Figure S1. Timeseries of relative TCO3 bias of CAMS reanalysis against FTIR data [(CAMS-obs)/obs] in % for the period January 2003 to April 2020. The stations on the y-axis are sorted by latitude from north (top) to south (bottom). Arctic stations (i.e. north of 60⁰N) are Eureka, Ny-Ålesund, Thule, Kiruna and Harestua.

A comparison of Arctic stratospheric ozone from the CAMS reanalysis with observations from limb scanning satellite instruments MLS (Livesey et al., 2018), Atmospheric Chemistry Experiment - Fourier Transform Spectrometer (ACE-FTS; Sheese et al., 2017) and MIPAS (Laeng et al., 2015) is shown in Figure S2. According to Hubert et al. (2016), MLS is among the most stable and accurate instruments for ozone above the UTLS and has a large coverage between 82° South and 82° North; MIPAS has also a good coverage but has some stability issues and tends to globally overestimate ozone concentrations. Both instruments were assimilated in the CAMS reanalysis and therefore should not be considered as fully independent observations. Ozone profile retrievals from ACE-FTS, which has a very sparse coverage, are an independent dataset not assimilated by CAMS. Availability of data from MLS starts in August 2004, ACE-FTS starts in February 2004, MIPAS ends in April 2012. In the middle upper stratosphere (10-30hPa, Figure S2a), the bias is generally smaller than ±5% for all instruments, except in 2003-2004 when no profile data were available for assimilation in the CAMS reanalysis leading to larger biases (Inness et al., 2019). As expected, the bias against MLS is very small (<1.5%), with a slight increase since 2018. This increase coincides with the change from using reprocessed MLS V4 data to using NRT MLS V4 data in the CAMS reanalysis in March 2018. A small negative bias against ACE-FTS and MIPAS is almost systematically present. In the middle lower stratosphere (30-70hPa, Figure S2b), the spread of the biases against the different instruments is slightly larger. Here, the bias against MLS is still low (<1.5%), the bias against MIPAS is negative (0-9%) and the bias against ACE-FTS is also negative (0-4%).