ECC ozonesondes have been used in OHP since 1991 to measure ozone vertical profiles on a weekly basis (generally around 09:00 UT). ECC generates an electrical current through the reaction of ozone in a KI solution. The ozone partial pressure PO3 (measured in mPa) is then obtained from the electrochemical current IM (measured in µA), the background current IB measured in the preparation laboratory with an ozone removal filter after the sonde was exposed to ozone, the gas volume flow rate ΦP (in cm³ s⁻¹) of the air sampling pump, the pump temperature TP (in K) and the total efficiency of the ozone sensor ηT . 1PO3=4.306×10-2TPηPηAηCΦP(I-IB) The total efficiency ηT consists of pump flow efficiency ηP, absorption efficiency ηA and conversion efficiency of the absorbed ozone ηC. The median value of the relative uncertainty in the ozone concentration measured by the ECC is on the order of 6 %–7 % in the stratosphere . Total column ozone (TCO) measurements by the UV-visible SAOZ (Système d'Analyse par Observation Zénithale) and the Dobson spectrophotometer are available at OHP. The so-called normalization factor is calculated as the ratio of the spectrophotometer TCO and the ECC TCO. Historically, ozonesonde profiles have often been normalized or scaled to an independent measurement of the TCO. Normalization of ECC sonde profiles is no longer recommended, however the normalization factor provides a useful indicator for the quality of ozonesonde profile data and the consistency in ozonesonde data time series. We use the dataset of the Dobson spectrophotometer from 1991 to 2004 and of the SAOZ from 2004 to 2023 to calculate normalization factor. The residual ozone above the 10 hPa pressure altitude or above the balloon burst altitude is derived from the satellite monthly mean climatology. The normalization factor is not calculated for balloon burst altitude lower than the 25 hPa pressure level or when the SAOZ measurement is not available. The corresponding sounding are nevertheless included in the OHP data set. This corresponds to 9.5 % of the 1524 OHP soundings from 1991 to 2023. Only ozonesonde profiles with a normalization factor below 1.3 are considered here (i.e. only 3 soundings out of the 1524 are excluded of the data base). The fraction of soundings with normalization factor higher than 1.2 is 1.3 %. Normalization factor are always larger than 0.83 and no filtering is applied for low values of the normalization factor.

The details of instrument changes and the homogenization are described in . Briefly, the main changes considered in this work are as the following: a major change in the sounding procedure occurred in March 1997 when the Science Pump Corporation (SPC) ozonesonde was replaced by the Environmental Science Corporation (EnSci) ozonesonde. In July 2007, the radiosonde type switched from Vaisala RS80 to Modem M10, together with a change of the position of pump temperature inserted in the pump hole instead of being taped to the pump. The TMAX electronical interface between the radiosonde and the ECC changed in 2002. As far as we know from available written information from this time period no other significant changes of the sounding preparation occurred during the time period March 1997 and July 2007. Correction of pump temperature has been implemented in to account for the changes in the position of the ECC pump thermistor in June 2007. However the homogenized ECC pump temperature TP remains inhomogeneous for the the period 1991–June 2007 and this correction must be carefully assessed. In this work, the ECC pump temperature correction is assessed by comparison of homogenized ECC ozone concentrations with other ozone measurements obtained at OHP in as close coincidence as possible.

At OHP, regular nighttime measurements (2–4 per week) by the Differential Absorption Lidar (DIAL) ozone lidar have been made since 1985 . The lidar is optimized for stratospheric ozone profiling between 10 and 50 km a.s.l. using an absorbed laser wavelength at 308 nm and a reference laser wavelength at 355 nm. Ozone profiles are obtained during the night in clear sky conditions and over a time range of 4 h. The best accuracy (5 %) is generally obtained in the 15–40 km altitude range where ozone concentrations are large enough to minimize lidar systematic errors and signal-to-noise ratio (SNR) is maximum . The coincidence is determined with a temporal criterion of ±12 h of the ECC sonde launch time for the comparison.

The Microwave Limb Sounder (MLS) on-board the Aura satellite, launched on 15 July 2004, measures vertical profiles of the abundances of key atmospheric species, including ozone. We use Aura MLS level 2 version 5.0x ozone volume-mixing ratio vertical profile observations for a comparison with the ECC ozonesonde time series. The accuracy from 68 to 4.6 hPa is 5 %–8 % . The coincidence with the ECC sonde launch is determined with spatial differences less than ±2.5° latitude and ±8° longitude, and temporal differences less than ±12 h. Using a smaller domain would significantly decrease the number of coincidences. Data screening based on the recommended values of the status flag, quality field and convergence field have been applied to the MLS data selected for the analysis of the stratospheric ozone profiles above OHP.

ECC pumping time for 100 mL of air is expected to change especially when switching ECC type or motor power supply. Its time evolution is shown in Fig. . There is indeed lower pumping time values from 2002–2007 in the range 27–29 s. However these values are within the 25–35 s recommended values in . The time evolution shown in Fig.  is taken into account in the ozone calculation using Eq. () and OHP metadata records are very reliable for this parameter. Therefore we do not expect the low ozone values observed at 25 km to be caused by incorrect values of the ECC flow rate used in the data processing.

reported that EnSci ECC total ozone column drop occurring at selected sites after 2013 could be related to ozonesonde production changes as well as station-specific factors due to ozonesonde preparation. The ECC serial numbers associated with sounding exhibiting TCO decreases larger than 3 %–5 % were above 20 000. The EnSci serial numbers are between 6200 and 12 800 for the OHP soundings made between January 2002 and June 2007. We therefore cannot identify a clear connection with the sonde production quality discussed in that paper.

ECC internal pump temperature, TP, correction is another key factor in calculating ozone partial pressure since the correction of ECC pump temperature applied in can be as large as 10 K to account for wrong thermistor position. Figure  shows its time evolution with three periods showing changes in the range of the homogenized ECC pump temperature (in this work the measured ECC pump temperature always refers to the homogenized ECC temperature retrieved in ). Overall, values are in the range 290–300 K after June 2007 while they are in the range 300–310 K before January 2002. The latter could be explained by a major change of the ECC installation in the Styrofoam box when switching from Vaisala RS80 to Modem M10 (new electronic interface and new batteries). However, the homogenized pump temperature frequently dropped within the range 280–290 K between January 2002 and July 2007 without a clear explanation for such a negative drop. The 8 K temperature positive correction in the stratosphere due to the position of the thermistor being taped instead of being inserted in the pump hole before July 2007 has been applied to all the sondes from January 1991 to June 2007 in . The unusually low values of the ECC pump temperature in the January 2002–June 2007 period corresponds to that of the ozone concentration observed in Sect. 3. An additional correction of the ECC pump temperature measurement for this time period might be needed. Other factors, namely, ozone concentration change in the stratosphere and the statistical distribution of the pump flow rate or the ECC current must be taken into account before we apply such correction.

First, the mean vertical distribution of the homogenized ECC temperature and of the ECC current for three periods (1991–2001, 2002–2007, 2008–2019) are shown in Fig.  to emphasize that the ECC temperature drop in 2002–2007 increases with altitude in the stratosphere as does the difference between internal and external ECC pump temperature shown in the analysis of . The differences of ECC current between 2002–2007 and 1991–2001 does not show a clear altitude dependency with no evidence of lower current difference at 30 km than at 20 km in 2002–2007 as we might expect if the ECC solution temperature is so cold that it becomes frozen. In fact the relative differences of ECC current between 2002–2007 and 1991–2001 are -5.4 % at 20 km, -5.7 % at 25 km and -6.8 % at 30 km. Wrong values of the recorded ECC pump temperature in 2002–2007 would thus remain as the possible primary cause of the negative ECC ozone concentration bias in the stratosphere.

Second, to check the magnitude of the ECC temperature drop for the period January 2002–June 2007 necessary to explain the ECC ozone concentration decrease at 25 km, the homogenized ECC temperature drop TP has been compared with a ECC temperature drop TPc calculated by using measured ECC current, pump flow and ozone concentration based on the Eq. (1). The ozone concentration measured by the stratospheric lidar at OHP is employed assuming that it is the best proxy to account for any long term ozone change in the stratosphere. We consider three periods: March 1997–December 2001 as period 1, January 2002–June 2007 as period 2 and July 2007–December 2023 as period 3 in this analysis. We did not use the data before March 1997 to avoid considering a change of the efficiency of the ozone sensor ηT when changing the type of ozonesonde in March 1997. The distributions and medians of the ECC temperature TP, the ECC current, the pump flow and the ozone concentration of lidar are shown in Fig.  and Table  at 25 km. The differences of ozone concentration observed by lidar among three periods are small (0 %–1 %).

The calculated percentage difference of homogenized ECC temperature derived from observed differences of ECC current and pump flow rate can be expressed as: 2ΔTPcTPc=ΔO3O3-Δ(IM-IB)(IM-IB)-Δ(1/ΦP)(1/ΦP) Taken period 1 as a reference percentage differences of all the parameters used in Eq. () are shown in Table  using measured values of period 2 and 3, i.e. when ozone underestimate change from -3.4 % to 0.1 %. Measured and calculated difference of the mean pump temperature is not large for period 3 (< 4 K) while the difference can be as large as -10 K for period 2.

Before we present the selected correction procedure of the ECC pump temperature it is important to review the known changes during this time period. First, there was a change of the electronics interface between the ECC and the radiosonde in January 2002 switching from TMAX-HMOS to TMAX-C. An update of the software to read the binary files was made to handle the dataframe from the TMAX-C interface but documentation about this step has been lost. We cannot exclude a bug in the implementation of this software. The use of the TMAX-C interface was discontinued in July 2007 when Modem radiosondes replaced the VAISALA radiosondes.

Second, we make the hypothesis that the thermistor was taped to the pump during this time period 2002–2007 in but this information has not been precisely recorded in the metadata and there is a large uncertainty about the position of the thermistor position in the styrofoam box. The pump temperature correction applied during the homogenization process performed in might not be correct.

Third, the NOAA STRATO version 8.66 has been implemented at OHP to produce level 1 files between February 2004 and June 2007 with two consequences: (i) a different processing software has been applied for the ozone retrieval without a recording of the ECC current in the level-1 files, (ii) the binary files with the ECC current have been lost. In order to include cell current in the data files, performed a reverse calculation of cell current since the necessary variables to back-calculate the cell current from the ozone partial pressure were available in the data file. We are confident that this process produced very small uncertainties in the ECC current retrieval. Use of the STRATO software has been discontinued in July 2007 when switching to the Modem data acquisition system.

The vertical profile of the pump temperature difference between the period 2002–2007 and the period 1991–2001 can be approximated by a linear evolution of -0.4 K km⁻¹ (Fig. ). We do not expect a perfectly linear altitude-dependent difference since atmospheric temperature profiles cannot be strictly identical for the two periods. The jump at 10 km in the pump temperature difference indeed corresponds to changes in the tropopause altitude. Several ECC pump temperature linear altitude-dependent corrections were therefore tested for the period January 2002–June 2007 in this section. Since according to Sect. 4.1, bias of the ECC pump temperature can be as large as -10 K in period 2 at 25 km, the additional correction applied to the homogenized ECC temperature TPcor is calculated as: 3TPcor(z)=TP(z)+αz where z is altitude (measured in km) and α the increasing correction coefficient in K km⁻¹. Since we do not have a clear explanation for the pump temperature drop off in 2002–2007 in the stratosphere, we decided to test three corrections: α=0.17 K km⁻¹ (ΔTP=0 K at 0 km and ΔTP=5 K at 30 km) called version 2, α=0.33 K km⁻¹ (ΔTP=0 K at 0 km and ΔTP=10 K at 30 km) called version 3 and α = 0.50 K km⁻¹ (ΔTP=0 K at 0 km and ΔTP=15 K at 30 km) called version 4. The homogenized ECC temperature without an additional correction is called version 1. The best correction will be chosen first by using the ECC/lidar and ECC/MLS comparisons at 25 km. Four time evolutions of corrected ECC temperature are displayed in Fig. . We can see that the observed drop of ECC temperature for the period January 2002–June 2007 is improved in version 3 and version 4.

To compare with the ozone concentrations of the stratospheric lidar and MLS, ozone concentrations are calculated by using three corrected ECC temperatures. Originally, median ozone concentration of the ozonesonde at 25 km for period 2 shows -3.4 % lower compared with those of the lidar and MLS (Figs. d and b). Ozone concentration corrected by the ECC temperature version 2 is still -2.0 % lower (Fig. ). The ECC correction version 3 (ΔTP=0 K at 0 km and ΔTP=10 K at 30 km) should reproduce ozone concentration which is equivalent to that of the stratospheric lidar according to Table . The ECC median ozone concentration corrected by the ECC temperature version 3 is -0.5 % lower than the lidar median ozone concentration during the period 2 (January 2002–June 2007). This -0.5 % bias lies between the median difference obtained during period 3 (0.1 %) and during period 1 (-2 %). Comparisons between the ozonesonde and MLS show similar improvement of the bias with insignificant differences for version 3 and 4 considering the large standard errors. The magnitude of the ozone correction is also shown as a function of altitude in Fig. . We can see that this correction is smaller than 1.5 % below 15 km and becomes negligible compared to others uncertainties in this altitude range, but the correction could impact ozone trend assessment in the stratosphere.

Normalization factor is also a useful indicator for the consistency in ozonesonde data time series. The normalization factor corrected by the ECC temperature version 4 is unable to maintain a stationary time evolution for the correction and the corresponding ECC ozone of the normalization factor from 1991 to 2024 since a well defined drop of the of the normalization factor occurs during the 2002–June 2007 period (Fig. d). The result supports that the correction by the ECC temperature version 3 is the best option to improve the current homogenization work of the OHP ECC time series from January 2002 to June 2007.

The ozone concentrations of the ECC ozonesonde version 3 are compared to the stratospheric lidar for a layer of 2 km thickness at 25 km in Fig. a–b. The corresponding ECC data can be accessed at repository under data DOI . The correction by the ECC temperature improves the difference from -3.4 % to -0.5 %. The correction only slightly improves the median difference between the ECC ozonesonde and MLS at 26.1 hPa from -0.5 % to -0.2 % (Fig. c–d).