Regional Scatter Diagrams
Western N America
Explanation of scatter diagrams (for details see full report)
In the scatter diagrams, the x-axis shows temperature changes in oC, the y-axis precipitation changes in percent. Each scatter point represents a single model-simulated temperature/ precipitation response to one forcing scenario. The scenario is depicted by the colour of the point (A1FI - red, A2 - grey, B1 - green and B2 - violet). The shape of the symbol defines the model: CCSRNIES (crosses), CSIRO Mk2 (squares), CGCM2 (X), ECHAM4/OPYC3 (diamonds), GFDL-R30 (snowflakes), HadCM3 (circles) and NCAR DOE PCM (triangles). Solid symbols correspond to responses directly inferred from AOGCM runs; open symbols are those calculated by the pattern-scaling method. Scaled points have been omitted from the diagrams for the earliest projection period 2010-2039. Since it is difficult to determine the exact locations of points depicted in the scatter diagrams, this information is also available to download in numerical format.
The ovals centred on the origin indicate the 95% Gaussian contour ellipses of the natural internal
tridecadal variability of temperature and precipitation, derived from unforced 1000-year AOGCM runs performed by CGCM2 (orange) and HadCM3 (blue); for further information see section 3.1 of the report. If a scatter point falls distinctly outside the ellipses, the model-simulated joint temperature/precipitation change is statistically significant. Note that inspection of the diagrams indicates statistically significant changes in the large majority of cases analyzed. This is due to strong temperature changes which usually lie well outside the range of natural variability. In contrast, precipitation changes exceed natural variability in far fewer cases. To assess the statistical significance of the temperature and precipitation changes separately, the dimensions of the contour ellipses on the x- and y-axes should be employed as described in section 3.1.
One must bear in mind that, in addition to the response to human-induced forcing, the modelled climate change includes contributions due to noise and, perhaps also, to climate drift. The influence of noise is strongest in the short-term projections, which is why pattern-scaling has not been applied for the earliest projection period. Moreover, present climate models are far from complete, and there are components of the climate system that we do not yet understand properly. Therefore, it is not impossible that the response of the real climate system could fall outside the range projected by the model simulations discussed in the report.
We are grateful to Dr. David Viner at the University of East Anglia and Drs. Francis Zwiers and Slava Kharin at the Canadian Centre for Climate Modelling and Analysis for providing data from AOGCM control simulations. Valuable comments were received from the following: E. Barrow, M. Hulme, L. Mearns, J. Mitchell, T. Mitchell, C. Rosenzweig, R. Stouffer and T. Wigley. Financial assistance for this work was provided by the Ministry for Foreign Affairs of Finland, the Government of Canada and the Intergovernmental Panel for Climate Change Trust Fund.