The problem of how to make useful forecasts at lead times between a week and a month is a challenging and often neglected one. Forecast information on these time scales is in great demand from users. This is an area that NOAA has traditionally not focused on in the past. CDC researchers have been addressing the problem on two fronts; 1) by trying to extract the maximum information from ensemble NWP model forecasts, and 2) by investigating statistical forecast methods that complement the NWP ensembles by exploiting predictable signals not well represented in current models. Our research thus far suggests that the NWP and statistical approaches are complementary, and provide information that is independent to some degree. The challenge is to combine the two in an optimal manner, yielding forecasts that are superior to either individually. Due to the low-frequency nature of the phenomena at these forecast ranges, determining the optimal combination would require generating a long (20+ year) dataset of ensemble forecasts with a fixed model to estimate the forecast error statistics with the necessary accuracy. Work is currently underway at CDC to create such a dataset, which will also be useful in several other applications not discussed here.
Contributed by: J. Barsugli, T. Hamill, H. Hendon, B. Liebmann, M. Newman, P. Sardeshmukh, and J. Whitaker.