A Spectral Analysis of Predictability and Error in an Operational Forecast System

Abstract

A spectral analysis of error at 24, 48 and 72 hours in an operational forecast system is carded out for a winter and a summer month. The scale-dependent behavior of the error and the several factors governing error growth, namely the nonlinear production term, the nonlinear transfer of energy between scales and the model-related error source term are evaluated. It is found that the largest errors occur at low wavenumbers where the amplitude of the flow is largest, while the largest relative errors occur at high wavenumbers when error is approaching saturation. Early in the forecast the growth of error is governed by the model error source term which supplies error at all scales. At later forecast times the nonlinear production and interaction terms become important. Error doubling times and the flux of error through wavenumber space are also calculated. The predictability of the flow resides in the low wavenumber region. The behavior of the error and its transient and stationary parts markedly with season at these scales.