1Institute for Climate and Atmospheric Science, University of Leeds, Leeds, UK
2Department of Meteorology and the Bert Bolin Centre for Climate Research, Stockholm University, Stockholm, Sweden
3University of Colorado, Boulder, CO, USA and NOAA-ESRL, Boulder, CO, USA
4Max Planck Institute for Meteorology, Hamburg, Germany
5Swedish Meteorological and Hydrological Institute, Norrköping, Sweden
6Met Office, Exeter, UK
Abstract. Observations made during late summer in the Central Arctic Ocean, as part of the Arctic Summer Cloud Ocean Study (ASCOS), are used to evaluate cloud and vertical temperature structure in the Met Office Unified Model (MetUM). The observation period can be split into 5 regimes; the first two regimes had a large number of frontal systems, which were associated with deep cloud. During the remainder of the campaign a layer of low-level cloud occurred, typical of Central Arctic summer conditions, along with two periods of greatly reduced cloud cover. The short-range operational NWP forecasts could not accurately reproduce the observed variations in near-surface temperature. A major source of this error was found to be the temperature-dependant surface albedo parameterisation scheme. The model reproduced the low-level cloud layer, though it was too thin, too shallow, and in a boundary-layer that was too frequently well-mixed. The model was also unable to reproduce the observed periods of reduced cloud cover, which were associated with very low cloud condensation nuclei (CCN) concentrations (<1 cm−3). As with most global NWP models, the MetUM does not have a prognostic aerosol/cloud scheme but uses a constant CCN concentration of 100 cm−3 over all marine environments. It is therefore unable to represent the low CCN number concentrations and the rapid variations in concentration frequently observed in the Central Arctic during late summer. Experiments with a single-column model configuration of the MetUM show that reducing model CCN number concentrations to observed values reduces the amount of cloud, increases the near-surface stability, and improves the representation of both the surface radiation fluxes and the surface temperature. The model is shown to be sensitive to CCN only when number concentrations are less than 10–20 cm−3.