The Challenges of Measuring Geyser Deformation with LiDAR

A 1.557Mb PDF of this article as it appeared in the magazine complete with images is available by clicking HERE

In September 2010 a multi-disciplinary team of geoscientists set up camp for four days at Yellowstone National Park’s Lone Star Geyser. The objective of the expedition was to use geophysical methods to characterize eruption cycles in the geyser system. The team deployed several instruments around the geyser cone, including infrared sensors, visible and infrared video, and microphones for capturing acoustic signals. Measurements of water velocity and channel cross-sectional area for the main streams flowing away from the geyser were used to estimate discharge and give an idea of the total water output from the system in each cycle.

A Leica ScanStation C10 was also set up on a hill above the geyser in order to map hydrothermal ground deformation. The geophysical instruments recorded distinct eruption cycles at Lone Star, but the LiDAR scans did not show clear cycles of ground inflation and deflation. The question arises: why was LiDAR unsuccessful here? In this article we will describe the scanning results and consider potential sources of error that may have obscured the deformation signal.

Lone Star Geyser
Geyser eruptions are driven by the systematic filling and emptying of an underground fluid storage system. Geothermally-heated water becomes superheated under pressure, eventually breaking through rock constrictions and flashing to steam, creating a jet of boiling water. The regularity of geyser eruptions is related to the geometry of the plumbing system and the time needed to fill and empty the system. These cycles result in ground inflation and deflation that can be sensed at the earth’s surface.

Lone Star is an isolated cone geyser that erupts regularly every three hours. Major eruptions are preceded by smaller water fountaining events starting about one hour before the main eruption. These pre-eruptions signify when the plumbing system has become partly or fully recharged. Following the 30 minute eruption period, there is a relaxation and recharge period of about 90 minutes (Karlstrom et al., in review). We would thus expect to see maximum inflation just before the eruption and maximum deflation about 30 minutes to an hour after the start of the eruption.

LiDAR scanning at Lone Star
The scanner was placed on a hill above the outwash plain, about 55 m from the geyser cone. The first three scans of the day were used to test the scan area and subsequent scans focused on just the region around the geyser (Figure 1). A complete scan was started every 20 minutes over the course of about 6 hours. The scanning continued through two major eruptions and likely captured two complete inflation-deflation cycles. The stream flow measurements estimate an average eruption volume of 21 4 m3 of water, which should correspond to a maximum of about 7 1 cm of ground deformation if we assume a simple cylinder model with a 10 m radius of influence around the geyser cone (Karlstrom et al., in review).

Measuring deformation
Deformation is assessed by comparing the vertical offsets between successive scans and between scan pairs before and after each eruption. Two major eruptions were captured during the scan period (Table 1). Scan 5 (SW5) was made just before the first eruption and Scan 7 (SW7) was made about 20 minutes after the end of the eruption. Scan 15 (SW15) was made just before the second eruption and scan 17 (SW17) was made about 20 minutes after the end of the eruption. Maximum offset should be visible between these scan pairs.

Offset was measured by cropping each point cloud to just the raised area around the geyser, located within a 10 m radius of the geyser cone, and not including the geyser cone itself. These points were used to create a TIN model of the surface in Leica Cyclone. The vertical distances between TINs for successive scans could then be measured at 5 cm intervals in a grid pattern to generate offset statistics.

Mean offset between TINs shows a slow deflation of about 8 mm occurring between SW5, just before the first eruption and SW11, about 2 hours after the eruption (Table 1). The second eruption shows a much stronger offset. Beginning at SW13, where minor eruptions indicate the system has reached saturation, there is inflation of about 11 mm up to just before the eruption. In the 40 minutes during and after the eruption there is a 26 mm drop in the surface. This is followed by 27 mm of inflation in the 40 minutes following the eruption, and SW19 returns to the same location as SW15. This pattern seems to match the expected 30-60 minute recharge period measured using the other geophysical methods.

Spatially, offset in the TINs seems focused on the geyser mound and decreases at the edge of the 10 m radius (Figure 2). The spatial distribution of offset between points in the whole scan area was also visualized using the open source CloudCompare package. Figure 3 shows areas of the scan where offset is minimal (< 5 mm), and reveals a strange cyclic pattern. If offset was affected by scanner angular accuracy, we would expect error to be smallest at the scanner "0" azimuth, in the center of the image. If offset was affected by wind, we would expect offset to increase from left to right across the image, as the scanner is increasingly shifted away from its set rotation track (Olsen, 2012). Neither of these patterns is evident here.

Correcting for potential scanner error
An attempt was made to correct for scanner error by realigning point clouds using a cloud-to-cloud approach. One potential problem with re-alignment is that the process may actually remove nearuniform deformation by trying to correct for it. To address this, the re-alignment was done using the whole scan area, rather than just the cropped region. The hope here is that most of the scan area used for the alignment is not deforming and will provide a good baseline.

Cloud-to-cloud alignments for SW5-SW7 and SW12-SW17 removed most noticeable offset. The mean vertical offset for SW5-SW7 is -5 mm for the unaligned scans and only -2 mm for the aligned scans. For the SW12-SW17 pair, mean offset for the unaligned scans is -16 mm and only +2 mm for the aligned scans. Offsets of 2 mm are both within scanner error, so there is no longer a significant offset after either eruption.

Statistics for the cloud-to-cloud alignment give some insight into how the scanner would need to be tilted to align the scan pairs. For the first eruption there is little significant rotation in either the X, Y, or Z directions. For the second eruption, the observed large offset is created by significant X and Z rotation. That is, the scanner would need to tilt forward and rotate from side to side. It is possible that a combination of horizontal wind motion and forward slumping of the scanner could have created this offset. Alternately, the cloud alignment may need to be done on a much larger scan area so that it can include points from features that remain stable throughout the eruption.

Signal or Error?
From this data set it is not immediately clear if the scanning captured ground deformation or not. Many of the offset measurements are below scanner error budget for the ScanStation C10, which has angular accuracy of 60 rad (about 3 mm at a distance of 50 m) and distance accuracy of about 4 mm. The second eruption shows a significant drop of 26 mm, followed by inflation back to 27 mm, which seems to match the geyser cycle and is reasonable, considering we expect up to 7 cm of ground deformation. However, in reality, the system is probably much broader than 10 m in radius, so the actual deformation is likely to be much less. Deformation may even be on the mm to sub-mm scale, in which case it would be indistinguishable from total scanner error.

It is strange, however, that offset should be lacking during the first eruption and so clear for the second eruption. Additionally, recharge seems to occur very slowly after the first eruption (2 hours) and very quickly after the second eruption (40 minutes). The geophysical methods indicate a standard deviation of 13 minutes for the length of the recharge period, which is too short to account for the difference between the first and second eruption. Estimates of ground deformation from stream discharge measurements range from about 5 cm to 9 cm for the nine eruption cycles where these measurements could be made (Karlstrom et al., in review). It is thus possible that the contrast in offset between eruptions is due to natural variations in eruption strength.

The cyclic distribution of error as shown in Figure 3 suggests that other forces, such as wind may have moved the scanner. However, wind picked up before the second eruption, and it is the first eruption that does not show a clear inflation-deflation signal. The regularity of the vertical offset for the second eruption also does not seem indicative of wind effects. For these reasons the scanner movement estimated in the cloud-to-cloud alignments seems a more likely explanation for the significant offset following the second eruption. This movement may be due to slumping of the scanner due to an unstable setup (e.g. the tripod legs were not securely fixed in the ground).

A significant unknown in this study is the spatial distribution of the underground fluid storage system and the nature of ground deformation on short timescales. Over what radius surrounding the geyser would we expect to see deformation? Was the scanner itself located on deforming ground? Does porosity in the ground take up deformation so that only a small amount is expressed at the surface? Is a 20-minute interval sufficient to capture deformation if deflation occurs rapidly after inflation? These unknowns are mostly related to our understanding of geyser systems.

Suggestions for the Future
The lesson learned here is that when working in outdoor environments, scanner movement can affect point cloud accuracy when measuring very small changes in a surface. It is therefore important to have a way to verify whether or not the scanner moved during scanning, particularly when doing repeat scans from one position. Without this information it may be hard to tell if measured offsets actually reflect geologic processes.

Several actions can be taken to deal with possible scanner motion. The easiest of these is to take thorough field notes that indicate when abnormal events, such as gusts of wind, may have happened. The scanner tripod also needs to be stabilized as much as possible to avoid translation of the scanner origin. A shield can be set up around the scanner to protect it from wind. Level compensator readings can be used to correct scanner rotation in real time. A wind station can be set up as close to the scanner as possible to give a sense of wind speed and direction. Targets can be placed in the scan area to allow better registration of scans. Complete stabilization is hard to do when working in protected areas such as national parks, but these other options may help indicate when movement has happened and allow us to correct for it.

Acknowledgements: We would like to thank Dr. Michael Olsen for his guidance in analyzing this data set and his helpful suggestions for preparing this article. Thanks also go to Chuck Meertens and Jim Normandeau of UNAVCO for organizing data collection and initial processing, and Shaul Hurwitz for organizing the logistics of the Lone Star Geyser experiment. We are also grateful to Leica Geosystems for providing Oregon State University with the software used in data analysis and to the developers of CloudCompare for providing their excellent open source software.

Susan Schnur is a 2nd year Ph.D. student in marine geology at the College of Earth, Ocean and Atmospheric Sciences at Oregon State University. She has an M.Sc. in Geography from the University of Zrich and a B.A. in Geology from Carleton College. Her research focuses on submarine volcanism, seamount chains and geospatial methods for analyzing seafloor topography.

Adam Soule is an Associate Scientist in the Department of Geology & Geophysics at Woods Hole Oceanographic Institute. He has a Ph.D. from the University of Oregon. His research focuses on physical volcanology in terrestrial and submarine volcanic systems with an emphasis on magmatictectonic interactions at mid-ocean ridges.

Karlstrom, L., Hurwitz, S., Sohn, R., Vandemeulebrouck, J., Murphy, F., Rudolph, M.L., Johnston, M.J.S., Manga, M., and McCleskey, R.B. Eruptions at Lone Star Geyser, Yellowstone National Park, USA, Part 1: Energetics and Eruption Dynamics (in review).

Olsen, M.J. (2010). Scannin’ in the Wind. LIDAR News eMagazine, 2(6).

A 1.557Mb PDF of this article as it appeared in the magazine complete with images is available by clicking HERE