Submit Now

Description:

Various examples of industrial workflows for uncertainty quantification demonstrated business value in inputting uncertainty measures into the risk analysis and decision-making processes. However, still there many misunderstandings and confusion in the community even on the definitions for uncertainty in the business context. This interdisciplinary workshop will focus on the quantification, analysis and use of uncertainties associated to seismic modelling and inversion (imaging and velocity model building). 

Methods to quantifying uncertainties on inverted Earth model parameters (e.g. velocity or reflectivity) become increasingly used in the industrial context of tomography and imaging, which is crucial to reduce exploration and production risks. Bayesian inference is an attractive framework to estimate those uncertainties from “input” information (observational, modelling and prior uncertainties). Ultimately one uses Monte Carlo (MC)-like techniques to sample the corresponding posterior probability distribution.

However this is not an easy task because of:

• The large computational cost in presence of a large number of parameters (commonly 1 million in the industry); the main bottleneck is then the cost of the forward modelling problem, which prevents a large number of MC simulations.
• The evaluation of the various sources of “input” uncertainties that still remains a challenge and can have dramatic impact on the results, thus on risk analysis.
  

In order to reduce the forward modelling cost various possibilities have been investigated:

• Construction of a “surrogate” (a simple statistical model).
• Sparser representations of Earth model parameters.
• Linearized approaches that proved to be efficient especially in the framework of tomography.
• The evaluation of the “input” uncertainties and their ranking has been sporadically investigated. The surrogate method allows an evaluation for the modelling uncertainties (through a propagation of the uncertainties from the Earth model to the predicted data).
  

The aim of the workshop is to address the challenge of a quantification of the uncertainties at a reasonable cost and a better evaluation of the “input” uncertainties:

• What are the current industrial applications?
• How to build accurate surrogates?
• Can we couple MC-like techniques and linearized ones to improve the efficiency?
• How to design sparser parameterizations?
• Can we rank and quantify the measurement and modelling uncertainties? 
  

We invite submissions on all aspects of uncertainty quantification across the fields of seismic modelling and inversion including theoretical advances, practical applications and case studies. Contributions are welcome from the seismological as well as the mathematical and statistical communities.

Workshop format:

Presentations followed by discussions. 

Who should attend:

Researchers interested in the estimation and use of uncertainties associated to seismic imaging.