[MGI30-P08] A data assimilation library with Python for parallel computing
キーワード：data assimilation、particle filter、parallel computing、Python
Parallel computing is essential to reduce the computational time in the ensemble-based data assimilation. However, it requires skills in parallel programming. It is sometimes a hard task to attain high computational efficiency with ensemble-based data assimilation. In particular, the particle filter (PF) algorithm, which is applicable to nonlinear and/or non-Gaussian problems, contains a procedure difficult to parallelize.
It is thus challenging to achieve high computational efficiency with the PF even for experienced users.
We have developed a Python library named P-cubed (Python Parallelized Particle Filter Library), that enables us to use parallel-ready PF algorithms with high parallel efficiency. Now we are also planning to attach a module of other data assimilation algorithms such as the ensemble Kalman filter to this library. In this presentation, we introduce the parallelized PF algorithms which are already available in P-cubed and explain the design and characteristics of the library. Future prospects of this library will also be discussed.
It is thus challenging to achieve high computational efficiency with the PF even for experienced users.
We have developed a Python library named P-cubed (Python Parallelized Particle Filter Library), that enables us to use parallel-ready PF algorithms with high parallel efficiency. Now we are also planning to attach a module of other data assimilation algorithms such as the ensemble Kalman filter to this library. In this presentation, we introduce the parallelized PF algorithms which are already available in P-cubed and explain the design and characteristics of the library. Future prospects of this library will also be discussed.