Conventional Filtering and Picking Methodologies Applied to DAS Microseismic Data

Fibre-optic Distributed Acoustic Sensing (DAS) is a new alternative technology that has been rapidly developing for both passive and active seismic monitoring. DAS systems are most sensitive to uniaxial strain along the fibre and therefore record data similar to a one-component geophone. However, unlike a geophone, they record this data at finely spaced intervals along the full fibre length which covers the entire wellbore. This dense spatial sampling can lead to very large volume datasets, up to terabytes per day, which can be technically difficult to handle and a challenge to accurately pick P- and S-wave arrivals from. Traditional manual picking approaches are not viable on such sizeable data volumes and hence automated solutions must be developed for efficient analysis of DAS data. Here we test how conventional filtering and picking methods from geophone seismology can be adapted, combined and applied to a DAS microseismic dataset in order to determine which combination can best automatically process and pick the characteristics of DAS recorded microseismic events. These approaches are tested on field data containing events from a single hydraulic fracture stage recorded by Silixa Ltd.'s iDAS System . Events are recorded on a horizontal cable of 3081 traces at 1m spacing and 1000Hz. 302 events are automatically detected and windowed, of which over 190 were of suitable quality for further investigation. We manually picked the events at 20m spacing to provide baseline results and use these picks to approximate a velocity model and then locate the events. We then test a variety of filtering (median, Weiner, bandpass and f-k) and automated picking methodologies including standard STA/LTA and a guided STA/LTA. We also test a migration algorithm which removes the need for any direct picking, instead using time-shifting and semblance-stacking for putative locations. The combined approaches that we have developed have produced equivalent results to manual picking but are significantly faster while analysing much higher density of traces.