br Methods and results ACE is a Canadian satellite

Methods and results ACE is a Canadian satellite mission [7] that measures atmospheric DMPO receptor spectra in the limb geometry using the Sun as a light source (solar occultation). The ACE orbit (74° inclination to the equator, 650 km altitude) concentrates measurements at high latitudes but also provides coverage of mid-latitudes and tropics [Fig. 6 in ref. 7]. The ACE-FTS is a high resolution (0.02 cm−1) Fourier transform spectrometer that covers the 750–4400 cm−1 region; the current processing version used in our analysis is 3.5/3.6 [as summarized in ref. 7]; see http://www.ace.uwaterloo.ca/ for the v.3.5/3.6 HCl and N2O microwindow list [15]. The ACE-FTS HCl data product is provided on a 1 km altitude grid and has a vertical resolution of about 3 km over the 6–57 km altitude range at the poles and 7–63 km in the tropics. The VMRs for 60°S–60°N were interpolated onto a pressure grid spaced at pi = 1000 × 10−i/6 (hPa) corresponding to an altitude spacing of about 2.7 km as was used by GOZCARDS [10] and in the recent trend analysis by Stolarski et al. [14]. After discarding large negative and positive values, the VMRs were filtered at each pressure level by removing all values that were outside 2.5 standard deviations from the median. Quarterly averages were then computed for Dec–Feb (DJF), Mar–May (MAM), Jun–Aug (JJA), Sep–Nov (SON) at each pressure level to make a time series for MAM 2004 to SON 2017. The time series at each pressure level was de-seasonalized by computing the quarterly average for the entire time series and subtracting this value from the corresponding quarter to obtain a time series of anomalies. These anomalies were converted back to a de-seasonalized VMR time series at each pressure level by adding the 2004–2017 average. A time series of N2O VMRs was derived using the same method described above for HCl. The HCl time series are displayed (Fig. 1) for the pressure levels 0.68, 2.2, 10 and 46 hPa at the approximate altitudes of 51, 42, 31 and 21 km, respectively, using the US standard atmosphere [16] to estimate altitudes (these approximate altitudes are only provided for the convenience of the reader and are not used in the analysis). The linear trend lines are also plotted and for 0.68 hPa two lines are used (see discussion below). Also included in Fig. 1 is the total effective tropospheric chlorine [17] lagged by 4 years. Linear trend values are derived from the de-seasonalized HCl time series for 13 pressure levels from 68 hPa to 0.68 hPa (about 19–51 km) for 60°S–60°N and are plotted with one standard deviation error bars in Fig. 2. Linear regression was used with a constant, a linear term and a term for the N2O VMR [as in ref. 14] (i.e., VMRHCl(t) = a + bt + cVMRN2O(t) with constants a, b and c). The term with the N2O VMR accounts for some of the HCl variability due to dynamics and improves the determination of the linear trend value, b. The de-seasonalized N2O time series was first corrected for a steady N2O VMR increase of +0.28%/year [14]. The trend error bars were estimated using the procedure of Weatherhead et al. [14], [18] and include the effects of first order autocorrelation in the time series. The standard error in the trend estimate b is given by σNn−3/2[(1 + φ)/(1- φ)]1/2, in which σN is the standard deviation of the residuals (i.e., the difference between observed and modeled HCl VMRs), φ is the autocorrelation of the residuals and n is the number of years of data [18]. The 13 VMR time series for HCl and N2O, and HCl trend values with errors are provided as supplementary data. In order to assess the effect of atmospheric dynamics on the trend values, the same analysis was carried out for 3 additional latitude bins (30–60°N, 30°S–30°N, 30–60°S) and these trend values are also provided in the supplementary data. All the values for the highest 3 pressure levels in the upper stratosphere (0.68, 1, 1.5 hPa) are within 0.5% of the 60°S–60°N bin but, as expected, there were substantial differences in trends in the lower stratosphere [15] even when N2O was included in the trend analysis.