Tunining of HARMONIE assimilation and forecast systems

Principal Investigator

Roger Randriamampianina
Norwegian Meteorological Institute
P.O. Box 43, Blindern
N-0313 Oslo
Norway

rogerr@met.no

Other researchers: Andrea Storto and Frank Thomas Tveter

Project description

Background

The installation and development of the HARMONIE system at the Norwegian Meteorological Institute is related to two ongoing research projects: to the IPY-TORPEX and to the ABC Projects. In the frame of the EUMETSAT-fellow ABC-project we study the impact of “superobs” derived from binary cloud observation from CloudSat on the HARMONIE analysis and forecast. Within the IPY-TORPEX Project, investigation of the impact of various satellite data on the HARMONIE analysis and forecast is being carried out, with special emphasis on IASI data. We also intend to participate in the EUCOS observation network design impact study in 2009. The installation of the HARMONIE system as well as the implementation of the (conventional and non-conventional) data is almost finished.

Objectives

Within the frame of the projects our aim is to develop a complex assimilation system, in terms of both configuration and scripting framework, suitable for operation application at the Norwegian Meteorological Institute. The support in the years 2009-2010 would be needed for achievement of the following project purposes:

1. To find the optimal way to implement and tune the assimilation of a variety of remote sensed observations, combining all these information together, in terms of relative weight, spatial resolution and thinning, targeting of sensitive region and vertical levels;
2. To perform Observing System Experiments (OSE) to evaluate the relative impact of various observation network designs as well as all the conventional and remote-sensed observations on the analysis and forecasts;
3. To test different algorithms for the assimilation of new observations (e.g. cloudiness observations, IASI);
4. To test more sophisticated assimilation scheme for adaptive bias correction (Variational bias correction) and for flow-dependent ensemble based background error statistics;
5. To extend the impact studies over Arctic’s; The IPY-THORPEX project aims to improve the weather forecast over Arctic by implementing the IASI data in the HARMONIE system. Tuning and bias correction studies with the IASI data are being performed for a smaller domain and the results will further be extended to the bigger domain, containing the part of the Arctic. This requires huge amount of calculation time;

Specificity of configuration set up:

Current configuration on HPCE supercomputer takes advantage of 24 processor cores for the analysis for cases without IASI / SEVIRI data or VARBC technique for bias correction. Otherwise (when IASI or SEVIRI data or VARBC technique is used) the analyses need 2 nodes of 24 processors instead of 1. For the forecasts 48 processor cores are needed requiring a maximum memory of 25 GB.

For the Norwegian (small) domain at 11 km resolution the upper-air three-dimensional variational data assimilation and surface optimum interpolation analysis normally takes about 20 minutes, while at about 30 min utes are needed for a 48 hours forecast model integration, including the pre-processing and post-processing tasks. Our preliminary test showed, that in case of large Norwegian at 16 km resolution domain the analysis and the forecast require 3 nodes of 24 processors and 2 nodes of 48 processors, respectively.

Additional information

New Project for 2009

Would accept support for 1 year only, if necessary.