Estimating the seasonal predictability of global precipitation: an empirical approach

Abstract

The asymptotic predictability of global land-surface precipitation is estimated empirically at the seasonal time scale at lead times from zero to 12 months. Predictability is defined as the maximum achievable predictive skill for a given model assuming all the relevant predictors are included and that an infinitely long training sample is available for parameter estimation, and represents an approximate upper bound to the predictive skill of statistical, and possibly dynamical, seasonal forecasting approaches. To estimate predictability, a simple linear regression model is formulated based on the assumption that land surface precipitation variability can be divided into a component forced by low-frequency variability in external boundary conditions which potentially can be predicted one or more seasons into the future, and a ‘weather noise’ component which originates from nonlinear dynamical instabilities in the atmosphere and which is not predictable beyond about 10 days. The external boundary condition is represented by an orthogonal (principal component analysis) transformation of the global sea surface temperature anomaly (SSTA) field, as this field constitutes the dominant driver of global precipitation variability.