Reconstruction of Quaternary temperature fields by dynamically consistent smoothing

Conventional methods of palaeoclimate reconstruction provide estimates of climatic parameters using proxy data which have originated from individual sites. These reconstructions yield information on the local environment but only limited information on spatial scales that are required for model-data intercomparisons. We present here a new approach that connects these different scales by an upscaling of the local palaeoinformation together with a dynamically consistent spatial smoothing. A probabilistic data-based method for local reconstructions is combined with a dynamic constraint on the reconstructed climate parameter which stabilises the reconstruction on the target scale. The variational analysis leads to climatological fields being optimised with respect to the proxy data and to the prescribed dynamics in a statistically consistent way. This method allows a probabilistic approach of quality control of the palaeodata in terms of their spatial consistency and homogeneity and for an estimation of reconstruction errors. The method was applied to palaeobotanical data to reconstruct near-surface temperature fields constrained by simple linear dynamics. An approximate approach was used to estimate the magnitude of reconstruction errors in terms of standard deviations. Reconstructed January and July mean temperature of the early Eemian (∼ 125,000 years bp) have errors with a median value of about 1.8°C in January and about 1.1°C in July. Reconstructions across Europe show positive temperature anomalies for Scandinavia and near the East coast of the Baltic Sea. In contrast, early Eemian temperatures were apparently quite similar to those found today in Central Europe, as no drastic differences were reconstructed between the Eemian and modern (1961–1990) climate. This implies somewhat stronger temperature gradients in the Eemian than are observed today.