An Investigation of the Dynamics of Forecast Uncertainty in Global Atmospheric Model Forecasts

We investigate the dynamics of the forecast uncertainty in global atmospheric model forecasts. In particular, we study the process by which uncertainties due to initial condition and model errors propagate toward the synoptic scales. We analyze ensemble forecast data from the TIGGE data set and carry out numerical experiments with CAM. The results suggest that there are important differences, especially at the shorter than 3-day forecast times, in the behavior of the different ensemble forecast systems included in TIGGE. The different operational global ensemble systems are clearly tuned to satisfy different optimality conditions. While the systems that employ stochastic schemes for the representation of the effect of model uncertainty perform better in the medium-range, they also have an elevated level of bias in the long-range.