Seismic hazard is used for national, regional, and local level to ensure safe constructions in specific areas. In the mining industry this information is valuable e.g. to  design infrastructure or rock support, to reduce the risk of rock burst and to minimise the risk of locating personnel in hazardous areas. Seismic hazard can be estimated by different approaches. Probabilistic Seismic Hazard Assessment (PSHA) is one approach to estimate the seismic hazard and is defined as the probability that an earthquake will occur within a certain area and time interval causing vibrations with an intensity larger than a given threshold. 

This thesis contains an introduction to various aspects of PSHA and highlights some of the limitations with current assumptions and methods, together with a summary of my scientific contributions to PSHA. These contributions aim to improve PSHA in mines at different steps of the calculation chain. Their primary focus is on obtaining reliable input and output parameters (i.e. with uncertainties) at each step in the calculation chain, necessary for a reliable hazard assessment. This is done by adopting a Bayesian workflow, with comprehensive model validation, and where the underlying uncertainties are included for proper weighting of the covariates in each step. Additionally, it contains a collection of three papers (Paper A, Paper B and Paper C) focusing on these aspects. The short summary of these papers follows.

Paper A Provides a path to reliable auto-processing of seismic events by describing how to capture the unknown and changing environment. It also highlights some of the human limitations with today's Routine Manual Processing (RMP) in terms of data truncation and discrepancies in processing results between individuals (e.g. in classification and hypocentre estimation). Additionally, the paper compares the automatic processing system BEMIS (developed by Wille Törnman and Jesper Martinsson) with RMP regarding event classification and hypocentre estimation when both approaches are subjected to the same data. This paper is an overview of the philosophy adopted in BEMIS, highlighting the strengths of using a Bayesian approach by: capturing, including, and propagating further the uncertainties in each step in the processing chain to obtain robust and valid estimates of the estimands of interest.

Paper B Describes a fully automatic and robust Bayesian method to estimate precise and reliable model parameters describing the observed S-wave spectra. These model parameters are essential for determination of source parameters of an earthquake (e.g. source radius, seismic moment, magnitude etc). The model includes the observed noise and a combined empirical Green’s function. It captures source-, receiver-, and path-dependent terms in the description of the observed spectra by combining a physical source and attenuation model with a spatially and event-size dependent empirical compensation. The proposed method propagates estimation uncertainties along the entire processing chain starting from the hypocentre location and delivers reliable uncertainty description of the estimands.

Paper C Describes the relationship between the recorded seismic activity and the: seismic decay time, planned production rate, production size and mining depth, for the seven largest orebodies in LKAB's iron ore mine in Malmberget. This relationship is described by a mine-wide Bayesian hieSavkararchical model and is an important part to individually customise the production rate for each orebody in the mine, make short-term predictions of future seismicity given planned productions, and to find out in what way the available predictors affect the seismicity. The model is validated using a comprehensive procedure and the results are precise and valid in terms of central tendency and dispersion.