A global perspective of the limits of prediction skill of NWP models

Abstract

This paper discusses the scale-dependent growth of the global forecast uncertainties simulated by the operational ensemble prediction system of the European Centre for Medium-Range Weather Forecasts. It is shown that the initial uncertainties are largest in the tropics and have biggest amplitudes at the large scales. The growth of forecast uncertainties (ensemble spread) takes place at all scales from the beginning of forecasts. The growth is nearly uniform in the zonal wave numbers 1–5 and strongly scale-dependent in the larger wave numbers. Moreover, the growth from initial uncertainties at large scales appears dominant over the impact of errors cascading up from small scales. A decomposition of the ensemble spread in components associated with the balanced and unbalanced dynamics shows that the initial uncertainties are primarily in the unbalanced modes, especially at the subsynoptic scales. The growth of uncertainties is found to be faster in the balanced than in the unbalanced modes and after 0.5–1 day of forecasts the balanced errors become dominant except at the subsynoptic scales.