Local multi-model prediction system (LaMPS)

Principal Investigator(s)

Piero Chessa,
Servizio Agro-Meteorologico per la Sardegna,
Viale Porto Torres,
119-07100 Sassari, Sardinia.
ITALY

chessa@sar.sardegna.it

Other investigators: M Bonavita, L Torrisi, M Ferri, G Ficca, C Dessy, C Castiglia and M Marrocu

Project Description

The LaMPS project goal is to investigate methodologies to improve the forecast capabilities in the field of operational early warning systems for intense/severe weather events. This is a hot research topic in meteorology and many competing approaches are currently under study and experimentation.

The method we envisage is to run an ensemble of different limited area models (LAM) over a European domain starting from different initial and boundary conditions (IC/BC) from general circulation models (GCM) and large-area regional models.

This approach has been chosen in order to allow a more complete sampling of atmospheric variability, taking into account both the analysis and model errors. This might result in an improvement with respect, for example, to a downscaling of GCM Ensemble Prediction Systems operated by means of a single regional model. The use of a multi-model, multi-analysis approach and a suitable post-processing technique both for deterministic output (e.g. Kalman Filtering, Multilinear Regression) and probabilistic outcomes (e.g. Bayesian Model Averaging, ROC Curve fitting, Dressing Methods, etc.) should help in coping with the analyses and model errors.

The main task of the present project is to provide well calibrated probabilistic forecasts at high and very high spatial resolutions. In the initial phase the LaMPS project will draw from the experience gained by some of the participants with the MuSE activity which is an internal SAR and CRS4 project aimed at implementing operationally a multimodel/multianalysis limited area ensemble (MuSE). MuSE is currently pre-operational and it is based on three Limited Area Models (BoLAM of the ISAC/CNR - Italy, MM5 of the Pennsylvania State University and RAMS of the Colorado State University) and three different IC/BC provided by ECMWF and NCEP GCMs.

In this special project MuSE will be enlarged allowing for more regional models (HRM of the DWD and, possibly, non-hydrostatic models) and BC/ICs (UGM 3D-var analysis). The models will be implemented and run at ECMWF and different model configurations will be explored in order to test different ensemble setups. As already mentioned the main effort will be devoted to calibrating the probabilistic forecasts and comparing them to those obtained by similar systems already in place.

The main focus will be put on surface or near surface parameters (temperatures, winds, precipitation and instability indexes) with special attention to intense/severe events liable to be of concern to civil protection authorities. The spatial domain for verification will be the Italian peninsula and its major islands and a high resolution network of ground stations will be used to this end.

For more details, please refer to the latest progress report.

Additional Information

Project started in 2006.