Polar WRF, ERA-Interim, and GFS can reproduce the general features of the observed mean atmospheric profiles, such as low-level temperature inversion, low-level jet (LLJ) and specific humidity inversion. The Polar WRF significantly improves the mean LLJ, with a lower and stronger jet and a larger turning angle than the forcing, which is likely related to the lower values of the boundary layer diffusion in WRF than in the global models such as ECMWF and GFS. The Polar WRF simulated relative humidity closely resembles the forcing datasets while having large biases compared to observations. This suggests that the performance of Polar WRF and its forecasts in this region are limited by the quality of the forcing dataset and that the assimilation of more and better-calibrated observations, such as humidity data, is critical for their improvement.
We investigate the potential of assimilating the SIZRS dropsonde dataset in improving the weather forecast over the Beaufort Sea. A simple local nudging approach is adopted. Along SIZRS flight cross sections, a set of Polar WRF simulations are performed with varying number of variables and dropsonde profiles assimilated. Different model physics are tested to examine the sensitivity of different aspects of model physics, such as boundary layer schemes, cloud microphysics, and radiation parameterization, to data assimilation. The comparison of the Polar WRF runs with assimilation and the runs without assimilation demonstrates the importance of SIZRS dropsonde data to the improvement of atmospheric analysis and reanalysis such as GFS and ERA-Interim, and consequently to the improvement of weather forecast in this region.