A conceptual framework for discrete inverse problems in geophysics

TL;DR

This paper proposes a comprehensive conceptual framework for discrete inverse problems in geophysics, addressing their properties, parameter identifiability, stability, and various analytical approaches to improve understanding and application.

Contribution

It introduces a formal, general framework for defining and analyzing discrete inverse problems in geophysics, including properties like identifiability, posedness, and conditioning.

Findings

Formal definitions of inverse problem properties

Discussion on the role of data and Bayesian approach

Framework supports sensitivity analysis and optimization

Abstract

In geophysics, inverse modelling can be applied to a wide range of goals, including, for instance, mapping the distribution of rock physical parameters in applied geophysics and calibrating models to forecast the behaviour of natural systems in hydrology, meteorology and climatology. A common, thorough conceptual framework to define inverse problems and to discuss their basic properties in a complete way is still lacking. The main goal of this paper is to propose a step forward toward such a framework, focussing on the discrete inverse problems, that are used in practical applications. The relevance of information and measurements (real world data) for the definition of the calibration target and of the objective function is discussed, in particular with reference to the Bayesian approach. Identifiability of model parameters, posedness (uniqueness and stability) and conditioning of the inverse problems are formally defined. The proposed framework is so general as to permit rigorous definitions and treatment of sensitivity analysis, adjoint-state approach, multi-objective optimization.