Climate change detection and attribution using multiple variables

The subject of climate change, particularly the possible contribution of human activities in forcing any such change, has generated much interest. Estimates of possible future climate change due to human forcing will be aided by attribution of recent observed changes to one or more causes.

Attribution of climate change is a difficult task. Studies that have sought to identify climate change using global indicators such as global-mean surface temperature have had difficulty in unequivocally establishing cause and effect. More recent studies have focussed on fingerprint methods, which make use of the spatial patterns of climate change to try and attribute the observed changes to one or more climate forcing factors. This study aims to review methods for the detection of climate change. In particular, it aims to develop two methods which allow for detection and attribution using the combination of several variables. It is hoped that the combination of multiple variables has more power to establish cause and effect than does the use of single variables alone.

Firstly, we seek to define a number of so-called global indices of climate change. These indices will hopefully provide us with a way of assessing climate change in a way that is both powerful, in terms of detection and attribution, as well as being easily understood. The indices have been selected based on earlier studies of climate change and on some of the climate change fingerprints that have been used commonly. Appropriate indices are those that capture key features of fingerprint patterns of climate change while still permitting relatively simple interpretation.

The second method involves a pattern type detection study based on principle component analysis of the spatial fields of several variables. These variables are selected in a similar manner to the indices. As with the indices, the use of more than one climate parameter will hopefully give us more information to attribute climate change. By making use of climate model experiments we can develop from these fields the modes or patterns of variability which represent the associated external forcing mechanisms and the dominant patterns of variability. These variable modes may then be compared with observations.

Multi-variate analysis of climate change is a relatively recent undertaking and this study aims to carefully identify appropriate variables for detection and attribution. Initial analysis of several climate indicators from observations and model data has been encouraging, with chosen indices of climate change showing a strong signal due to greenhouse gas changes and trends comparable with observations for the last 100 years. Analysis of changes in mean surface temperature, the surface temperature contrast between land and ocean, diurnal temperature range and meridional temperature gradient will be presented. Changes in the amplitude of the annual cycle in temperature over land will also be discussed as will important hemispheric differences in the identified fingerprints.

Results will be shown for 150 years of observations and from CSIRO9 Coupled Climate Model simulations and Hadley Centre Climate Model (HadCM2) simulations. Results from the Coupled Model Intercomparison Project (CMIP2) will also be discussed.