The Parallel Algorithm of the Non-differentiable Maximum Likelihood Ensemble Filter on Cloud Environments

By employing the non-differentiable Maximum Likelihood Ensemble Filter (MLEF) method, a parallelized data assimilation system without the requirement of linearity and differentiability is developed. Under this system, several experiments are done on cloud environments with one-dimensional nonlinear advection model and simulated observations. The experimental results show that the parallel MLEF algorithm works efficiently with satisfied speedups and robust performance, indicating that it is appropriate to apply the MLEF method to cloud.