Peter Cervelli

Estimating source parameters from deformation data, with an application to the March 1997 earthquake swarm off the Izu Peninsula, Japan

We have applied two Monte Carlo optimization techniques, simulated annealing and random cost, to the inversion of deformation data for fault and magma chamber geometry. These techniques involve an element of randomness that permits them to escape local minima and ultimately converge to the global minimum of misfit space. We have tested the Monte Carlo algorithms on two synthetic data sets. We have also compared them to one another in terms of their efficiency and reliability. We have applied the bootstrap method to estimate confidence intervals for the source parameters, including the correlations inherent in the data. Additionally, we present methods that use the information from the bootstrapping procedure to visualize the correlations between the different model parameters. We have applied these techniques to GPS, tilt, and leveling data from the March 1997 earthquake swarm off of the Izu Peninsula, Japan. Using the two Monte Carlo algorithms, we have inferred two sources, a dike and a fault, that fit the deformation data and the patterns of seismicity and that are consistent with the regional stress field.

Earthquakes triggered by silent slip events on Kīlauea volcano, Hawaii.

Slow-slip events, or ‘silent earthquakes’, have recently been discovered in a number of subduction zones including the Nankai trough in Japan, Cascadia, and Guerrero in Mexico, but the depths of these events have been difficult to determine from surface deformation measurements. Although it is assumed that these silent earthquakes are located along the plate megathrust, this has not been proved. Slow slip in some subduction zones is associated with non-volcanic tremor, but tremor is difficult to locate and may be distributed over a broad depth range. Except for some events on the San Andreas fault, slow-slip events have not yet been associated with high-frequency earthquakes, which are easily located. Here we report on swarms of high-frequency earthquakes that accompany otherwise silent slips on Kīlauea volcano, Hawaii. For the most energetic event, in January 2005, the slow slip began before the increase in seismicity. The temporal evolution of earthquakes is well explained by increased stressing caused by slow slip, implying that the earthquakes are triggered. The earthquakes, located at depths of 7–8 km, constrain the slow slip to be at comparable depths, because they must fall in zones of positive Coulomb stress change. Triggered earthquakes accompanying slow-slip events elsewhere might go undetected if background seismicity rates are low. Detection of such events would help constrain the depth of slow slip, and could lead to a method for quantifying the increased hazard during slow-slip events, because triggered events have the potential to grow into destructive earthquakes.

Sudden aseismic fault slip on the south flank of Kilauea volcano

One of the greatest hazards associated with oceanic volcanoes is not volcanic in nature, but lies with the potential for catastrophic flank failure. Such flank failure can result in devastating tsunamis and threaten not only the immediate vicinity, but coastal cities along the entire rim of an ocean basin. Kilauea volcano on the island of Hawaii, USA, is a potential source of such flank failures and has therefore been monitored by a network of continuously recording geodetic instruments, including global positioning system (GPS) receivers, tilt meters and strain meters. Here we report that, in early November 2000, this network recorded transient southeastward displacements, which we interpret as an episode of aseismic fault slip. The duration of the event was about 36 hours, it had an equivalent moment magnitude of 5.7 and a maximum slip velocity of about 6 cm per day. Inversion of the GPS data reveals a shallow-dipping thrust fault at a depth of 4.5 km that we interpret as the down-dip extension of the Hilina Pali–Holei Pali normal fault system. This demonstrates that continuously recording geodetic networks can detect accelerating slip, potentially leading to warnings of volcanic flank collapse.

Inversion of GPS data for spatially variable slip‐rate on the San Andreas Fault near Parkfield, CA

We analyze GPS data collected from 1991-1998 at 35 sites near the Parkeld segment of the San Andreas Fault. Inverting the resultant site velocities for the distri- bution of interseismic slip-rate on the San Andreas reveals an area of low slip-rate on thefault extendingfrom between Middle Mountain and Carr Hill to southeast of Gold Hill. This slip-rate patternis similar to thatfoundbyHarris and Segall(1987) using trilateration data collected between 1966 and 1984. We infer a deep slip-rate (33 mm/yr) and depth of the transition between seismogenic and non-seismogenic slip(14km)thatagree betterwithindependentgeologic ev- idence than those found in the 1987 study. In contrast to Harris and Segall(1987), wendnoevidenceoffault-normal contraction.

The mechanics of unrest at Long Valley caldera, California: 1. Modeling the geometry of the source using GPS, leveling and two-color EDM data

We surveyed 44 existing leveling monuments in Long Valley caldera in July 1999, using dual frequency global positioning system (GPS) receivers. We have been able to tie GPS and leveling to a common reference frame in the Long Valley area and computed the vertical deformation by differencing GPS-based and leveled orthometric heights. The resurgent dome uplifted 74±7 cm from 1975 to 1999. To define the inflation source, we invert two-color EDM and uplift data from the 1985–1999 unrest period using spherical or ellipsoidal sources. We find that the ellipsoidal source satisfies both the vertical and horizontal deformation data, whereas the spherical point source cannot. According to our analysis of the 1985–1999 data, the main source of deformation is a prolate ellipsoid located beneath the resurgent dome at a depth of 5.9 km (95% bounds of 4.9–7.5 km). This body is vertically elongated, has an aspect ratio of 0.475 (95% bounds are 0.25–0.65) and a volume change of 0.086 km3 (95% bounds are 0.06–0.13 km3). Failure to account for the ellipsoidal nature of the source biases the estimated source depth by 2.1 km (35%), and the source volume by 0.038 km3 (44%).

The reawakening of Alaska's Augustine volcano

Augustine volcano, in south central Alaska, ended a 20-year period of repose on 11 January 2006 with 13 explosive eruptions in 20 days. Explosive activity shifted to a quieter effusion of lava in early February, forming a new summit lava dome and two short, blocky lava flows by late March (Figure 1). The eruption was heralded by eight months of increasing seismicity, deformation, gas emission, and small phreatic eruptions, the latter consisting of explosions of steam and debris caused by heating and expansion of groundwater due to an underlying heat source.

Volcano monitoring using the Global Positioning System: Filtering strategies

Permanent Global Positioning System (GPS) networks are routinely used for producing improved orbits and monitoring secular tectonic deformation. For these applications, data are transferred to an analysis center each day and routinely processed in 24-hour segments. To use GPS for monitoring volcanic events, which may last only a few hours, real-time or near real- time data processing and subdaily position estimates are valuable. Strategies have been researched for obtaining station coordinates every 15 min using a Kalman filter; these strategies have been tested on data collected by a GPS network on Kilauea Volcano. Data from this network are tracked continuously, recorded every 30 s, and telemetered hourly to the Hawaiian Volcano Observatory. A white noise model is heavily impacted by data outages and poor satellite geometry, but a properly constrained random walk model fits the data well. Using a borehole tiltmeter at Kilaueas summit as ground-truth, solutions using different random walk constraints were compared. This study indicates that signals on the order of 5 mm/h are resolvable using a random walk standard deviation of 0.45 cm/ AAA p . Values lower than this suppress small signals, and values greater than this have significantly higher noise at periods of 1-6 hours.

Vulcanology: Interaction between Kilauea and Mauna Loa

After almost a decade of very slow rates of deformation, Mauna Loa in Hawaii, the largest volcano on Earth, began inflating in May 2002; at the same time, a high-volume effusive episode began at its neighbour Kilauea. We have found a correlation between these events at a very short timescale, detected by continuous deformation monitoring. This remarkable observation suggests that there is a crustal-level interaction between the magma systems of Mauna Loa and Kilauea, reviving a century-old controversy over the relationship between these two volcanoes on the basis of differences in their lava chemistry and in their patterns of eruptive behaviour.

Renewed inflation of Long Valley Caldera, California (2011 to 2014)

Slow inflation began at Long Valley Caldera in late 2011, coinciding with renewed swarm seismicity. Ongoing deformation is concentrated within the caldera. We analyze this deformation using a combination of GPS and InSAR (TerraSAR-X) data processed with a persistent scatterer technique. The extension rate of the dome-crossing baseline during this episode (CA99 to KRAC) is 1 cm/yr, similar to past inflation episodes (1990–1995 and 2002–2003), and about a tenth of the peak rate observed during the 1997 unrest. The current deformation is well modeled by the inflation of a prolate spheroidal magma reservoir ∼7 km beneath the resurgent dome, with a volume change of ∼6 × 106 m3/yr from 2011.7 through the end of 2014. The current data cannot resolve a second source, which was required to model the 1997 episode. This source appears to be in the same region as previous inflation episodes, suggesting a persistent reservoir.

Inflation rates, rifts, and bands in a pāhoehoe sheet flow

The margins of sheet flows—pāhoehoe lavas emplaced on surfaces sloping Inflation and rift-band formation is probably cyclic, because the pattern we observed suggests episodic or crude cyclic behavior. Furthermore, some inflation rifts contain numerous bands whose spacing and general appearances are remarkably similar. We propose a conceptual model wherein the inferred cyclicity is due to the competition between the fluid pressure in the flow9s liquid core and the tensile strength of the visco elastic layer where it is weakest—in inflation rifts. The viscoelastic layer consists of lava that has cooled to temperatures between 800 and 1070 °C. This layer is the key parameter in our model because, in its absence, rift banding and stepwise changes in the flow height would not occur.