Sensitivity of model output statistics on the Lorenz system

Abstract

This project studies the sensitivity of model output statistics (or MOS) in numerical weather prediction to lead times and parameters. In order to study MOS in an idealized setting, I use the simplified Lorenz system of differential equations as the physical model to explore the sensitivity of linear MOS in this non-linear system. MOS finds a statistical relationship between imperfect model variables and true variables using linear regression. The effect on the forecast error of lead time and parameter perturbation are investigated.