A Diurnal Predictability Barrier for Weather Forecasts

Abstract

AbstractIn this study, we investigate a diurnal predictability barrier (DPB) for weather predictions using an idealized model and observations. This DPB is referred to a maximum drop of predictability (e.g., autocorrelation) at a particular time of the day, regardless of the initial time. Previous studies demonstrated that a strong seasonal cycle of El Niño-Southern Oscillation (ENSO) growth rate is responsible for the seasonal predictability barrier of the ENSO in spring. This led us to investigate whether or not a strong diurnal cycle may generate a DPB. We study the DPB using an idealized model, the Lorenz 1963 model (Lorenz63), with the addition of a diurnal cycle. We find that diurnal growth rate can generate a DPB in this chaotic system, regardless of the initial error. Finally, by calculating the autocorrelation function using the hourly data of surface temperature, we explore the DPB at two stations in Wisconsin, USA and Beijing, China. A clear DPB feature is found at both stations. The dramatic drop of predictability at a specific time of the day is likely due to the diurnal variation of the system. This is a new feature that needs further study for short-term weather predictions.