Quantifying the Impacts of Initial and Wind Forcing Uncertainties on Oceanic ForecastUncertainties in the Gulf of Mexico.

Mohamed Iskandarani, University of Miami - RSMAS, Miami, FL, United States, Guotu Li, Duke, Mechanical Engineering and Material Science, Durham, NC, United States, Matthieu Le Henaff, CIMAS/University of Miami, Miami, FL, United States, Justin Winokur, Sandia National Laboratories, Albuquerque, NM, United States, Olivier P. Le Maitre, Laboratoire d'Informatique pour la Mécanique et les Science de l'Ingénieur, Orsay, France and Omar M Knio, King Abdullah University of Science and Technology, Thuwal, Saudi Arabia
Abstract:
Polynomial Chaos Expansions (PCE) are used to quantify the impacts of uncertainties in initial conditions and wind forcing on HYCOM forecasts of the circulation in the Gulf of Mexico. The input uncertainties are encoded in the amplitude of variability modes identified using Empirical Orthogonal Function decompositions: one for a model state time series, and the other for a wind time series. An ensemble of simulations is used to build accurate surrogates for the forecasted sea surface height and mixed layer depth using a basis pursuit denoising method to filter model noise. The surrogates are then used to estimate various statistical measures of the model output. A variance analysis shows that, for the 30 days time scale considered in the study, the uncertainties in the Sea Surface Height are dominated by uncertainties in initial conditions whereas uncertainties in the mixed layer are dominated by uncertainties in the wind stress.
Back to: Developments and Ocean Applications of Data Assimilation, Uncertainty, and Sensitivity Analyses III