Leading NWP centers have agreed to create a database of their operational ensemble forecasts and open access to researchers to accelerate the development of probabilistic forecasting of high-impact weather.Objectives and cOncept. During the past decade, ensemble forecasting has undergone rapid development in all parts of the world. Ensembles are now generally accepted as a reliable approach to forecast confidence estimation, especially in the case of high-impact weather. Their application to quantitative probabilistic forecasting is also increasing rapidly. In addition, there has been a strong interest in the development of multimodel ensembles, whether based on a set of single (deterministic) forecasts from different systems, or on a set of ensemble forecasts from different systems (the so-called superensemble). The hope is that multimodel ensembles will provide an affordable approach to the classical goal of increasing the hit rate for prediction of high-impact weather without increasing the false-alarm rate. This is being taken further within The Observing System Research and Predictability Experiment (THORPEX), a major component of the World Weather Research Programme (WWRP) under the World Meteorological Organization (WMO). A key goal of THORPEX is to accelerate improvements in

Abstract. The limited-area ensemble prediction system COSMO-LEPS has been running every day at ECMWF since November 2002. A number of runs of the non-hydrostatic limited-area model Lokal Modell (LM) are available every day, nested on members of the ECMWF global ensemble. The limited-area ensemble forecasts range up to 120 h and LM-based probabilistic products are disseminated to several national and regional weather services. Some changes of the operational suite have recently been made, on the basis of the results of a statistical analysis of the methodology. The analysis is presented in this paper, showing the benefit of increasing the number of ensemble members. The system has been designed to have a probabilistic support at the mesoscale, focusing the attention on extreme precipitation events. In this paper, the performance of COSMO-LEPS in forecasting precipitation is presented. An objective verification in terms of probabilistic indices is made, using a dense network of observations covering a part of the COSMO domain. The system is compared with ECMWF EPS, showing an improvement of the limited-area high-resolution system with respect to the global ensemble system in the forecast of high precipitation values. The impact of the use of different schemes for the parametrisation of the convection in the limited-area model is also assessed, showing that this have a minor impact with respect to run the model with different initial and boundary condition.

No abstract

A high-resolution ensemble system, based on five runs of a limited-area model (LAM), is described. The initial and boundary conditions for the LAM integrations are provided by the representative members (RMs) selected from the European Centre for Medium-Range Weather Forecasts Ensemble Prediction System (EPS). @S meinbeax are grouped in five clusters; then, from each cluster, an RM is selected, according to the methodology described in the companion paper. The ability of the high-resolution ensemble system to predict the occurrence of heavy rainfall events (either five or six days ahead) is tested for four cases of floods over the Alpine region. Results show that, in two case-studies, the LAM integration corresponding to the RM of the highly populated cluster predicts the observed rainfall with a very good degree of time and spatial accuracy. In the other two cases, the extreme events are captured by at least one of the runs nested on the members of the less populated clusters. Probability maps constructed from LAM integrations provide great detail on the location of the regions affected by heavy precipitation and the information gained with respect to EPS probability maps and LAM deterministic forecasts is highlighted. The probabilistic estimates based on the LAM ensembles are also shown to be of valuable assistance to forecasters in issuing early flood alerts, contributing to the definition of a flood-risk alarm system. t In November 2000, the horizontal resolution of the operational EPS was increased to T~2 5 5 , Corresponding to a grid scale of approximately 80 km. @ Royal Meteorological Society, 2001. 209s 2096 C. MARSIGLI et al.and boundary conditions provided by the representative members (hereafter, RMs) of the ECMWF EPS. The RMs are selected first by applying a cluster analysis to the 51member EPS to define five clusters and, then, by identifying the RM of each cluster. Clusters are defined by considering the atmospheric flow at 700 hPa and by using the wind vector as clustering variable. Once the five clusters have been constructed, for each cluster the RM is defined as the member closest to all members of its own cluster and most distant from the members of the other clusters, with distances computed using an Ll norm applied to the precipitation field. The reader is referred to the companion paper, Molteni et al. (2001), and to Marsigli (1998) for a detailed description of the selection methodology. LEPS is based on integrations of the limitedarea model LAMBO (Limited Area Model Bologna), operational at ARPA-SMR since 1993. LAMBO runs are performed at high horizontal resolution (about 20 km) in order to resolve those orographic and mesoscale processes responsible for heavy -precipitation events. A probability of occurrence is assigned to each scenario, based on the population of the corresponding EPS cluster. In this way, it is possible to combine the ability of the EPS to highlight a set of possible evolution scenarios (keeping account of the intrinsic predictability of a particular synoptic situation...

No abstract