Uncertainty Quantification in Ocean-Atmosphere Simulations Using Polynomial Chaos Methods and Gaussian Process Regression

Explore uncertainty quantification techniques in ocean-atmosphere simulations through this 48-minute conference presentation by Mohamed Iskandarani from the University of Miami. Discover how Gaussian Process Regression and Polynomial Chaos methods can be applied to quantify uncertainties in climate simulations and data analysis over the past decade. Learn about the strengths and limitations of these techniques for building model surrogates that explore subsets of model parameter spaces to perform uncertainty analysis. Examine two practical applications: calibrating model parameters using observational data from Typhoon Fanapi, and reconstructing velocity fields from massive surface drifter releases using Gaussian Process Regression. Gain insights into how these mathematical approaches can enhance the reliability and accuracy of earth system simulations by providing systematic methods for handling uncertainty in complex atmospheric and oceanic models.