Journal on Uncertainty Quantification
May 18, 2012
In very broad terms, UQ in computational science and engineering has to do with describing the effects of error and uncertainty on results based on simulation and prediction of the behavior of constructed models of phenomena in physics, biology, chemistry, ecology, engineered systems, politics, and so on. Results from mathematical modeling are subject to errors and uncertainty emanating from a variety of sources; among them are uncertainty in data obtained from experiment and observation; limitations of physical modeling, including uncertain coefficients, discretization, approximation, and the need for emulation; problems in computer codes; and the difficulty of combining models into integrated systems. Quantifying the effects of these uncertainties is crucial to bringing to fruition the dream of being able to accurately model and predict real complex processes through computational simulators.
In specific terms, UQ embraces problems in a number of areas, including:
- Code verification
- Model validation and estimation of structural model error
- Computational error estimation for numerical solutions, e.g., a posteriori error analysis
- Data assimilation and model calibration
- Detection and forecasting of high-impact, rare events
- Emulation of computer models and dimension reduction
- Inference with complex multiscale, multiphysics models
- Representation of uncertainty and error, and integration of different types of uncertainty, e.g., parameter uncertainty, numerical error, and structural model error
- Inverse problems, decision making and optimization under uncertainty
- Treatment of high-dimensional spaces
The mathematical and statistical research required to address such problems is of great technical difficulty. General mathematical components of UQ include probability, measure theory, functional analysis, differential equations, graph and network theory, approximation theory, and ergodic theory. At the same time, nearly all aspects of the statistical sciences---including ANOVA expansions, stochastic processes, time series, classic inference, Bayesian analysis, importance sampling, nonparametric techniques, rare and extreme events, multivariable techniques---are relevant to UQ. Moreover, much of this research is necessarily carried out in interdisciplinary settings.
Research articles appropriate for publication in JUQ will present significant mathematical, statistical, algorithmic, and application advances in uncertainty quantification in the context of simulation, prediction, control, and optimization in science and engineering, and in related fields, such as sensitivity analysis, model validation, model calibration, data assimilation, and code verification. A key goal of JUQ is to nurture synergistic interactions between the mathematical, statistical, computational, and applications communities involved in uncertainty quantification and related areas. To this end, JUQ solicits papers describing new ideas that could lead to significant progress in methodology, computational/algorithmic aspects, and fully conceived applications of uncertainty quantification, as well as review articles on particular aspects.
Submissions for JUQ are now welcome at http://juq.siam.org. The list of the journal's associate editors---which includes many leaders working on problems relevant to UQ---and criteria for publication can be found at the journal website (www.siam.org/journals/juq.php). All articles will be rigorously reviewed, to the standards of SIAM and ASA. Our intention is to have the entire decision process for an article completed within six months, and initial reviews within three months.---Jim Berger, Don Estep, and Max Gunzburger.