On the existence of multiple climate regimes

Abstract New techniques are presented for testing the three main hypotheses about the probability distribution of the climate system: multinormal (single regime), unimodal but not multinormal (single regime), and multimodal (multiple regimes). Rather than searching for evidence that confirms the multimodal hypothesis expected from the chaos and other strongly nonlinear paradigms, our strategy is to try and reject the simplest single‐regime hypothesis of multinormality expected for aggregate indices of many local weather degrees of freedom. Concerning multiple climate regimes in the northern hemisphere, we find no strong evidence in the available monthly mean reanalysis data for rejecting the single‐regime multinormal hypothesis in favour of the multimodal hypothesis. A simple non‐parametric method is presented for transforming state space into a more homogeneous probability space that makes regimes easier to identify. A spatial point process test is used in this space to demonstrate that the hemispheric clusters are not significantly different to what could be expected from sampling a unimodal distribution. Based on the observed data, the single‐regime multinormal hypothesis can not be rejected at the 5% level of significance and so provides the simplest useful model for the probability distribution for the northern hemisphere geopotential‐height field. Copyright © 2004 Royal Meteorological Society