Intercomparison of long-term ground-based measurements of total, tropospheric, and stratospheric ozone at Lauder, New Zealand

Abstract. Long-term, 21st century ground-based ozone measurements are crucial to study the recovery of stratospheric ozone as well as the trends of tropospheric ozone. This study is performed in the context of the LOTUS (Long-term Ozone Trends and Uncertainties in the Stratosphere) and TOAR-II (Tropospheric Ozone Assessment Report, phase II) initiatives. Within LOTUS, we want to know why different trends have been observed by different ground-based measurements at Lauder. In TOAR-II, intercomparison studies among the different ground-based datasets are needed to evaluate their quality and relevance for trend studies. To achieve these goals, we perform an intercomparison study of total column ozone and its vertical distribution among the ground-based measurement instruments available at the Lauder station from 2000 to 2022, which are a Fourier transform infrared (FTIR) spectrometer, a Dobson spectrophotometer, a UV2 (ultraviolet double monochromator), a microwave radiometer (MWR), ozonesondes, and a stratospheric lidar. Because only the latter two provide high-vertical-resolution profiles, the vertical ozone distribution is validated using partial columns, defined to provide independent information: one tropospheric column and three stratospheric columns. Because FTIR provides total columns and vertical information covering all partial columns as well as high temporal sampling, the intercomparisons (bias, scatter, and drift) are analyzed using FTIR as the reference. Very good agreement between the FTIR and Dobson (FTIR and UV2) total column ozone records is apparent in the high Pearson correlation of 0.97 (0.93), low biases of −3 % (−2 %), and 2 % (3 %) dispersions, which are within the respective systematic and random uncertainties. The small observed drifts 0.4 % (0.3 %) per decade are “non-significant” (or rather a low certainty in a 95 % confidence interval) and show good stability of the three ozone total column series at Lauder. In the troposphere we find a small bias of −1.9 % with the ozonesondes but a larger one (+10.7 %) with Umkehr, which can be explained by the low degrees of freedom for signal (0.5) of Umkehr in the troposphere. However, no significant drift is found among the three instruments in the troposphere, which proves their relevance for trend studies within TOAR-II. The negative bias observed in total columns is confirmed by negative biases in all stratospheric columns for all instruments with respect to FTIR (between −1.2 % and −6.8 %). This, confirmed by the total column biases, points to a 2 %–3 % underestimation of the infrared spectroscopic line intensities. Nevertheless, the dispersion between FTIR and all techniques is typically within 5 % for the stratospheric partial columns, in close agreement with the given random uncertainty budgets. We observe no significant drift in the stratosphere between ozonesondes and FTIR for all partial columns, with ozonesonde trends being less negative than in LOTUS (Godin-Beekmann et al., 2022, further referred to as the LOTUS22). The only significant drift in the lower-stratospheric columns is obtained between FTIR and Umkehr, as was already found in LOTUS22. Two significant positive drifts are observed in the middle stratosphere (2 % and 3 % per decade) with lidar and MWR, respectively, while two significant negative drifts are observed in the upper stratosphere (−3 % and −4 % per decade) with Umkehr and lidar, respectively. While remaining drifts are still present, our study explains roughly half of the differences in observed trends in LOTUS22 by the different sampling, vertical sensitivity, or time periods and gaps. In addition, the FTIR data in the present work have been improved since LOTUS22, reducing the differences in the upper-stratospheric and tropospheric trends. This shows the necessity for continuous review and improvement of the measurement and retrieval processes. This study also reflects the importance of super sites such as Lauder for cross-validating the long-term ozone measurements. Our study demonstrated that well-harmonized, optimized, well-characterized instruments that show very good agreement in terms of bias, dispersion, and correlation are capable of detecting trends that agree within their respective measurement uncertainties.