Strategies for Optimizing the Barotropic-Baroclinic Splitting of the Time-Stepping Algorithm in High Resolution Ocean Models

Siddhartha Bishnu1,2, Mark R Petersen3 and Bryan Quaife2, (1)Los Alamos National Laboratory, Computer, Computational and Statistical Sciences, Los Alamos, United States, (2)Florida State University, Scientific Computing, Tallahassee, FL, United States, (3)Los Alamos National Laboratory, Los Alamos, NM, United States
Abstract:
My research involves improving the barotropic-baroclinic splitting of the time-stepping algorithm of the Model for Prediction Across Scales - Ocean (MPAS-Ocean) with a view to improving the numerical stability and solution accuracy while reducing the computational time. More specifically, I have been studying (a) different filters for time-averaging the intermediate instantaneous barotropic modes and including the 'mean' solution in the time derivative for the next baroclinic (large) time step, and (b) variations of the forward-backward time-stepping algorithm for advancing these barotropic modes. The primary purpose of the time-averaging filters in (a) is to minimize the aliasing and mode-splitting errors while ensuring the stability of the time-stepping scheme. The forward-backward algorithm in (b) consists of a predictor and an optional corrector stage with a set of weighting parameters, an optimum combination of which can enhance solution accuracy. I have programmed a one-dimensional shallow water equation solver in object-oriented Python for simulating the propagation of a surface gravity wave, where I have tested a number of filters and time-stepping schemes. In this poster, I will compare the efficiency and accuracy of various designs in the simplified code and global MPAS-Ocean simulations.