A local particle filter and its Gaussian mixture extension implemented with minor modifications to the LETKF

Abstract. A particle filter (PF) is an ensemble data assimilation method that does not assume Gaussian error distributions. Recent studies proposed local PFs (LPFs), which use localization, as in the Kalman filter, to apply PF efficiently for high-dimensional dynamics. Among others, Penny and Miyoshi (2016) developed LPF form of transform matrix (LETKF). The LETKF has been widely accepted various geophysical systems, including numerical weather prediction (NWP) models. Therefore, implementing consistently with existing code useful. This study develops a software platform its mixture extension (LPFGM) by making slight modifications simplified global climate model known Simplified Parameterizations, Primitive Equation Dynamics (SPEEDY). series idealized twin experiments were accomplished under ideal-model assumption. With large inflation relaxation prior spread, showed stable performance dense observations but became unstable sparse observations. LPFGM more accurate than both In addition parameter, regulating resampling frequency amplitude kernels was important LPFGM. spatially inhomogeneous observing network, superior sparsely observed regions, where background spread non-Gaussianity larger. SPEEDY-based LETKF, LPF, systems are available open-source on GitHub (https://github.com/skotsuki/speedy-lpf, last access: 16 November 2022) can be adapted models relatively easily, case LETKF.