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

2.1.2 Conventional observations. The simulations have been evaluated against hourly observations from weather stations provided by the Norwegian Meteorological Institute (MET Norway), and from drifting buoys provided by Canada’s Integrated Science Data Management (ISDM) centre. Drifting buoys have been deployed by different international programs, the largest being the Global Drifter Program (GDP), which is the result of an international collabo- ration under the World Meteorological Organization (WMO) and the United Nations Educa- tional, Scientific and Cultural Organization (UNESCO) umbrella. The GDP has been deploying surface Velocity Program Lagrangian drifters equipped with barometers that mea- sure mean sea level pressure (SLP) every hour [39]. The main advantages of using conventional observations as “truth data” are that they directly measure meteorological variables and they have high temporal resolution, which is essential to capture PL development. The observations from weather stations used to verify the simulations are SLP, 2-m temper- ature, 2-m dewpoint temperature, 2-m relative humidity, 10-m wind speed and direction, 10-m maximum wind gusts, and 1-h accumulated precipitation. The registered wind speed and direction are averaged over the last ten minutes before the observation time, and the maxi- mum wind gust is the maximum wind registered during the ten minutes before the observa- tion time. For drifting buoys, only SLP is available. Care should be taken when comparing the

observed 10-m maximum wind gusts with the simulation and ERA5 wind gusts since the latter two are instantaneous wind gusts that are output every hour. 2.1.3 ERA5. The reanalysis ERA5 is produced by the EMCWF using a 4D-Var data assimila- tion scheme and the IFS Cy41r2 [12]. ERA5 has a grid spacing of 31 km and 137 levels to 0.01 hPa, and it provides hourly data. It covers the period from 1978 to the present, and there is a pre- liminary version from 1950 to 1978 [40]. Among other data, ERA5 assimilates conventional observations from surface stations and drifting buoys [see Fig 4 of 12]. Some studies have found that ERA5 shows a good performance in the Arctic [41, 42]. For example, Graham et al. [41] found that, compared to other reanalyses, including ERA-I, ERA5 had the smallest biases and root mean square errors (RMSEs), and the highest correlation coefficients at capturing the tem- perature, wind speed and specific humidity in the Fram Strait. Nevertheless, some studies have found limitations of ERA5 over Arctic sea ice. Since ERA5 does not represent a snow layer on top of the sea ice, the conductive heat flux from the ocean to the atmosphere is overestimated. As a result, like other reanalyses, ERA5 sea-ice surface temperature shows a warm bias during clear- sky conditions in winter [43]. This is consistent with the large warm bias of ERA5 2-m tempera- ture over Arctic sea ice during the cold season compared to observations from drifting buoys [44]. 2.2 Data preparation We have prepared all the data from surface stations and drifting buoys available in the domain of

the simulations in order to have complete time series of the variables whenever possible. Regard- ing surface stations, only data with acceptable quality has been selected, and outliers have been discarded. Therefore, some of the time series were incomplete either because there was already missing data or because some observations were discarded due to their low quality. In the case of noisy variables (10-m wind, 10-m wind gusts, and 1-h accumulated precipitation), the time series with one or more missing data have been completely discarded. In the case of smooth or continu- ous variables (SLP, 2-m temperature, 2-m relative humidity and 2-m dewpoint temperature), the time series with more than three missing values have been discarded. For the time series with three or less missing values, these values have been computed doing a linear temporal interpola- tion using the closest previous and following available observations, including sub-hourly observa- tions. When the time between the previous or following observation and the missing observation was longer than one hour, the time series was discarded. Finally, since both wind speed and 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 5 / 28