A Practical Two-Phase Approach to Improve the Reliability and Efficiency of Markov Chain Monte Carlo Directed Hydrologic Model Calibration

Abstract

Markov chain Monte Carlo (MCMC) methods are widely used in hydrology and other fields for posterior inference in a Bayesian framework. A properly constructed MCMC sampler is guaranteed to converge to the correct limiting distribution, but convergence can be very slow. While most research is focused on improving the proposal distribution used to generate trial moves in the Markov chain, this work instead focuses on efficiently finding an initial population for population-based MCMC samplers that will expedite convergence. Four case studies, including two hydrological models, are used to demonstrate that using multi-level single linkage implicit filtering stochastic global optimization to initialize the population both reduces the overall computational cost and significantly increases the chance of finding the correct limiting distribution within the constraint of a fixed computational budget.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Mar 01, 2020
Accession Number
AD1092579

Entities

People

  • Brian E. Skahill
  • Jeffrey S. Baggett

Tags

Communities of Interest

  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Algorithms
  • Bayesian Networks
  • Case Studies
  • Computational Science
  • Data Science
  • Drainage Basins
  • Information Science
  • Markov Chains
  • Markov Processes
  • Models
  • Monte Carlo Method
  • Numerical Analysis
  • Probabilistic Models
  • Probability
  • Probability Distributions
  • Random Variables
  • Stochastic Processes

Readers

  • Computational Fluid Dynamics (CFD)
  • Computational Modeling and Simulation
  • Statistical inference.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference
  • AI & ML - Machine Learning Algorithms