Exploring the Limits of Earthquake Offset Detection in GPS Geodesy

Abstract:

When an earthquake occurs, coseismic deformation moves the Earth surface at nearby GPS stations, causing a sharp, immediate discontinuity in the position time series. These offsets in GPS time series give key information about the earthquake, yet also need to be corrected when estimating important parameters such as velocity. But how far away from an earthquake are GPS stations affected at a detectable level?

The Nevada Geodetic Laboratory (NGL) maintains a list of potential step events for >18,000 GPS time series based on station locations and earthquake epicentral location and magnitude (M). Currently, NGL estimates step size at earthquake event times if the station-to-epicenter distance is less than an empirical radius of influence defined as r₀=10^0.5M-0.8, where r₀ is distance in degrees. This circular domain around the event is purposefully simple and large enough to include areas where offsets are likely above the threshold of resolvability, but may also include stations below that threshold. GPS stations within the radius may not be observed to have a significant offset at these times; e.g. step magnitude may be effectively zero if the station is distant enough from an earthquake hypocenter or in a direction along the azimuth of rupture. Conversely, offsets may be resolvable beyond r₀ when the source directs displacement along some azimuths more than others. The true perimeter of the region of offset resolvability depends on the source parameters and complexity, and Earth response.

We evaluate the effectiveness of our current radius of influence, plus additional new models, for detecting offsets caused by earthquake events using the 2019 M7.1 Ridgecrest and 2020 Monte Cristo Range, Nevada events as case studies. For Ridgecrest, r₀ flagged 1,113 GPS time series inside a 562 km radius, and we found resolvable offsets at >500 km from the epicenter. We compare the pattern of stations with detectable steps inside a 2r₀ distance to identify missed offsets and compare them to Okada model predicted offsets to identify where they become indistinguishable from noise. Our findings will help build an improved model to automatically flag potential steps used to calculate GPS time series offsets for recent events, impact NGL’s offset and velocity estimations, and have applications for rapid GPS data products following earthquakes.