Towards efficient probabilistic characterisation of induced seismic sources in the groningen gas field

Abstract

We present a probabilistic scheme to invert for hypocenters and moment tensors (MTs) of induced seismic events in 2D heterogeneous media. Our scheme is based on a variant of the Hamiltonian Monte Carlo scheme that uses the linearization of a misfit function. This algorithm, however, requires a sufficiently accurate prior, which could be taken from earthquake catalogues. In our case, a detailed subsurface velocity model allows the hypocenter prior to be obtained via a first arrival based algorithm. Fixing the induced event's location to this prior, the MT's prior is subsequently obtained. Having all necessary priors, we then recover the posterior probability density of the induced event's source characteristics, i.e., hypocenter and MT. By comparing the recovered posterior with the directly modelled events, we infer that our scheme is efficient and practical to invert for source characteristics of induced seismic events.