface platforms from which to acquire said data, and UAVs provide that platform from heights of 10 to 120 meters above ground level (AGL). From the scientist’s perspective, UAVs are merely sensor platforms, and their utility is confined to the sensors they’re capable of sustaining in flight. Advances in both UAV and sensor technologies, however, ensure expanded access for scientists to ever-more powerful and affordable data collection tools. Bureau scientists currently employ UAV technology for the following applications:
General Mapping
The hundreds of RGB images routinely captured in a single drone flight can easily be assembled to make suitable orthomosaic base maps for a variety of geoscience studies. Taking advantage of unprecedented resolutions—from centimeter- to millimeter-levels of precision, depending on project demands—geomorphic, vegetative, mineralogic, and hydrologic landscape features once indiscernible in aerial images are now routinely imaged and mapped.
Topographic Mapping
Digital photogrammetry—especially structure from motion (SfM) techniques—enable geoscientists to generate accurate, precise terrain models from series of overlapping images. The camera’s azimuth and position are recorded for each image during flight. Back in the lab, software is employed to merge images using this location and orientation information and pattern-matching algorithms to identify common features imaged in overlapping photos. Necessarily, every feature in the target area—small shrubs, abandoned bottles, sidewalk corners, highway centerlines, etc—must be captured in multiple images, enabling the software to determine each feature’s distance from the multiple camera locations. From this information—camera location, camera orientation, and feature distance from cameras—a digital terrain model is generated. If ground control points (GCPs) are collected, they can be included in the processing steps to improve spatial accuracy. High resolution terrain models are essential datasets for a wide variety of earth science inquiry including beach morphology studies, fault monitoring efforts, subsidence studies, and sinkhole monitoring, among many other uses and applications.
Field Measurements
It is not uncommon for researchers to monitor the efficacy of classification algorithms—applied to vegetation, soils, lithology, minerology, etc—with site visits to ensure field measurements align with predicted or modelled results. Images captured via drone at intermediate scales—between regional-scale airborne imagery and in-situ field measurements, for example—can provide the requisite precision to visually identify feature details against which the regional-scale classification efforts can be monitored, yet provide orders of magnitude more samples compared to in-situ field measurements. In our example below, approximately 40 shrubs were classified and mapped (red dots) over two hours by researchers in the field. A twenty minute drone flight, for comparison, can image thousands of shrubs with a precision enabling near-100% accuracy in visual species identification back in the office on computer (image below). So, compared to traditional field work, data capture via drone can provide additional or alternative methods for tweaking classification algorithm parameters and improve results.
Outcrop Capture and Modelling
Detailed digital outcrop models from SfM-derived surfaces enable precise and accurate geologic mapping in three dimensions. Before digital photogrammetry, conventional photography was used to take variable-scale imperfect photos of outcrops, and geologists used these photos to interpret and map faults, formations, facies, and other features of interest to geoscientists. But the mapping and interpretation--as with the underlying 'basemap' photos--were not in real-world space, scales were approximate and variable across the photos, and the location and orientation of features within the photos were likewise approximations. But using the same data capture and processing techniques described above—albeit with images captured from oblique rather than map-view orientations—detailed vertical surface structures and precise image mosaics (<5cm precision) provide geologists unparalleled data for 3D geologic mapping. These models and interpretation can be correlated to subsurface data including geophysical well logs, drillers reports, and seismic data.
Additional Sensors
Bureau researchers deployed a FLIR (forward-looking infrared)-equipped UAV recently to monitor spring flow in the Devils River watershed. Temperature gradients and spring inflow are evident in the image, below left. The image to the right is a standard RGB orthomosaic of the same area. Lidar, hyperspectral, magnetometer, and other geophysical sensors are on our deployment horizon as sensor size, weight, and affordability continue to evolve.
References
Abolt, C., Caldwell, T., Wolaver, B., and Pai, H., 2008, Unmanned aerial vehicle-based monitoring of groundwater inputs to surface waters using an economical thermal infrared camera: Opt. Eng. 57(5), 053113 (2018), https://doi.org/10.1117/1.OE.57.5.053113
Paine, J. G., Andrews, J. R., Morris, J. N., Saylam, K., and Kyle, J. R., 2023, Airborne and ground surveys of the April 2023 Daisetta sinkhole, Liberty County, Texas: The University of Texas at Austin, Bureau of Economic Geology, Open-File Report 23–1, 16 p, https://doi.org/10.23867/OFR23-1
Young, M.H., Andrews, J.R., Caldwell, T.G., Saylam, K., 2017, Airborne lidar and aerial imagery to assess potential burrow locations for the Desert Tortoise (Gopherus agassizii). Remote Sens., 9, 458. https://doi.org/10.3390/rs9050458