ENSO - WorldWeatherAttribution/wwa-wiki GitHub Wiki

Detrended Niño3.4 index

The Niño3.4 index is defined as the mean sea surface temperature (SST) anomaly from 5S-5N, 120W-170W. To remove the trend induced by global warming, we subtract the mean tropical SST anomaly from 20S-20N, following van Oldenborgh et al. (2021). We're interested in how the index co-varies with our main time series rather than using fixed thresholds to identify El Niño or La Niña events, so the individual calendar months are not renormalised to have the same variances as the standard Niño3.4 index.

Defining factual and counterfactual ENSO states for climate models

During the attribution step, the detrended Niño3.4 index derived from the climate model is standardised so that the subset covering the same period as the observations has mean 0 and variance 1; the ‘factual’ climate is then defined as having the model’s GMST in the event year and the observed value of the detrended Niño3.4 index, standardised in the same way. This removes any potential biases in the results due to differences between the amplitude of the modelled Niño3.4 index and that observed.

Evaluating climate models

When the statistical model used in the attribution step includes ENSO, we evaluate the climate models for both the variance of the (unstandardised) detrended Niño3.4 index and the correlation between the detrended Niño3.4 index and the hazard index. Uncertainties are obtained by bootstrapping and the range of variances and correlations is compared to the range in the observed data, as for the other model parameters.