Evaluation of NWP models results in the Ross Sea (Antarctica) and Bay of Bothnia (Baltic Sea) based on small unmanned aerial system (sUAS) measurements.

Numerical weather prediction (NWP) models struggle with the representation of atmospheric boundary layer (ABL) over sea ice. Particularly difficult to resolve are near-surface temperature, wind speed, and humidity, including diurnal and vertical changes of those properties. Many of the complex processes happening at the interface of sea ice and atmosphere, i.e. vertical fluxes, turbulence, atmosphere - surface coupling are poorly parameterized or not represented in the models at all, largely due to a limited number of observations from the sea ice-covered regions. There is a growing need to determine which aspects of the models configuration are responsible for the biggest errors and which parameterizations need to be improved. In order to this, we have to combine the analysis of the models results with observations. In the past, the majority of observations in the polar regions came from weather stations, which provided us with point measurements from few meters above the ground, not sufficient for the comprehensive evaluation of models performance. Nowadays, the usage of unmanned aerial systems (UAS) opened up a range of new possibilities, providing us with 3D observations of the ABL over multiple areas in both hemispheres.

In the following study, we use data collected during HAOS (Hailuoto Atmospheric Observations over Sea ice) (Wenta et. al., 2021) and HAOS2 (22 - 27 March 2021) campaigns carried out near Hailuoto island (Bothnia Bay, Baltic Sea) and SUMO (Small Unmanned Meteorological Observer) UAS measurements from the Ross Sea (Cassano, 2020) to evaluate regional NWP models predictions. In the case of Baltic Sea campaigns the following models forecasts are verified: AROME (Applications of Research to Operations at Mesoscale) - Arctic, HIRLAM (High-Resolution Limited Area Model), and WRF (Weather Research and Forecasting), whereas for the Ross Sea we use Antarctic Mesoscale Prediction System (polar WRF) data. Models ability to simulate vertical changes of humidity, temperature, wind speed, and air pressure is analyzed during various states of the atmosphere including clear sky and cloudy conditions, anticyclone passing, and throughout the periods of very cold and relatively mild temperatures. Our goal is to assess which ABL state, stable, neutral, or unstable, poses the biggest challenge for the models and why? It is found that generally, models struggle with an accurate representation of surface-based temperature inversions, often overestimating its strength, and vertical variations of humidity during the day, as they tend to overestimate the amount of water vapor/liquid water in the air. Furthermore, the low-level jets (LLJ) are also misrepresented in the models results, together with temporal changes of wind speed. In many cases the properties of the upper layers of the atmosphere are better resolved, than the lower layers, adjacent to the surface, thus indicating (as many other studies) that surface-atmosphere interactions in the models are not represented correctly. In order to determine whether different models configurations might improve models ability to predict certain ABL conditions we focus on the WRF model, a connecting factor between Bothnia Bay and Ross Sea analysis, and using WRF Single-Column-Model version evaluate how different planetary boundary layer parameterizations (MYJ, YSU, MYNN, QNSE), vertical resolution, and more accurate representation of surface conditions affect the results. Experiments with WRF SCM are also used to determine the possible reasons behind model biases. Preliminary results show that accurate representation of sea ice conditions, including thickness, surface temperature, albedo, and snow coverage is crucial for increasing the quality of NWP model forecasts. We emphasize the importance of further development of parametrizations focusing on the processes at the sea ice-atmosphere interface.