A new implementation of digital filtering initialization schemes for HIRLAM

Xiang-Yu Huang and Xiaohua Yang

Abstract

A new implementation of Digital Filtering Initialization (DFI) schemes has been made for the HIRLAM forecast system and preliminary version has been implemented in the reference system as a beta release (version 5.0.3). The motivations for the new implementation are three-fold:

Earlier DFI implementations require separate runs with forecast models using different namelists. They also involve unnecessary model state input/output (I/O). Errors often occur when making changes to the scripts. It is therefore desirable for the maintenance and development to have a simpler script for DFI, which activates only a single model run, with no extra I/O, as in the previous setup (version 5.0.2).
The earlier DFI implementations in HIRLAM contain many DFI-related statements in the main program, GEMINI, with the initialization step controlled at script level. This makes it difficult to understand and to modify. For further development, it is desirable to have a clean main program with only one subroutine call to DFI.
Many digital filter schemes have been tested since the first DFI implementation, but only the most effective one has been kept in the current reference model. Considering the fact that other DFI schemes may be more advantageous for different purposes, it seems useful to enable the choice of different schemes and filters.

In this technical report, we present the recoded initialization interface which addresses the above-mentioned issues. The new codes are validated with parallel assimilation runs. Since this report is mainly aimed for being used as a technical manual for users interested in DFI schemes in the HIRLAM system, we also examine some of the new features available in the modified code. For example it is shown that the incremental DFI schemes significantly improve the spin-up aspects for precipitation, with only moderately increased initial noise level.

download report