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

lifetimes ranging from three to 36 hours [3]. They develop over the open water near the snow- covered landmasses or the sea-ice edge during marine cold air outbreaks (MCAOs). PLs are associated with severe weather conditions, including gale-force winds and heavy snowfall. These conditions can lead to large waves [e.g., 4], low visibility, snow avalanches, and icing on infrastructures. Therefore, PLs pose a threat to coastal populations, infrastructures, transport, and economic activities, and in some cases they have led to casualties [e.g., 5]. Producing accu- rate weather forecasts of PLs is thus critical to provide communities with enough time to prepare. Weather forecasting in the polar regions remains a challenge since conventional observa- tions are sparse, with weather stations being mainly concentrated along the coast [6], and data assimilation often fails to optimally use the available observational datasets [7]. The small tem- poral and spatial scales–horizontal and vertical scales of 100 km and 1 km, respectively –of PLs makes them particularly hard to forecast and to reanalyse [8]. Global reanalyses have low reso- lution (> 30 km of grid mesh), so they often fail to capture observed PLs. For instance, the reanalysis of the European Centre for Medium-Range Weather Forecasts (ECMWF) known as ERA-Interim [ERA-I, 9], which has a grid mesh of 0.75˚, fails to capture many PLs [10, 11]. The fifth-generation ECMWF reanalyses ERA5 [12], which has a grid mesh of 31 km and hourly outputs, captures more PLs than its predecessor [13]. Regional reanalyses such as the

Arctic System Reanalysis [ASR, 14] are likely to be more adequate to represent PLs than global reanalyses given their higher resolution, and the fact that they are adapted to a particular region. For example, the first version of the ASR, which has a grid mesh of 30 km, captures more PLs than ERA-I [15]. Limited-area high-resolution atmospheric models are also a useful tool to study PLs since they can represent more PLs compared to the coarser reanalysis used as initial and boundary conditions [e.g., 10]. PL forecasting has been improved recently thanks to the development of high-resolution, non-hydrostatic atmospheric models. Compared to large-scale models, convection-permitting models (CPMs) provide a better representation of convection as well as surface and near-sur- face processes [16], which play an important role in the development of PLs. Indeed, Stoll et al. [17] found that, compared to the ECMWF global model HRES based on the Integrated Fore- cast System (IFS) cycle 32r3 [18], which has a grid mesh of 25 km, the regional model ARO- ME-Arctic [19], which has a grid mesh of 2.5 km, represented better the small-scale features associated with a PL such as individual convective clouds. The emergence of high-resolution atmospheric models comes with its challenges. The increased resolution of the models requires that the model parameterisations be adapted to the resolution of the CPMs [16, 20]. In the polar regions, the parameterisation of surface fluxes needs to be optimised [21]. Furthermore, to make correct forecasts, atmospheric models need

to be initialised with good observed conditions. Initial conditions uncertainties affecting the prediction of small-scale weather systems are mainly associated with convective and mesoscale instabilities [22]. The initial conditions of moisture at the mesoscale are especially significant for PL forecasting [23]. The initialisation time also seems to have an impact on the representa- tion of PLs, as shown by case studies of the PL developed on 3 March 2008 [24, 25]. McInnes et al. [24] found that the simulations with the MetUM using a grid mesh of 4 km showed better performance when the simulations were initialised at around 42 hours before the PL formed compared to the simulations initialised 24 hours later. The authors argued that this could indi- cate that initialising the simulations at an earlier stage may be necessary to reproduce the syn- optic-scale atmospheric conditions leading to the PL development. Nevertheless, Wagner et al. [25] obtained opposite results using the Polar Weather Research and Forecasting (WRF) model with a grid mesh of 2 km. In effect, the authors found that the simulations that per- formed better were those whose initialisation time was closer to the genesis time of the PL. 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 2 / 28 are available from NASA (https://ladsweb.modaps. eosdis.nasa.gov/). The observations from surface stations can been downloaded using MET Norway Frost API (https://frost.met.no/index.html), and the

observation from drifting buoys can be requested to Canada’s ISDM centre (https://www.dfo-mpo. gc.ca/science/data-donnees/drib-bder/index-eng. html). The ERA5 global reanalysis from ECMWF is available at https://www.ecmwf.int/en/forecasts/ datasets/reanalysis-datasets/era5. The Global Self- consistent, Hierarchical, High-resolution Geography Database (GSHHG), available at https:// www.ngdc.noaa.gov/mgg/shorelines/, has been used to represent the coastlines. The divergent colourmap used in Fig 2 is provided by the Texas Advanced Computing Center at https://sciviscolor. org/. The simulation output and the coordinates of the manually obtained tracks are available at Borealis, the Canadian Dataverse Repository (doi: 10.5683/SP3/6E3ITE). Funding: This work was supported by the Discovery Grant program of the Natural Sciences and Engineering Research Council of Canada (NSERC) under Grant 707337, by the project “Marine Environmental Observation, Prediction and Response” (MEOPAR) of the Networks of Centres of Excellence (NCE) of Canada, by the UQAM’s Faculty of Sciences under the programme “faculty financial support”, and by the excellence scholarship of the Trottier Family Foundation. The operation of the supercomputer Beluga is funded by the Canada Foundation for Innovation (CFI), Ministère de l’E´conomie et de l’Innovation du Que´bec (MEI) and les Fonds de recherche du Que´bec (FRQ). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. There was no additional external funding received for this study.

Competing interests: The authors have declared that no competing interests exist.