The EKF-AUS-NL algorithm implemented without the linear tangent model and in presence of parametric model error

Abstract

In this paper we propose a C++-software package implementing the algorithm EKF-AUS-NL (Extended Kalman Filter with Assimilation in the Unstable Space with NonLinear evolution) designed to perform data assimilation in the unstable space when the Jacobian of the differential equation cannot be calculated. We also propose a simple approach to take into account the presence of the model error in the framework of the EKF-AUS-NL. The software performs the data assimilation using the EKF-AUS-NL algorithm with a dynamical systems defined as a generic time evolution routine separately implemented. We present two illustrative examples based on the Lorenz96 and SLAM systems. Keywords: Kalman-Filter, AUS, Lorenz96, SLAM