1Climate Change Research Centre, University of New South Wales, Sydney, Australia
2ARC Centre of Excellence for Climate System Science, University of New South Wales, Sydney, Australia
3International Pacific Research Center, University of Hawaii, Honolulu, Hawaii, USA
4Department of Atmospheric and Environmental Sciences, University at Albany, Albany, NY, USA
5Geophysical Fluid Dynamics Laboratory/NOAA, Princeton, New Jersey, USA
Abstract. It is vital to understand how the El Niño–Southern Oscillation (ENSO) has responded to past changes in natural and anthropogenic forcings, in order to better understand and predict its response to future greenhouse warming. To date, however, the instrumental record is too brief to fully characterize natural ENSO variability, while large discrepancies exist amongst paleo-proxy reconstructions of ENSO. These paleo-proxy reconstructions have typically attempted to reconstruct ENSO's temporal evolution, rather than the variance of these temporal changes. Here a new approach is developed that synthesizes the variance changes from various proxy data sets to provide a unified and updated estimate of past ENSO variance. The method is tested using surrogate data from two coupled general circulation model (CGCM) simulations. It is shown that in the presence of dating uncertainties, synthesizing variance information provides a more robust estimate of ENSO variance than synthesizing the raw data and then identifying its running variance. We also examine whether good temporal correspondence between proxy data and instrumental ENSO records implies a good representation of ENSO variance. In the climate modeling framework we show that a significant improvement in reconstructing ENSO variance changes is found when combining information from diverse ENSO-teleconnected source regions, rather than by relying on a single well-correlated location. This suggests that ENSO variance estimates derived from a single site should be viewed with caution. Finally, synthesizing existing ENSO reconstructions to arrive at a better estimate of past ENSO variance changes, we find robust evidence that the ENSO variance for any 30 yr period during the interval 1590–1880 was considerably lower than that observed during 1979–2009.