The enigma of multidecadal to centennial global vs. local temperature variability in models and proxies

Climate variability, resulting from natural radiative forcing and interactions within the climate system, is a major source of uncertainty for regional climate projections. Constraining the amplitude of these natural variations is fundamental to assess the range of plausible future scenarios. As the instrumental record is limited to the last two centuries, information about climate variations on multi-decadal to millennial timescales relies on the analysis of climate proxy records and climate model simulations. However, current results from systematic model-proxy comparisons of natural variability seem contradictory. Several studies suggest that simulated local temperature variability is consistently smaller than proxy-based reconstructions and conclude that climate models might have major deficiencies. Other studies find agreement in global temperature variability across timescales and argue that current models can faithfully simulate climate variability. 

Here, we review the evidence on the strength of natural temperature variability during recent millennia. We identify systematic biases in the reconstructions that may contribute to the model-proxy discrepancy but are likely not sufficient to reach consistency. Instead, we propose that the seemingly contradictory  findings on the (dis)agreement between proxies and simulations can be reconciled assuming that regional climate variations persist on longer time scales than currently simulated by climate models. The combined evidence argues for deficiencies in the simulation of internal variability but a faithful response of climate models to natural radiative forcing. We propose a strategy to test our hypothesis and discuss the implications for future climate projections.