Item

Abstract

Reanalyses are especially useful in providing climatic information in remote and inhospitable regions, such as the Antarctic continent. While there are relatively few direct measurements from Antarctica to assess the accuracy of reanalyses in simulating precipitation at sub-daily frequencies, present and past weather reports from meteorological stations allow a quantitative assessment of the skill of reanalyses in getting the correct timing of precipitation. This is an important consideration when it comes to utilising reanalyses to examine extreme precipitation events or in interpreting climate signals in ice cores. Here, we use a forecast verification methodology based on non-probabilistic forecasts of discrete predictands – that is, simply, there is precipitation or there isn’t - to examine the capability of ERA5 to correctly reproduce the timing of Antarctic precipitation at a 6-hourly temporal resolution. The assessment is undertaken by comparing reanalysis output to 20-years of present and past weather reports at six Antarctic meteorological stations. Using three different ‘definitions’ of precipitation in ERA5 (>0.0 mm, >0.1 mm, >1.0 mm) across a 6-hourly period, we examine how these impact on scalar attributes of the forecast skill, such as accuracy, bias, reliability, discrimination, and the Peirce Skill Score. Unsurprisingly, the precipitation definition that provides the best skill is different for an Antarctic Plateau station, where small amounts of precipitation predominantly fall as diamond dust, and coastal stations, which have greater precipitation associated with frontal systems. We also investigate whether there is any seasonal variability in the skill of ERA5 to reproduce sub-daily frequency Antarctic precipitation.