Sensitivity of climate change detection and attribution to the characterization of internal climate variability

The Intergovernmental Panel on Climate Change (IPCC) “very likely” statement that anthropogenic emissions are affecting climate is based on a statistical detection and attribution methodology that strongly depends on the characterization of internal climate variability. In this paper, we test the robustness of this statement in the case of global mean surface air temperature, under different representations of such variability. The contributions of the different natural and anthropogenic forcings to the global mean surface air temperature response are computed using a box diffusion model. Representations of internal climate variability are explored using simple stochastic models that nevertheless span a representative range of plausible temporal autocorrelation structures, including the short-memory first-order autoregressive (AR(1)) process and the long-memory fractionally differencing (FD) process. We find that, independently of the representation chosen, the greenhouse gas signal remains statistically significant under the detection model employed in this paper. Our results support the robustness of the IPCC detection and attribution statement for global mean temperature change under different characterizations of internal variability, but also suggest that a wider variety of robustness tests, other than simple comparisons of residual variance, should be performed when dealing with other climate variables and/or different spatial scales.