Comparison of observed borehole temperatures in Antarctica with simulations using a forward model driven by climate model outputs covering the past millennium

Abstract

Abstract. The reconstructed surface-temperature time series from boreholes in Antarctica have significantly contributed to our understanding of multidecadal and centennial temperature changes and thus provide a good way to evaluate the ability of climate models to reproduce low-frequency climate variability. However, up to now, there has not been any systematic model–data comparison based on temperature from boreholes at a regional or local scale in Antarctica. Here, we discuss two different ways to perform such a comparison using borehole measurements and the corresponding reconstructions of surface temperature at the West Antarctic Ice Sheet (WAIS) Divide, Larissa, Mill Island, and Styx Glacier in Antarctica. The standard approach is to compare the surface temperature simulated by the climate model at the grid cell closest to each site with the reconstructions in the time domain derived from the borehole temperature observations. Although some characteristics of the reconstructions, for instance the nonuniform smoothing, limit to some extent the model–data comparison, several robust features can be evaluated. In addition, a more direct model–data comparison based on the temperature measured in the boreholes is conducted using a forward model that simulates explicitly the subsurface temperature profiles when driven with climate model outputs. This comparison in the depth domain is not only generally consistent with observations made in the time domain but also provides information that cannot easily be inferred from the comparison in the time domain. The major results from these comparisons are used to derive metrics that can be applied for future model–data comparison. We also describe the spatial representativity of the sites chosen for the metrics. The long-term cooling trend in West Antarctica from 1000 to 1600 CE (−1.0 ∘C) is generally reproduced by the models but often with a weaker amplitude. The 19th century cooling in the Antarctic Peninsula (−0.94 ∘C) is not reproduced by any of the models, which tend to show warming instead. The trend over the last 50 years is generally well reproduced in West Antarctica and at Larissa (Antarctic Peninsula) but overestimated at other sites. The wide range of simulated trends indicates the importance of internal variability in the observed trends and shows the value of model–data comparison to investigate the response to forcings.