Abstract
While gridded seasonal pressure reconstructions poleward of 60°S extending back to 1905 have been recently completed, their skill has not been assessed prior to 1958. To provide a more thorough evaluation of the skill and performance in the early 20th century, these reconstructions are compared to other gridded datasets, historical data from early Antarctic expeditions, ship records, and temporary bases. Overall, the comparison confirms that the reconstruction uncertainty of 2–4 hPa (evaluated after 1979) over the Southern Ocean is a valid estimate of the reconstruction error in the early 20th century. Over the interior and near the coast of Antarctica, direct comparisons with historical data are challenged by elevation‐based reductions to sea level pressure. In a few cases, a simple linear adjustment of the reconstruction to sea level matches the historical data well, but in other cases, the differences remain greater than 10 hPa. Despite these large errors, comparisons with continuous multi‐season observations demonstrate that aspects of the interannual variability are often still captured, suggesting that the reconstructions have skill representing variations on this timescale, even if it is difficult to determine how well they capture the mean pressure at these higher elevations. Additional comparisons with various 20th‐century reanalysis products demonstrate the value of assimilating the historical observations in these datasets, which acts to substantially reduce the reanalysis ensemble spread, and bring the reanalysis ensemble mean within the reconstruction and observational uncertainty.