engleski

Dynamics of the Adriatic in Real-Time (DART) : multi-model superensembles prediction in a coastal ocean regime

Multi-model Super-Ensembles (SE) aimed at combining optimally different models have been shown to improve significantly atmospheric weather predictions. In the coastal ocean, complex, yet poorly understood dynamics, the presence of small-scales processes, the lack of real-time data at appropriate spatio-temporal resolution and limited reliability of operational models so far prevented the application of SE methods. Here, we report results from state-of-the-art super-ensemble techniques based on dynamic combinations of SEPTR (a trawl-resistant bottom mounted platform transmitting in near real-time) data and a series of eight operational models ran during the DART (Dynamics of the Adriatic in Real-Time) experiment in a coastal area. Kalman filter and particle filter SE-based methods which allow for dynamic evolution of weights and associated uncertainty show increased temperature prediction skill (+10%) as compared to single models. The latter method copes with non-Gaussian error statistics and reduces the forecast uncertainty by a further 30%.

Adriatic; super-ensembles; models; prediction

nije evidentirano