Optimization of Kinematic GPS Data Analysis for Large Surface Deformation from the July 2003 Dome Collapse at Soufrière Hills Volcano, Montserrat

Several volcanic systems in the western US and Alaska (part of PBO) as well as the Soufriere Hills volcano on Montserrat (CALIPSO) have spatially dense continuous GPS networks that have been operating for close to a decade. Because GPS signals are affected during transmission through the atmosphere, it is important to resolve any contribution of atmospheric effects to apparent changes in position and therefore to obtain the best estimate of both. This is especially critical in the Caribbean or other tropical regions, where the effect of tropospheric water vapor is large as well as spatially and temporally variable. Several proximal cGPS sites (<10 km from the vent) collected data at 30 sec intervals during the 12-13 July 2003 eruption and massive dome collapse of Soufrière Hills Volcano (SHV). Data were originally processed treating the antennae as a kinematic buoy using GIPSY-OASIS-II (v. 5) and high-rate (30 s) final, precise orbit, clock, and earth orientation parameter products from JPL. In the original GOA-II analysis, the parameters for the random walk of the wet zenith delay, elevation cutoff, troposphere horizontal gradient and the rate of change of the random walk of position were kept at the default values suggested by JPL for precise kinematic positioning. After reviewing the position time-series, one GPS station, HERM recorded a maximum vertical displacement of -1.98 m from its mean, with negligible horizontal movement, rebounding within an hour. This estimate of vertical site displacement was an order of magnitude larger than those estimated at other sites on SHV. We report here our revised processing using GOA-II (v. 6.2), updated processing procedures, including the use of VMF1 grid files and APCs for the antenna/radome combinations, and newly released IGS08 data products from JPL. We have reprocessed all available cGPS from the July 2003 dome collapse event on SHV using a grid-search method to examine the appropriate stochastic atmosphere and position parameters to increase the precision of GPS position estimates during the eruption. BGGY, a station located 48 km northeast on Antigua, was used as a control to optimize the parameters for modeling the atmospheric variations more accurately for this type of environment, since BGGY is subjected to the similar weather patterns but was unaffected by volcanic activity at SHV. The final stochastic parameters were selected to yield the lowest variance in the kinematic position time-series at BGGY, then, HERM was reprocessed using the same parameters. The apparent vertical movement at HERM has been reduced substantially, and now has a maximum of 2.5 cm with a variation of 30 cm in the zenith wet troposphere estimate. We conclude that the original default parameters used to process that GPS observations over-constrained possible atmospheric variation for this tropical environment, producing apparently large dynamic position changes. Our new results now reflect actual dynamic ground deformation during the massive dome collapse and may be used to develop improved models for volcanic processes that occur over time scales of minutes to hours at SHV and other tropical volcanoes.