Are simulations and reconstructions of past climate and its variabilityconsistent with each other? We assess the consistency of simulations andreconstructions for the climate of the last millennium under the paradigm ofa statistically indistinguishable ensemble. In this type of analysis, thenull hypothesis is that reconstructions and simulations are statisticallyindistinguishable and, therefore, are exchangeable with each other. Ensembleconsistency is assessed for Northern Hemisphere mean temperature, CentralEuropean mean temperature and for global temperature fields. Reconstructionsavailable for these regions serve as verification data for a set ofsimulations of the climate of the last millennium performed at the Max PlanckInstitute for Meteorology.Consistency is generally limited to some sub-domains and some sub-periods.Only the ensemble simulated and reconstructed annual Central European meantemperatures for the second half of the last millennium demonstratesunambiguous consistency. Furthermore, we cannot exclude consistency of anensemble of reconstructions of Northern Hemisphere temperature with thesimulation ensemble mean.If we treat simulations and reconstructions as equitable hypotheses aboutpast climate variability, the found general lack of their consistency weakensour confidence in inferences about past climate evolutions on the consideredspatial and temporal scales. That is, our available estimates of past climateevolutions are on an equal footing but, as shown here, inconsistent with eachother.
展开▼