Geology topics

George Hilley

Publications and source records attributed to George Hilley.

4 recordsLinked to original sources

USGS2026

SlideDetect: Spatio-temporal landslide detection using a three-dimensional convolutional neural network

Landslides pose a serious and ongoing threat to both human lives and infrastructure worldwide; therefore, it is of interest to predict where and when landslides are likely to occur. Advances in machine learning techniques have spurred numerous studies aimed at estimating relative landslide propensity, but are limited to spatial (as opposed to temporal) prediction due to the sparsity of landslide timing data. We address this data gap by training SlideDetect, a 3-dimensional convolutional neural network (3D CNN), to identify landslides based on their spatial and temporal occurrence within multitemporal image stacks. We use an inventory of landsides triggered by the 2018 Hokkaido earthquake and two years of monthly composite optical imagery spanning this event. The model can identify not only landslide location but also landslide date with an area under the precision-recall curve (PR-AUC) of 0.84. We further present a new standard for presenting PR curve results that explicitly compares model performance at different confidence thresholds, allowing for clearer model evaluation and comparison. Our new approach to constraining landslide timing paired with this more consistent and objective method for evaluating model performance shows considerable promise, and with further application and testing, SlideDetect could enhance the data availability and tools needed to advance landslide hazard and risk assessments.

JGR Machine Learning and Computation↗

USGS2020

High-resolution imaging of hydrothermal heat flux using optical and thermal Structure-from-Motion photogrammetry

Quantifying hydrothermal heat flux at meter-scale resolution over N0.25 km2 is required to bridge in-situ heat flux and satellite-based measurements. We advance a methodology that blends ground-based daytime optical and nighttime thermal infrared (TIR) imagery using Structure-from-Motion photogrammetry to map radiant hydrothermal heat flux over these scales at sites with low signal-to-noise ratios that would otherwise be difficult to characterize using, for example, unmanned aerial systems. The improved method uses a computerized telescopic mount to relocate and align daytime optical acquisitions with nighttime TIR imagery, thereby enabling TIR acquisition from multiple camera orientations positioned throughout a study area. This facilitates mapping of thermal features at sites of varying size and complexity and helps to ameliorate topographic occlusion effects and geometric distortions that can bias radiant hydrothermal heat flux estimates derived from the resulting orthorectified thermal maps. We assessed detection thresholds of this method at three sites across central California, which range in size, topography, and heat flux conditions. We found that blending of optical and thermal acquisitions successfully detected anomalous heat flux, even in cases where temperatures were only slightly greater than background. This approach might be applied to a variety of volcanic and hydrothermal systems to quantify the spatial distribution of heat flux, and how this may relate to factors such as the distribution of ground fractures and lava flow rheology.

Journal of Volcanology and Geothermal Research↗

USGS2016

High spatio-temporal resolution observations of crater-lake temperatures at Kawah Ijen volcano, East Java, Indonesia

The crater lake of Kawah Ijen volcano, East Java, Indonesia, has displayed large and rapid changes in temperature at point locations during periods of unrest, but measurement techniques employed to-date have not resolved how the lake’s thermal regime has evolved over both space and time. We applied a novel approach for mapping and monitoring variations in crater-lake apparent surface (“skin”) temperatures at high spatial (~32 cm) and temporal (every two minutes) resolution at Kawah Ijen on 18 September 2014. We used a ground-based FLIR T650sc camera with digital and thermal infrared (TIR) sensors from the crater rim to collect (1) a set of visible imagery around the crater during the daytime and (2) a time series of co-located visible and TIR imagery at one location from pre-dawn to daytime. We processed daytime visible imagery with the Structure-from-Motion photogrammetric method to create a digital elevation model onto which the time series of TIR imagery was orthorectified and georeferenced. Lake apparent skin temperatures typically ranged from ~21 to 33oC. At two locations, apparent skin temperatures were ~ 4 and 7 oC less than in-situ lake temperature measurements at 1.5 and 5 m depth, respectively. These differences, as well as the large spatio-temporal variations observed in skin temperatures, were likely largely associated with atmospheric effects such as evaporative cooling of the lake surface and infrared absorption by water vapor and SO2. Calculations based on orthorectified TIR imagery thus yielded underestimates of volcanic heat fluxes into the lake, whereas volcanic heat fluxes estimated based on in-situ temperature measurements (68 to 111 MW) were likely more representative of Kawah Ijen in a quiescent state. The ground-based imaging technique should provide a valuable tool to continuously monitor crater-lake temperatures and contribute insight into the spatio-temporal evolution of these temperatures associated with volcanic activity.

Kawah Ijen volcano↗

USGS2015

Integrated thermal infrared imaging and Structure-from-Motion photogrametry to map apparent temperature and radiant hydrothermal heat flux at Mammoth Mountain, CA USA

This work presents a method to create high-resolution (cm-scale) orthorectified and georeferenced maps of apparent surface temperature and radiant hydrothermal heat flux and estimate the radiant hydrothermal heat emission rate from a study area. A ground-based thermal infrared (TIR) camera was used to collect (1) a set of overlapping and offset visible imagery around the study area during the daytime and (2) time series of co-located visible and TIR imagery at one or more sites within the study area from pre-dawn to daytime. Daytime visible imagery was processed using the Structure-from-Motion photogrammetric method to create a digital elevation model onto which pre-dawn TIR imagery was orthorectified and georeferenced. Three-dimensional maps of apparent surface temperature and radiant hydrothermal heat flux were then visualized and analyzed from various computer platforms (e.g., Google Earth, ArcGIS). We demonstrate this method at the Mammoth Mountain fumarole area on Mammoth Mountain, CA. Time-averaged apparent surface temperatures and radiant hydrothermal heat fluxes were observed up to 73.7 oC and 450 W m-2, respectively, while the estimated radiant hydrothermal heat emission rate from the area was 1.54 kW. Results should provide a basis for monitoring potential volcanic unrest and mitigating hydrothermal heat-related hazards on the volcano.

California↗