Statistical predictability of Euro-Mediterranean subseasonal anomalies: The TeWA approach

Abstract

It is widely known from energy balances that global oceans play a fundamental role in atmospheric seasonal anomalies via coupling mechanisms. However, numerical weather prediction models still have limitations in long-term forecasting due to their nonlinear sensitivity to initial deep oceanic conditions. As the Mediterranean climate has highly unpredictable seasonal variability, we designed a complementary method by supposing that (1) delayed teleconnection patterns provide information about ocean–atmosphere coupling on subseasonal timescales through the lens of (2) partially predictable quasi-periodic oscillations, since (3) forecast signals can be extracted by smoothing noise in a continuous lead-time horizon. To validate these hypotheses, subseasonal predictability of temperature and precipitation was analyzed at 11 reference stations in the Mediterranean area in the 1993–2021 period. The novel method, presented here, consists of combining lag-correlated teleconnections (15 indices) with self-predictability techniques of residual quasi-oscillation based on Wavelet (cyclic) and ARIMA (linear) analyses. The prediction skill of this Teleconnection-Wavelet-ARIMA (TeWA) combination was cross-validated and compared to that of the SEAS5-ECMWF model (3 months ahead). Results show that the proposed TeWA approach improves the predictability of first-month temperature and precipitation anomalies by 50–70% compared with the forecast of SEAS5. On a moving-averaged daily scale, the optimum prediction window is 30 days for temperature and 16 days for precipitation. The predictable ranges are consistent with atmospheric bridges in teleconnection patterns (e.g., ULMO) and are reflected by spatial correlation with SST. Our results suggest that combinations of the TeWA approach and numerical models could boost new research lines in subseasonal-to-seasonal forecasting.