Polar lows (PLs) are intense mesoscale maritime cyclones. They produce hazardous winds in the Arctic, but their mesoscale structure is difficult to represent in atmospheric reanalyses. We evaluate the second-generation Copernicus Arctic Regional Reanalysis (CARRA2) for ten well-documented polar lows from 2002–2024 in the Nordic Seas, Barents Sea, Denmark Strait and Pacific Arctic. Reference events were compiled from published case studies and satellite imagery. PLs were tracked using 850-hPa vorticity in three-hourly 2.5-km analyses, and equivalent neutral 10-m winds were compared with QuikSCAT and ASCAT scatterometer observations at common valid locations. Retained tracks matched nine of the ten events, with mean position errors of 15.8–104.6 km (unweighted mean 55.1 km). The unmatched event of 7 March 2005 showed enhanced native-grid vorticity that spatial smoothing and persistence criteria removed. Across 59 scatterometer overpasses from nine events, median wind speeds were highly correlated (correlation 0.960) but 1.25 m s^-1 lower in CARRA2 on average. The difference rose to 2.18 m s^-1 for the 95th percentile and exceeded 8 m s^-1 in individual intense states. Scatterometer area fractions above 15 and 20 m s^-1 exceeded CARRA2 by 10.14 and 6.08 percentage points, respectively. CARRA2 therefore represents most selected polar lows and their typical wind variability, but its strong winds and their extent are lower than scatterometer estimates.