ALBERT

Description: To assess weather forecast and climate models, it is necessary to have observations going beyond standard variables, to reveal underlying processes. However, such observations are scarce over the central Arctic Ocean, with not even accurate observations of atmospheric vertical structure. Therefore, it is crucial to gather detailed atmospheric observations during icebreaker-based research expeditions to the central Arctic. In this study, we analyze extensive observations collected during the Arctic Ocean 2018 expedition on the icebreaker Oden. This took place in summer of 2018, with a focus on a month-long period from August 12 to September 14 while drifting with sea-ice near the North Pole. We evaluate 125 3-day forecasts from the European Centre for Medium-Range Weather Forecasts (ECMWF) Integrated Forecast System (IFS). Our analysis reveals a distinct systematic vertical error structure, with a boundary layer that is too warm with a kilometer-deep layer immediately above that is too cold. This error grows as the melt ends and the surface begins to freeze and also displays a diurnal variation. Additionally, the IFS overestimate low-troposphere clouds, with strong effects on the surface energy budget. Clear periods that occur in reality never materialize in the model. Initially errors are smaller, likely due to assimilation of the expeditions soundings, but grow rapidly and persist throughout the forecasts. We propose that the errors are due to parameterized sub-grid scale convection. Since the evaluated forecast model (Cy45r1) is almost identical to that used for ERA5, we expect these errors to be present also in the reanalysis.