Probability-based methods for quantifying nonlinearity in the ENSO

Abstract.

Robust statistical tools have been used to investigate non-normality and nonlinearity of the El Niño Southern Oscillation (ENSO) in observations and coupled model simulations. The analysis confirms previous suggestions that the observed Niño-3 sea surface temperature (SST) anomalies are positively skewed. The non-linearity is estimated using a simple nonlinear stochastic model, which relates the sea surface temperature anomalies to the observed thermocline depth anomalies in the Niño-3 region. There is evidence that saturation of SST only occurs when the thermocline is deep. The nonlinearity has also been estimated for the Niño-3 SST indices from twenty four different coupled models participating in the El Niño Simulation Intercomparison Project (ENSIP). Large differences are found between models and observations. In particular, the majority of the coupled models underestimate the nonlinearity seen in the observed Niño-3 sea surface temperature index. More than half of the models have Niño-3 SST indices that are normally distributed at 99% confidence level. Only a few models exhibit significant nonlinearity yet this tends to be rather different in character from the nonlinearity seen in the observations. Furthermore, no significant association is found between the means and the spread nor between the spread and the skewness for the different coupled model Niño-3 SST indices.