Assessment of simulations of a polar low with the Canadian Regional Climate Model

direction are needed to verify the simulations, only the data of stations that provide both wind speed and direction have been retained for the verification of the wind field. In the case of drifting buoys, no time interpolation of the missing data has been done. The simulated and ERA5 atmospheric fields have been spatially interpolated from the model grid to the observation points using either bilinear–for noisy variables–or bicubic inter- polation–for smooth variables. A simple height correction has been applied to the simulated and the ERA5 temperature and dew point temperature to account for the difference in height between the real topography and the topography of the model. The lapse rate of the simula- tions and ERA5 at the lowest levels of the atmosphere has been used for the height correction of their respective temperature fields. 2.3 Verification First, the track, size and intensity of the PL captured by the simulations and ERA5 have been compared to that of the observed PL. The track of the observed PL has been manually obtained using IR radiance satellite images from the Moderate Resolution Imaging Spectroradiometer (MODIS), the Advanced Very High Resolution Radiometer (AVHRR/3) and the Visible Infrared Imaging Radiometer Suite (VIIRS). The coordinates of the centre of the observed PL have been estimated at each hour from the genesis until the dissipation of the PL. The track that has been initially obtained using the satellite images has been improved by ensuring that the track is consis-

tent with the conventional observations of SLP and 10-m wind. The tracks of the PL in the simu- lations and ERA5 have been manually obtained using the SLP field. The criteria to identify the beginning of the PL is the presence of at least three SLP closed contours in a map showing the SLP isobars every 1 hPa. The size of the PL has been estimated in all the datasets by measuring the diameter of the cloud signature during the mature stage of the PL. The intensity of the PL in the simulations and ERA5 is given by the SLP minimum at its centre. In the case of observations, the SLP minimum corresponds to the SLP observation from the surface station that is the closest to the centre of the PL. Since some stations are too far from the centre of the PL, only the observa- tions from stations within a distance of 25 km from its centre have been considered. Second, all the simulations and ERA5 have been verified against observations from surface stations and drifting buoys affected by the PL. Therefore, we have only used observations obtained within a distance of 300 km– which approximately corresponds to the radius of the cloud signature of the PL at its mature stage –from the centre of the observed PL. The total number of observations used are 352 for SLP, 860 for 2-m temperature, 820 for 2-m dewpoint temperature, 483 for 2-m relative humidity, 534 for 10-m wind, 318 for 10-m wind gusts, and 448 for 1-h accumulated precipitation. From the genesis of the PL until 25 March at 11:00 UTC or 12:00 UTC, depending on the variable, the number of observations is no more than

10, or there are no observations at all. The number of observations notably increases when the PL gets closer to the Norwegian coast (Fig 1). Therefore, the results are mainly representative of the mature and dissipation stages of the PL. The statistics computed to measure the perfor- mance of the simulations are the mean error (ME), the root mean square error (RMSE), and the Spearman correlation coefficient (r) [45]. Since the wind is a vector, the root mean square wind-vector-difference error has been computed [RMSE-WVD; e.g., 46]. The correlation coef- ficient has only been computed when at least three observations were available. 3. Results and discussion 3.1 Description of the life cycle of the PL Northerly winds on the cold side of a synoptic-scale low located over the Barents Sea caused a MCAO in Fram Strait at the end of March 2019 (Fig 2A). Cold northerly winds to the west of PLOS ONE Assessment of simulations of a polar low with the canadian regional climate model PLOS ONE | https://doi.org/10.1371/journal.pone.0292250 October 5, 2023 6 / 28