Available online at www.sciencedirect.com International Journal of Forecasting 27 (2011) 1004–1005 www.elsevier.com/locate/ijforecast Discussion Validation and forecasting accuracy in models of climate change: Postscript Robert Fildes ∗ , Nikolaos Kourentzes Lancaster Centrefor Forecasting, Lancaster University, UK We welcome the comments made by Keenlyside and McSharry and have no disagreements with either the substance or the tone. In this brief postscript we reflect on what we have learnt and what we believe is the overall message from the research. The paper was initially intended to inform forecasters of the state of forecasting as it is currently applied in the context of climate change. However, it developed a life of its own with the discovery of Smith et al.’s (2007) research, which provided forecasts which could reasonably be compared with the forecasts produced by statistical methods. However, the purpose of the paper remains methodological: to show how forecasting methods can be used to illuminate and improve on the forecasts produced by global circulation models. We believe that the results, tentative and limited though they are, have clear implications for the climate modelling community. • Before decadal forecasts are adopted by policy makers, the forecasts should be shown to be more effective than those of alternative statistical methods. ∗ Corresponding author. E-mail address: R.Fildes@Lancaster.ac.uk (R. Fildes). – This requires an extension of the simulations and supporting methodology beyond that proposed in the forthcoming IPCC assessment report. • If the combination of GCM based forecasts and statistical forecasts is shown to lead to an improved accuracy, this approach should be the basis of policy. – This may have implications for the effectiveness of relying on policies for controlling CO 2 emissions; see Pielke et al. (2009). • Where the forecasts from a GCM are enhanced by the inclusion of a statistical model, this is evidence that the GCM is not capturing all of the important elements of the climate system, and therefore that the model should be reconfigured. In statistical terms, the model is mis-specified. While part of the climate modelling community is becoming more open to a consideration of the value of statistical analyses, some of the reviews of this paper demonstrated a degree of hostility and rejection of the fundamental ideas of forecasting which suggests that much progress still has to be made. Climate change is a major threat to us all, and this requires the IPCC community of climate modellers to be more open to developments and methods from other fields of research. Equally, it is a field to which researchers 0169-2070/$ - see front matter c ⃝ 2011 International Institute of Forecasters. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.ijforecast.2011.07.001