Ambiguity of large scale temperature reconstructions from artificial tree growth in millennial climate simulations

The ambiguity of temperature reconstructions is assessed using pseudo tree growth series in the virtual reality of two simulations of the climate of the last millennium. The simple, process-based Vaganov-Shashkin-Lite (VS-Lite) code calculates tree growth responses controlled by a limited number of climatic parameters. Growth limitation by different ambient climate conditions allows for possible nonlinearity and non-stationarity in the pseudo tree growth series. Statistical reconstructions of temperature are achieved from simulated tree growth for random selections of pseudo-proxy locations by simple local regression and composite plus scaling techniques to address additional ambiguities in paleoclimate reconstructions besides the known uncertainty and shortcomings of the reconstruction methods. A systematic empirical evaluation shows that the interrelations between simulated target and reconstructed temperatures undergo strong variations with possibly pronounced misrepresentations of temperatures. Thus (i) centennial scale inter-annual correlations can be very weak; (ii) the decadal range of reconstructed temperatures may be as large as the range of the temperature variations over the considered time-period; (iii) decadal variability is under-represented in the reconstructions. The misrepresentations are in part due to the temporally varying temperature-growth relations and to an apparent lack of decadal scale variability in the simulated pseudo-growth series compared to the local temperatures.