Implicit and explicit treatments of model error in numerical simulation
Danny Smyl
TL;DR
The paper addresses the pervasive problem of model error in numerical simulations by distinguishing implicit (uncertainty-absorbing) and explicit (discrepancy-correcting) treatments. It surveys foundational methods across Bayesian inverse problems, data assimilation, probabilistic numerics, multi-fidelity modeling, and machine-learning–driven corrections, and it shows how these approaches can be integrated into PDE solvers and large-scale inference workflows. Key contributions include formal taxonomies, critical discussions of identifiability and calibration challenges, and practical guidance on combining implicit and explicit strategies to improve predictive performance and uncertainty quantification. The work underscores the growing role of discrepancy modeling in engineering and Earth-system science, and it outlines future directions such as automated discrepancy discovery, real-time model-error correction in digital twins, and diffusion-model–based discrepancy representations to augment uncertainty estimates.
Abstract
Numerical simulations of physical systems invariably suffer from model errors stemming from unmodeled physics, idealizations, and discretization. This article provides a review of techniques developed in the past two decades to approximate and account for these model errors, both implicitly and explicitly. Beginning from fundamental definitions of model-form versus numerical error, we frame model error in inverse problems, data assimilation, and predictive modeling contexts. We then survey major approaches: the Bayesian approximation error framework for implicit error quantification, embedded internal error models for structural uncertainty, probabilistic numerical methods for discretization uncertainty, model discrepancy modeling in Bayesian calibration and its recent extensions, machine-learning-based discrepancy correction, multi-fidelity and hybrid modeling strategies, as well as residual-based, variational, and adjoint-driven error estimators. Throughout, we emphasize conceptual underpinnings of implicit versus explicit error treatment and highlight how these methods improve predictive performance and uncertainty quantification in practical applications ranging from engineering design to Earth-system science. Each section provides an overview of key developments with an extensive list of references to facilitate further reading. The review is written for practitioners of large-scale computational physics and engineering simulation, emphasizing how these methods can be incorporated into PDE solvers, inverse problem workflows, and data assimilation systems.