Predictability in complex terrain: a preliminar study

It is widely known that the atmospheric behaviour is chaotic and, thereby, a single result is not exactly predictable. The forecast skill depends not only on the accuracy of the initial conditions and on the model itself, but also on the instabilities of the flow. The smallest imperfection in the forecast model or the smallest error in the initial conditions will lead to a total loss of skill in the weather forecast. To take into account the stochastic nature of the evolution of the atmosphere, ensemble modelling have been developed to explicitly acknowledge the uncertainty of atmospheric model predictions. At the beginning, global ensemble prediction system have been developed and regional ensemble systems followed later on.

In this preliminary study different ensemble systems have been used. A PEPS (Poor's man Ensemble Prediction System) and a REPS (Regional Ensemble Prediction System) have been the basic two ones. The PEPS system uses the results from different models: GFS, ECMWF, WRF, MSM, ISM at various grid resolutions and different physical parameterizations. The REPS system have been set to separately use 10 runs coming from the GEPS (Global Ensemble Prediction System), operationally run at Epson Meteo Centre every day.

Recently, a version of REPS which uses the ensemble Kalman filter technique (EnKF) have been developed for research use. The EnKF technique is a relatively new method of data assimilation and generation of ensemble perturbations. It has been set with our MSM model to generate the mesoscale ensemble predictions. The forecast model parameters are the same of the current operational setting. If EnKF converges, it is expected that the analysis members well represent actual error structures and EnKF may be considered an ideal method to generate ensemble perturbations.

Very preliminary results have been obtained from selected case studies in a complex terrain environment in order to investigate the behaviour of the systems and analyze the predictability of mesoscale phenomena.