This website stores cookies on your computer. These cookies are used to collect information about how you interact with our website and allow us to remember your browser. We use this information to improve and customize your browsing experience, for analytics and metrics about our visitors both on this website and other media, and for marketing purposes. By using this website, you accept and agree to be bound by UVic’s Terms of Use for web and social media privacy.  If you do not agree to the above, you can configure your browser’s setting to “do not track.”

Skip to main content
People Departments Maps Buildings Rooms A-Z
University
of Victoria
Apply Library
Sign in Online tools Sign out
Faculty of Graduate Studies

Diego Peña Palma

  • B.Sc. (Universidad Nacional Autónoma de México, 2024)
Notice of the Final Oral Examination for the Degree of Master of Science

Topic

An Adaptive Simultaneous Update Method to Accelerate the Calibration of the Stochastic Multicloud Model

Department of Mathematics and Statistics

Date & location

  • Thursday, October 8, 2026
  • 11:00 A.M.
  • Clearihue Building, Room B007

Examining Committee

Supervisory Committee

  • Dr. Boualem Khouider, Department of Mathematics and Statistics, University of Victoria (Supervisor)
  • Dr. Farouk Nathoo, Department of Mathematics and Statistics, UVic (Member)

External Examiner

  • Dr. Jahrul Alam, Department of Mathematics and Statistics, Memorial University of Newfoundland

Chair of Oral Examination

  • Dr. Brian Pollick, Department of Art History and Visual Studies, UVic

Abstract

The Stochastic Multicloud Model (SMCM) of Khouider et al. [2010] provides an effective probabilistic method for parameterizing unresolved tropical convection in General Circulation Models (GCMs). The SMCM improves the simulation of climate phenomena, but its effectiveness depends on the calibration of cloud transition timescales that are difficult to find just from observational data. In De La Chevrotière et al. [2014], calibrating the SMCM using Bayesian inference requires evaluating a likelihood function defined by solving the system of Kolmogorov differential equations, a computationally intensive process involving several matrix exponential computations. While this parallel implementation relies on CPU-based sparse matrix-vector operations via PETSc, it is computationally demanding.

In this thesis, we present a new efficient Bayesian inference approach to accelerate the calibration. The Adaptive Simultaneous Update Method (ASUM) is based on a reconstructed block updating [Knorr-Held and Rue, 2002] inference algorithm to benefit from data parallelism, implementing optimized sparse matrix-vector multiplication (SpMV) routines, efficient memory access patterns, and refined state space indexing routines for both CPU and GPU implementation. Benchmark results obtained with the Digital Research Alliance of Canada Trillium Cluster demonstrate a 240× speedup on a single CPU node compared to the original paper results, reducing Bayesian parameter estimation runtimes from weeks to hours. This performance enables Bayesian calibration across bigger cloud lattices and longer observational time series.

Back to oral exams