An Information-Theoretic Framework for Improving Imperfect Dynamical Predictions Via Multi-Model Ensemble Forecasts

Michal Branicki, Andrew J. Majda

Research output: Contribution to journalArticle

Abstract

This work focuses on elucidating issues related to an increasingly common technique of multi-model ensemble (MME) forecasting. The MME approach is aimed at improving the statistical accuracy of imperfect time-dependent predictions by combining information from a collection of reduced-order dynamical models. Despite some operational evidence in support of the MME strategy for mitigating the prediction error, the mathematical framework justifying this approach has been lacking. Here, this problem is considered within a probabilistic/stochastic framework which exploits tools from information theory to derive a set of criteria for improving probabilistic MME predictions relative to single-model predictions. The emphasis is on a systematic understanding of the benefits and limitations associated with the MME approach, on uncertainty quantification, and on the development of practical design principles for constructing an MME with improved predictive performance. The conditions for prediction improvement via the MME approach stem from the convexity of the relative entropy which is used here as a measure of the lack of information in the imperfect models relative to the resolved characteristics of the truth dynamics. It is also shown how practical guidelines for MME prediction improvement can be implemented in the context of forced response predictions from equilibrium with the help of the linear response theory utilizing the fluctuation–dissipation formulas at the unperturbed equilibrium. The general theoretical results are illustrated using exactly solvable stochastic non-Gaussian test models.

Original languageEnglish (US)
Pages (from-to)489-538
Number of pages50
JournalJournal of Nonlinear Science
Volume25
Issue number3
DOIs
StatePublished - Jun 1 2015

Fingerprint

Multi-model
Imperfect
Forecast
Ensemble
Prediction
Uncertainty Quantification
Framework
Relative Entropy
Linear Response
Dynamical Model
Prediction Error
Information Theory
Prediction Model
Convexity
Forecasting
Information theory
Entropy

Keywords

  • Information theory
  • Multi-model ensembles
  • Time-dependent prediction
  • Uncertainty quantification

ASJC Scopus subject areas

  • Applied Mathematics
  • Modeling and Simulation
  • Engineering(all)

Cite this

An Information-Theoretic Framework for Improving Imperfect Dynamical Predictions Via Multi-Model Ensemble Forecasts. / Branicki, Michal; Majda, Andrew J.

In: Journal of Nonlinear Science, Vol. 25, No. 3, 01.06.2015, p. 489-538.

Research output: Contribution to journalArticle

@article{9a3dfa87378f4c798ea2d25cf5c7bdbc,
title = "An Information-Theoretic Framework for Improving Imperfect Dynamical Predictions Via Multi-Model Ensemble Forecasts",
abstract = "This work focuses on elucidating issues related to an increasingly common technique of multi-model ensemble (MME) forecasting. The MME approach is aimed at improving the statistical accuracy of imperfect time-dependent predictions by combining information from a collection of reduced-order dynamical models. Despite some operational evidence in support of the MME strategy for mitigating the prediction error, the mathematical framework justifying this approach has been lacking. Here, this problem is considered within a probabilistic/stochastic framework which exploits tools from information theory to derive a set of criteria for improving probabilistic MME predictions relative to single-model predictions. The emphasis is on a systematic understanding of the benefits and limitations associated with the MME approach, on uncertainty quantification, and on the development of practical design principles for constructing an MME with improved predictive performance. The conditions for prediction improvement via the MME approach stem from the convexity of the relative entropy which is used here as a measure of the lack of information in the imperfect models relative to the resolved characteristics of the truth dynamics. It is also shown how practical guidelines for MME prediction improvement can be implemented in the context of forced response predictions from equilibrium with the help of the linear response theory utilizing the fluctuation–dissipation formulas at the unperturbed equilibrium. The general theoretical results are illustrated using exactly solvable stochastic non-Gaussian test models.",
keywords = "Information theory, Multi-model ensembles, Time-dependent prediction, Uncertainty quantification",
author = "Michal Branicki and Majda, {Andrew J.}",
year = "2015",
month = "6",
day = "1",
doi = "10.1007/s00332-015-9233-1",
language = "English (US)",
volume = "25",
pages = "489--538",
journal = "Journal of Nonlinear Science",
issn = "0938-8974",
publisher = "Springer New York",
number = "3",

}

TY - JOUR

T1 - An Information-Theoretic Framework for Improving Imperfect Dynamical Predictions Via Multi-Model Ensemble Forecasts

AU - Branicki, Michal

AU - Majda, Andrew J.

PY - 2015/6/1

Y1 - 2015/6/1

N2 - This work focuses on elucidating issues related to an increasingly common technique of multi-model ensemble (MME) forecasting. The MME approach is aimed at improving the statistical accuracy of imperfect time-dependent predictions by combining information from a collection of reduced-order dynamical models. Despite some operational evidence in support of the MME strategy for mitigating the prediction error, the mathematical framework justifying this approach has been lacking. Here, this problem is considered within a probabilistic/stochastic framework which exploits tools from information theory to derive a set of criteria for improving probabilistic MME predictions relative to single-model predictions. The emphasis is on a systematic understanding of the benefits and limitations associated with the MME approach, on uncertainty quantification, and on the development of practical design principles for constructing an MME with improved predictive performance. The conditions for prediction improvement via the MME approach stem from the convexity of the relative entropy which is used here as a measure of the lack of information in the imperfect models relative to the resolved characteristics of the truth dynamics. It is also shown how practical guidelines for MME prediction improvement can be implemented in the context of forced response predictions from equilibrium with the help of the linear response theory utilizing the fluctuation–dissipation formulas at the unperturbed equilibrium. The general theoretical results are illustrated using exactly solvable stochastic non-Gaussian test models.

AB - This work focuses on elucidating issues related to an increasingly common technique of multi-model ensemble (MME) forecasting. The MME approach is aimed at improving the statistical accuracy of imperfect time-dependent predictions by combining information from a collection of reduced-order dynamical models. Despite some operational evidence in support of the MME strategy for mitigating the prediction error, the mathematical framework justifying this approach has been lacking. Here, this problem is considered within a probabilistic/stochastic framework which exploits tools from information theory to derive a set of criteria for improving probabilistic MME predictions relative to single-model predictions. The emphasis is on a systematic understanding of the benefits and limitations associated with the MME approach, on uncertainty quantification, and on the development of practical design principles for constructing an MME with improved predictive performance. The conditions for prediction improvement via the MME approach stem from the convexity of the relative entropy which is used here as a measure of the lack of information in the imperfect models relative to the resolved characteristics of the truth dynamics. It is also shown how practical guidelines for MME prediction improvement can be implemented in the context of forced response predictions from equilibrium with the help of the linear response theory utilizing the fluctuation–dissipation formulas at the unperturbed equilibrium. The general theoretical results are illustrated using exactly solvable stochastic non-Gaussian test models.

KW - Information theory

KW - Multi-model ensembles

KW - Time-dependent prediction

KW - Uncertainty quantification

UR - http://www.scopus.com/inward/record.url?scp=84939948968&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84939948968&partnerID=8YFLogxK

U2 - 10.1007/s00332-015-9233-1

DO - 10.1007/s00332-015-9233-1

M3 - Article

VL - 25

SP - 489

EP - 538

JO - Journal of Nonlinear Science

JF - Journal of Nonlinear Science

SN - 0938-8974

IS - 3

ER -