Uncertainty Quantification in Induced Seismicity Forecasting

Some human industrial activities like oil and gas extraction induce earthquakes. These earthquakes are occasionally large enough to cause concern about the possibility of building damage. We address the need for uncertainty quantification in the forecasting of such earthquakes. We use an integrated modeling applied to the largest producing natural gas field in western Europe which is close to the city of Groningen in the Netherlands. Gas production started in the early 1960s and started to induce detectable seismicity 30 years later. We propose and assess the performance of an algorithm for uncertainty quantification in the forecast of earthquake numbers and magnitudes. We forecast the number of future earthquakes and an estimate of the most probable maximum magnitude based on a hypothetical future gas extraction scenarios.

Forecast the number of events and uncertainty quantification.

You see in this figure the forecast of seismicity and uncertainty quantification. The blue line is the observed number of events in each year in the Groningen gas field. The white region is used as the training set to find our model parameter estimation to forecast the seismicity for the period of the grey region. The cyan is our maximum likelihood estimate. the red lines are inside the 94% percent of the epistemic uncertainty. The black lines are >90 percent confidence interval of the number of events accounting for both the aleatoric and epistemic uncertainity.


Forecast magnitude of events.

Here you the evolution of the expected value (with respect to model parameters) of the probability (with respect to the distribution) of exceeding certain magnitudes for nontapered (solid lines) and tapered (dashed) distributions. 

Paper:

Hojjat Kaveh, Pau Batlle, Mateo Acosta, et al. Induced Seismicity Forecasting with Uncertainty Quantification: Application to the Groningen Gas Field. Authorea. June 06, 2023.