From June 2004, following the installation of an indium gallium arsenide (InGaAs) detector, an existing Bruker 120HR FTS (referred to as lauder01 in the TCCON archive) was used in a time-sharing mode to make routine measurements of solar absorption in the mid-infrared (MIR, for the NDACC) and near-infrared (NIR, for the TCCON). Acquisition of spectra in the two spectral regions (MIR and NIR) requires a beam splitter exchange, necessitating daily venting and re-evacuation of the instrument (and loss of observation time for the original MIR measurement programme). This was the operational TCCON instrument from 20 June 2004 until 28 February 2010. From 20 June 2004 to 30 May 2006, each retrieval consisted of 10 co-added interferograms measured at a scanner velocity of 20 kHz. After this period retrievals were made from individual interferograms measured at 40 kHz.

A prototype active tracking system was installed in the IFS 120HR solar tracker telescope in 2006, which adjusted the position of a 45∘ mirror immediately before the instrument input port based on feedback from quadrant diodes. The in-house-developed NIWA360 tracker as used by the IFS 125HR instrument was not installed for the IFS 120HR until 2011, after the end of the lauder01 dataset.

After installation of the IFS 125HR instrument (see below) 1 month of measurements using both instruments have been released to the TCCON archive to allow users to assess any biases between instruments. An assessment of bias between instruments is also made in Sect. .

Measurements in TCCON mode are still made periodically (approximately monthly) with this instrument to provide an independent check of the primary TCCON instrument. These data are not released to the TCCON archive.

In 2009 NIWA purchased a Bruker IFS 125HR spectrometer dedicated to TCCON measurements (lauder02 in the TCCON data archive). This instrument began routine TCCON measurements in February 2010.

Active tracking for this instrument is achieved using the in-house-developed NIWA360 system, which directs a small amount of the incoming solar radiation through a telescope and onto quadrant photo diodes, which steer the tracker to keep the solar image centred on the FTS input aperture to within 10 % of the radial extent of the solar disc, an accuracy of better than 0.025∘. The NIWA360 system has been used for the IFS 125HR since installation.

As with most TCCON sites, this instrument was originally supplied with a faulty ECL03 laser sampling board, resulting in spurious “ghost” signals, which interfere with the spectral fitting. The electronics board in the Lauder IFS 125HR was upgraded to an ECL05 board in January 2011 and the ghost-to-parent ratio (GPR) has since been monitored regularly and minimised whenever necessary. A full discussion on the treatment of this issue is given in Sect. .

Apart from some early testing between February and October 2010 when some measurements were taken at scanner velocities of 10 and 40 kHz, the majority of the IFS 125HR time series have been taken at 20 kHz. Following the installation of new control electronics in October 2015, measurements were transitioned to 10 kHz at the end of March 2016.

In order to satisfy the requirements of a TCCON station, and to carry out retrievals, a number of auxiliary measurements of meteorological parameters are required. A NIWA climate monitoring station is sited at Lauder, approximately 220 m from the FTS instruments. This includes measurements of

pressure: Vaisala PTB100A ±0.3 hPa.

The atmospheric research station at Lauder hosts a wide range of atmospheric remote sensing and in situ instrumentation. With respect to atmospheric greenhouse gas measurements, these include in situ FTS measurements of CO2, CO, CH4 and N2O , non-dispersive infrared (NDIR) measurements of CO2 using a LI-COR LI-7000 instrument and flask samples for subsequent measurements of CO2, CO, CH4, N2O and stable isotopologues of CO2, which are collected between 15:00 and 16:00 NZST when the wind speed is above 5 ms-1 in order to sample a well mixed boundary layer . The in situ measurements at Lauder are described more fully in Appendix A of . The site also hosts lidar measurements of ozone and aerosols.

The IFS 125HR FTS makes use of a helium–neon (HeNe) metrology laser running parallel to the measured solar beam in the interferometer to allow accurate measurement and control of the optical path difference by counting interference fringes. A fault with the original ECL03 laser amplifier board supplied with this instrument meant that the metrology laser was not sampled correctly at the zero crossing point between the peak and trough of the resulting sine wave . As a result, some of the spectral information above the Nyquist frequency of 7899 cm-1 was folded below it and vice versa, causing features known as ghosts.

During installation the LSE was minimised for a scanner velocity of 40 kHz, resulting in a small LSE at a rate of 20 kHz used for the majority of operational measurements.

Following the discovery of this fault, Bruker supplied all TCCON sites with replacement ECL05 laser amplifier boards, which allowed the zero level (i.e. the level at which the electronics identify the zero crossing point) to be adjusted. The rectified electronics board was installed in the Lauder IFS 125HR in January 2011. Since this time, regular (approx. every 3 months) checks have been conducted to check for the presence of ghosts and to minimise them by adjusting the zero level if necessary. This is achieved by taking measurements of a tungsten lamp, placing a band pass filter (5500–6600 cm-1) between the interferometer and detector, and measuring the signal aliased at 9800 cm-1. If any signal is detected, the zero level is iteratively adjusted in order to minimise it.

The 2014 release of the GGG processing software I2S algorithm makes use of the spectra recorded using the Si detector, which is wholly contained in the upper half of the alias, to derive the magnitude of the LSE and allow the spectra of both detectors to be resampled, removing the ghost signal. This method is described fully in .

Figure  shows the time series of LSE determined using the I2S algorithm over the period of measurements made using the IFS 125HR. Periods of instrument alignment are denoted by the shaded vertical bands whilst other notable instrument changes are indicated by the various vertical lines. The solid vertical line in early 2011 represents the change of laser amplifier board and subsequent improvement in LSE. Also of note is the change in LSE performance with individual metrology lasers, with the performance of the unit installed between April 2012 and January 2015 being poorer and less stable than other devices as indicated by the relative scatter in the derived LSEs.

The performance and alignment of both instruments is monitored through monthly retrievals of the instrumental line shape (ILS) from spectra obtained from a lamp source with a reference hydrogen chloride (HCl) cell, using the LINEFIT 14 software . The retrieved ILS describes the deviation from a theoretical ideal ILS, which is assumed by the GFIT retrieval code, in terms of the modulation efficiency (ME) and phase error, both of which vary as a function of optical path difference (OPD) of the interferometer. In the following discussion, ME will refer to the modulation efficiency at the maximum OPD of 45 cm used for TCCON measurements.

Figure  shows a time series of the retrieved ME and maximum phase error for both instruments. For the IFS 125HR, the mean of the ME is 0.9995 with a SD of 0.0106 indicating the stability of the instrument optical alignment over time. This value is significantly better than 4 % deviation in ME, which demonstrated was necessary in order to achieve the XCO2 retrieval accuracy required by the TCCON. The results for the IFS 120HR show more scatter and less stability, with a mean of 1.0124 and SD of 0.0256. Regular ILS retrievals from lamp spectra have been carried out since August 2010.

Due to the instability of the IFS 125HR instrument's internal clock, there have been a number of occasions when timing errors have been noted in the retrieved data. These are listed in Table . The errors have generally occurred when updates to the operating system or experimental set-up have prevented the instrument control computer from periodically resetting the time in the instrument's firmware. Such errors cause the retrieval scheme to attempt to fit the spectra to an incorrect air mass due to small differences in the calculated and true SZA. At the magnitude at which these errors occurred, the effect on the retrieved Xgas values is negligible as these are scaled by the retrieved column of O2 in order to remove such instrumental artefacts. However, the values of Xair that are retrieved and used as a diagnostic and quality indicator within TCCON are perturbed, as in this case it is the theoretical column of dry air that is scaled by the retrieved O2 column. By artificially introducing a timing error into retrievals that are known to have accurate timing, it has been shown that for a timing error of 60 s and a SZA less than 85∘, the perturbation of XCO2 is less than 0.02 %, compared to 1 % for Xair. Therefore, for the 2014 TCCON data release, data subject to these errors have been included without any attempt to correct them.

As noted previously, for an ideal retrieval, Xair would be unity, but due to spectroscopic limitations there is a TCCON wide bias and SZA dependence. These are, however, stable both temporally and across the network. Thus, Xair is a useful diagnostic of the quality of measurements, with retrievals deviating more than 1 % from the nominal value of 0.98 indicating systematic errors, e.g. the timing errors mentioned above, or deviations from an ideal ILS.

Figure  shows the time series of Xair retrievals from the IFS 125HR instrument that have satisfied the data quality criteria for release to the TCCON archive. It can be seen that most of these values fall within 1 % of 0.98 (indicated by the dashed lines); however, there are some periods with values outside of these limits, which have been identified as having timing errors. These periods are highlighted in green and defined in Table .

In Fig.  we show the hourly mean and SD (upper panel) of Xair values for the entire Lauder TCCON dataset covering both instruments.

There is a discontinuity in the SD of the hourly mean time series corresponding to a change from performing retrievals using 10 co-added interferograms sampled at a scanner velocity of 20 kHz to single scans at 40 kHz in May 2006. The discontinuity in both the mean and SD that occurs in early 2010 indicates the change from the IFS 120HR to the IFS 125HR and demonstrates the improvement in performance with the newer instrument.

In order to assess any biases between the two instruments used to collect TCCON data at Lauder, we have compared days on which measurements were taken with both instruments between October 2009 and December 2012. Figures  to show in the upper panels the daily mean retrieved values for XCO2, XCH4 and XCO respectively. Error bars represent the SD about the daily means. The lower panels show the difference (IFS 120HR – IFS 125HR) between the daily means, and the error bars are the sum in quadrature of the SDs about the daily means. Also plotted are the mean of the daily mean differences and the associated SD and standard error. These values are listed as percentages for all compared species in Table . For XCO2 the mean bias is 0.068 % with a standard error of 0.015 %.

Here we present a subset of retrieval time series available in the Lauder TCCON data archive. Figures  to show the hourly means of the retrieved Xgas, along with their SD in the upper panels. These figures are made up of data from both of the Lauder instruments with the change-over occurring on 1 February 2010.

The time series of XCO2 values is shown in Fig. . Over the length of this time series XCO2 has been increasing by approximately 2 µmolmol-1yr-1. There is a seasonal cycle due to summer drawdown, but it is smaller than in the Northern Hemisphere . Since measurements using the IFS 125HR began, the SD about the hourly mean has been better than 0.1 % for 96.81 % of retrievals and better than 0.25 % for 99.96 % of retrievals. The precision of the IFS 120HR was poorer, with only 14.95 % of hourly SDs better than 0.1 and 95.93 % better than 0.25 %. The precision of both instruments compared to the TCCON targets are shown in Table .

Figure  shows the XCH4 time series. XCH4 has a smaller positive trend of approximately 5.6 nmolmol-1yr-1 and a seasonal cycle of similar magnitude with a summer minimum due to oxidation by photochemically produced OH radicals. The Lauder TCCON XCH4 time series starts around 2 years before the end of a period of stable atmospheric CH4 concentrations that has been observed in a number of different time series (, and references therein) and this is also apparent in the data shown here.

The time series of XCO shown in Fig.  has a small, negative trend and a stronger seasonal cycle with a summer minimum due to OH oxidation and a springtime peak attributed to CO transported from Southern Hemisphere biomass burning sources .

Figure  shows the time series of the remaining retrieved species.