Accurate and reliable weather forecasts and warnings are essential for the public authorities, the economic sector and the private individuals in order (for them) to take weather-dependent decisions in timely manner, or to plan their outdoor activities. Especially in the Alpine region, a very high spatial resolution is required to accurately predict local weather events such as thunderstorms, thermally induced mountain and valley wind systems or the Foehn effect. The model COSMO-2 with a 2.2 km grid used by MeteoSwiss until now has been able to represent major Alpine valleys and orographic structures with an extension of at least 10 km. However, it was not always capable of simulating the development of storm clouds or wind conditions in small or medium-size valleys such as, for example, the Valle Maggia.
COSMO-1 in operation as from today
Today, following a six months long test period, MeteoSwiss puts COSMO-1into operation. The new forecasting model with a grid spacing of 1.1 km is an important improvement in the prediction of local weather phenomena in the Alpine region and is expected to produce more detailed regional forecasts and more accurate local warnings. As of now, even wind conditions in narrow valleys such as the Valle Maggia will be simulated with a high degree of precision.
COSMO-1 will be run every three hours for up to +33 hours into the future. For warnings concerning the following day, it will even run up to +45 hours ahead once daily. The operation of COSMO-2 will cease in autumn 2016.
New forecasting model is run on a new supercomputer
COSMO-1 requires 20 times the computing power of COSMO-2; it is therefore run on the new supercomputer “Piz Kesch”, which came into operation in September 2015 at the Swiss National Supercomputing Centre (CSCS) in Lugano. “Piz Kesch” was custom-built by MeteoSwiss in collaboration with CSCS, the Center for Climate Systems Modeling (C2SM) at the ETH Zurich, and the companies Cray and NVIDIA to fit the exact purposes of MeteoSwiss. Worldwide, MeteoSwiss is the first national meteorological service to opt for a new computer architecture for its operational forecasting models: by using graphic processors (GPUs) and a software which is optimised for GPUs, the supercomputer not only calculates simulations significantly faster but is also much more energy efficient, which allows for an increased computing capacity at same costs.