Advice on the computation of peakground-velocity confidence regions for events in gas fields other than the Groningen gas field

On request by the Technical committee on ground movement (Tcbb), a study was carried out to define the region affected by induced earthquakes related to gas exploitation1 from fields other than the Groningen gas field. This request is based on advise that the Tcbb formulated in 2019, on a national approach for handling damage due to mining.

We built a database of a few thousand peak-ground velocity (PGV) recordings, both from induced events inside and outside Groningen. Subsequently, we assessed which existing ground-motion model best fits this Dutch PGV database.

It was found that the empirical PGV model for Groningen (Bommer et al., 2019) was equally suited for induced events in- and outside Groningen and showed the best fit to the database. There was no marked difference for events with and without a shielding overburden. The selected model was adapted at short distance and for low magnitudes, outside the applicable magnitude range of the original model, to obtain an improved fit to the PGV database. A depth dependence was added to make the updated model (BMR2) applicable to source regions with nucleation depths other than 3 km.

The BMR2 model has the same aleatory variability as the original empirical PGV model. This variability incorporates, among others, epicentral uncertainty and variation in site effects, radiation pattern and stress drop. The BMR2 model, together with local recordings, is used for computing PGV contours for 2 mm/s until the maximum level, with a 1, 10 and 50% probability of exceedance. In the current PGV database, the lowest magnitude event for which 2 mm/s has been measured, is 1.9. For a 1% probability of exceedance, the BMR2 model reaches 2 mm/s at a local magnitude M of 1.53. Therefore, only for M ≥ 1.5 or when a PGV recording is larger than 2 mm/s, a PGV threshold region computation is started up. The recordings are used to estimate and remove an event term. Aside from a small nugget, the local PGV recordings are used as ground truths at their recording site and away from that site they are combined with the model to obtain a combined PGV field. The radius of influence of these recordings, that is the distance at which the weight of the model and recording is equal, is computed with the PGV database to be 2.7 km.

The computation of PGV contours is illustrated numerically and with three events that occurred outside the Groningen gas field. For one of these events (Dalen 17-7-2018) the event term could largely be computed, but there were no near-source recordings to locally perturb the modeled PGV field. For the Warder 04-06-2018 event there was a lack of both local and regional PGV recordings which made it necessary to use the full variability of the model, leading to relatively large PGV threshold regions. For the Roswinkel 25-10-2000 event, the availability of 4 recordings within a few kilometers from the source led to a near-epicentre PGV field dominated by the measurements.