Data assimilation in a sparsely observed one-dimensional modeled MHD system

A one dimensional non-linear magnetohydrodynamic (MHD) system has been introduced to test a sequential optimal interpolation assimilation technique that uses a Monte-Carlo method to calculate the forecast error covariance. An ensemble of 100 model runs with perturbed initial conditions are used to construct the covariance, and the assimilation algorithm is tested using Observation Simulation Experiments (OSE’s). The system is run with a variety of observation types (magnetic and/or velocity fields) and a range of observation densities. The impact of cross covariances between velocity and magnetic fields is investigated by running the assimilation with and without these terms. Sets of twin experiments show that while observing both velocity and magnetic fields has the greatest positive impact on the system, observing the magnetic field alone can also effectively constrain the system. Observations of the velocity field are ineffective as a constraint on the magnetic field, even when observations are made at every point. The implications for geomagnetic data assimilation are discussed.