Long-term winter temperatures in central Mediterranean: forecast skill of an ensemble statistical model

This study proposes a forecast skill scheme for winter temperatures in the central Mediterranean Sub-regional Area. An original series of mean winter (December to February) temperatures spanning the period 1698–2010 is the basis for developing long-term climate prediction. A procedure was identified where a predictable structure was first provided by reducing noise via the Empirical Mode Decomposition method and then an Ensemble Climate Prediction (ECP) was applied. The predictability limit of the decomposed series was assessed by applying a collection of methods from linear and nonlinear time series analysis (Hurst coefficients, Lyapunov exponents and wavelet patterns). The analysis was based on a set of tools that are suitable to discover the manifestation of a possible trajectory of projected temperature change. The progress of the winter temperature forecast was inferred by an Exponential Smoothing with a Damped (ESD) multiplicative trend model. ECP-ESD hindcast experiments were tested and ensemble forecasts run until the year 2040. ECP-ESD yields an ensemble mean path that consists: (1) of a sharp decline in winter temperature in the current decade; (2) more widespread predictions for 2020–2030, with fluctuations near the present multi-decadal average; and (3) a sharp warming for the 2030–2040 decade. Our results are compatible with projections for European winters and major climatic indices such as the Pacific Decadal Oscillation and the Atlantic Multi-decadal Oscillation.