Evolutionary computation for climate and ocean forecasting: “El Niño forecasting”

It has been recently postulated that the irregular dynamics of El Niño-Southern Oscillation (ENSO) may embed a low order chaotic process. If true, some aspects of the ENSO variability are predictable. Here, from observations of the annual average Sea Surface Temperature (SST) of the tropical Pacific Ocean (1949-1981) and using an evolutionary genetic analysis, we characterise the chaotic nature of ENSO. The mathematical model of the ENSO dynamics obtained is cross-validated with the observations for the next fifteen years (1982-1997). The extracted ENSO dynamics show a chaotic behaviour characterised by a correlation dimension d=3.5, in agreement with numerical simulations. An eight years forecast (1997-2005) of the ENSO shows a strong cooling (La Niña) during 1999 and a significant El Niño in 2001. Finally, a longer term forecast reveals a warming of about 0.2 Celsius per century that might be associated with very low frequency variability of ENSO.