This chromatographic system was built based on the Shimadzu Nexis GC-2030 equipped with two parallel detectors: BID and TCD (Fig. 1). The carrier gas was controlled by an advanced flow controller (AFC) connected to a split/splitless injector. Between the AFC and the injector, a two-position six-port valve with a 1 mL (or 2 mL) sample loop was placed on the carrier line.

The gas chromatograph oven was equipped with an additional cryogenic option (CRG) where liquid nitrogen (LN₂) was used as a cooling agent, which allowed for separation at below-ambient temperatures. The samples from the AS-210 greenhouse gas autosampler (SRI Instruments Europe GmbH) were transferred to the sampling valve through a stainless-steel transfer line continuously warmed to 110 °C with heating tape to prevent moisture contamination. The presence of moisture in the samples results in the loss of sorption capacity of carbon molecular sieves which are used commonly for separation of gases (Fastyn et al., 2003). The injection was performed by valve rotation. The sample was transferred from the loop (1 or 2 mL) through the injector at a total flow of 10 mL min⁻¹ and was then split 1:7 just before the column inlet. This was sufficient to achieve a good peak shape with sufficient area.

Additionally, the injection sample was then divided with using a T-joint connector between two porous layer open tubular capillary columns filled molecular sieve 5A (RT-Msieve 5A 30m×0.32mm×30 µm; Restek, USA, catalogue no. 19722) and fused silica (Carboxen 1010 PLOT 30m×0.53mm×30 µm; Supelco, USA, catalogue no. 25467). The dimensions of the columns were selected to achieve a splitting ratio of 1:5, directing most of the sample to the Carboxen 1010 PLOT and BID. Corresponding calculations were performed in Shimadzu AFT (Advanced Flow Technology) software (Fig. 2).

Extremely low baseline noise (signal-to-noise ratio (S / N) always above 10) was achieved by a combination of two factors: a high-purity carrier gas helium of grade 5.0 connected to the Valco helium purifier HP2 (VICI, Valco Instruments Co. Inc.) and particle traps (2.5m× 0.32 mm o.d.) mounted on the columns' outlets. The presence of the traps protected the detectors from particles dislodging from the porous layer open tubular (PLOT) capillary column, which can cause spikes.

Both of the detectors used are concentration dependent; therefore, to obtain the highest sensitivity on the BID channel, the discharge gas flow rate was decreased from a default of 50 to 30 mL min⁻¹, which is the lowest possible flow. Below this value the plasma flame is not stable and tends to flicker or is extinguished. Detection levels did not need to be taken into consideration with the TCD.

The linearity of both detectors' response was controlled and maintained during all the measurements with minimal R2=0.99 applying at least two standard gases and zero point.

The temperature programme for gas chromatography analyses started at 100 °C for 13 min and later increased to 200 °C at a rate of 25 °C min⁻¹ with the oven set at 200 °C for 1 min. The temperature of the split/splitless injector was 120 °C. The TCD and BID were at an equal temperature of 220 °C. The carrier gas pressure was 70 kPa and the column flow was 5.3 mL min⁻¹. Linear velocity was 41.2 cm s⁻¹ and purge flow was 1 mL min⁻¹. The total flow for split ratios 1, 2, 3, 4 and 5 was 11.5, 16.8, 22.0, 27.3 and 32.5 mL min⁻¹, respectively.

Standard gas mixtures used for testing and final determination of the measurements precision were atmospheric air from Wrocław (Poland) (analyses of N₂, O₂, CH₄, CO₂ and N₂O at ambient atmospheric concentrations) and a special gas mixture from Messer (CH₄ 10 ppm, CO₂ 1000 ppm, N₂O 50 ppm, diluted in pure N₂). The in-house standard of compressed air from Wrocław (Poland), which contained natural moisture (water vapour), was stored in the 10 L gas cylinder. It was prepared using an oil-free compressor for diving cylinders. The second standard was ordered for Messer Polska Sp. z o.o. and is the commercial product prepared in Switzerland according to ISO6141:2015. This standard was prepared in pure N₂, without moisture, and in a volume of 8 L, and it contains an F10 filter, which protects the outer valve from the possible water vapour or solid particles. The standards were directly connected by a 1/8 in. capillary to the AS-210 greenhouse gas autosampler. The sample loops used for tests of standard gases were 1 and 2 mL. The atmospheric air was tested for splits 1, 2, 3, 4 and 5. The special gas mixture from Messer was tested for splits 3, 4 and 5.

The chromatographic system for GHG analyses using a single BID was initially tested for application of the SH-Q-BOND column (30m×0.53mm i.d. ×20 µm df; catalogue no. 221-75765-30) from Shimadzu, which allows for separation of CH₄, CO₂ and N₂O and is resistant to water vapour contamination. The scheme showing the GC configuration for testing of SH-Q-BOND and RT-Msieve 5A is presented in Appendix A (Fig. A1). In this configuration most of the parameters were exactly the same (length of the column, diameter, film thickness, flow parameters, split ratio) as in the configuration using the Carboxen 1010 PLOT column. The only differences were the usage of a SH-Q-BOND column for separation of CH₄, CO₂, and N₂O and a different column oven temperature programme.

Separation of these gases was tested at different low temperatures of the column oven (30, 35 and 40 °C). The lowest temperature (30 °C) was difficult for the oven to achieve quickly without using a cryogenic trap. At the laboratory's normal temperature (22 °C), it was possible to decrease the oven temperature rapidly to 35 and 40 °C, but unfortunately both temperatures were insufficient to separate CO₂ from N₂O at a retention time interval longer than 30 s, which appeared very problematic while analysing real samples of elevated CO₂ concentration. When the CO₂ concentration was high (e.g. 700 ppm), the tail of the CO₂ peak partially covered the N₂O peak, as shown below in Fig. 3a and b, and ultimately the N₂O peak area was understated. Moreover, this set-up did not allow detection of CH₄ in atmospheric concentrations.

Therefore, after basic tests of the SH-Q-BOND column at different temperatures, it was decided to check the retention times of individual CH₄, CO₂ and N₂O gases on the Carboxen 1010 PLOT column. The Carboxen 1010 PLOT column offered very good separation of CO₂ from N₂O, even at a very high concentration (CO₂ 1000 ppm), as shown in Fig. 4 (peaks of CH₄ and N₂O are visible only when zoomed). Moreover, the ambient CH₄ was very well separated from the N2+O₂ peak (Fig. 4). The longer programme of separation guaranteed the ideal separation of CH₄, CO₂ and N₂O (more than 2 min between each gas). The time for one single analysis is 18 min, but this is necessary because a low flow of the carrier gas is recommended for the Carboxen column by its manufacturer. A carrier gas flow that is too high (e.g. above 20 mL min⁻¹) causes faster destruction of the column and contamination of the particle trap, and subsequently of the detector, with fragments of the column filling. However, the disadvantage of the application of the Carboxen column is also its low resistance to water vapour. Therefore, before starting the analyses, all lines of the AS-210 greenhouse gas autosampler as well as the sample loop were carefully heated using heating tape and a gun heater to remove water from the stainless-steel capillaries and metal parts of the valves. The parallelly connected column Rt-Msieve 5A and TCD allow for determination of N₂ and O₂ (and if necessary CH₄) concentrations in the range from 0.2 % to 100 % (example shown in Fig. 5).

The analyses of the compressed air standard with split ratios 1, 2, 3, 4 or 5 and using 1 and 2 mL sample loops showed a different precision, expressed as a coefficient of variation (CV; calculated as the standard deviation divided by the mean value, expressed in %). The CV of the CH₄ measurement (1.895 ppm) was in the range of 3.31 % to 8.57 % for the 1 mL sample loop and 1.51 % to 2.76 % for the 2 mL sample loop. In the case of CO₂ (411 ppm), the CV ranged from 0.91 % to 2.83 % for the 1 mL sample loop and from 0.57 % to 1.87 % for the 2 mL sample loop. The CV of the N₂O measurement was lower for the 2 mL sample loop and ranged from 7.09 % to 12.67 %, compared with the 1 mL sample loop where the CV of the N₂O measurement ranged from 9.15 % to 20.19 %. Generally, it was observed that gases at low detection limits (CH₄, CO₂) were measured more precisely using the 2 mL sample loop.

Measurement of N₂O in split ratio 3 resulted in a significantly higher CV when compared with the results obtained in split ratios 4 and 5 (Table 1). This is because the injection of a higher amount of water vapour contained in the sample partially covered the peak area of the N₂O (by increasing the baseline level), similarly to the measurement in split ratio 2 (CV 20.19 %). In this case the lower CV at split ratio 1 was only calculated for three measurements to avoid unnecessary contamination of the column.

The gases analysed using the TCD, O₂ and N₂, were characterised by a narrow CV ranging from 0.10 % to 0.39 %. The highest CV (0.39 %) was observed for the N₂ measurement with the 2 mL sample loop, where the peak area was very large. The results of the measurement (peak area, SD, CV) are presented in Table 1.

The CH₄, CO₂ and N₂O measurements of the special gas mixture standard at split ratios 3, 4 and 5 were characterised by a repeatable CV within a narrow range from 0.11 % to 3.22 %. The CV of the CH₄ measurement (10 ppm) using the 1 mL sample loop ranged from 0.11 % to 0.55 %, while for the 2 mL sample loop the CV was in the range of 0.34 % to 1.79 %. The CV of CO₂ (1000 ppm) for the 1 mL sample loop was between 2.05 % and 3.08 %, and for the 2 mL sample loop it was between 1.57 % and 3.22 %. The N₂O measurement (50 ppm), which is a very high concentration (rare in the natural environment), was characterised by a CV in the range of 0.18 % to 0.44 % for the 1 mL sample loop and 0.85 % to 2.14 % for the 2 mL sample loop. These values clearly show that measurements of the gas mixtures with relatively high concentrations of N₂O using BID were repeatable for all splits 3, 4 and 5 and were slightly better using the 1 mL sample loop. However, the application of the sample loops (1 and 2 mL) at split ratios 3, 4 and 5 guaranteed the achievement of repeatable results. Table 2 shows all the data of the standard gas mixture testing measurements.

Another testing of the GC system was carried out with the application of splits 3, 4 and 5 with a 2 mL sample loop (Table 3). The ambient air from the laboratory on 1 d was analysed directly from the AS-210 greenhouse gas autosampler (empty plate for vials, which enabled direct sampling of the ambient air from the needle to the line connected with the GC's sample loop). The tests were performed with splits 3, 4 and 5 (splits 1 and 2 were omitted to avoid excessive introduction of air containing natural moisture into the Carboxen column).

The CV of CH₄ was in the range of 3.40 % to 4.10 % (the highest value for split 4). The CV of CO₂ was in range of 1.31 % to 2.29 %, whereas for N₂O it was between 3.11 % and 4.04 %. These CV values are close to the results obtained during measurements of the compressed air standard (Sect. 3.3, Table 1). The difference between the two experiments is that compressed air always had the same composition and gas concentrations, whereas the CH₄ and CO₂ concentrations in ambient air could change slightly over time (daily variability).

To verify the long-term stability of the measurements and the system performance for real samples, a 55 h long measurement series of ambient laboratory air was performed. The subsequent air samples were measured in parallel with the GC set-up and with the optical instruments dedicated to analyses of GHG concentrations and isotopic signatures (Picarro G5131-i for isotopic N₂O and Picarro G2201-i for CO₂ and CH₄) (Picarro, Santa Clara, USA). The reference methods were applied to check whether slight changes in GHG concentrations over the day/night period can be monitored well with this GC system. The reference instruments – isotopic Picarro – showed a quite narrow range of possible concentration measurements (Picarro G5131-i isotopic N₂O up to 2000 ppb, Picarro G2201-i up to 2000 ppm CO₂ and up to 12 ppm CH₄) but a very high precision for ambient concentrations, without the need for calibration. Therefore, they served here as an ideal reference method.

During this 55 h time series, the Picarro measurement was performed every 3 min and the GC measurement every 19 min. For GC measurements, split 4 was applied. The concentration trends for CO₂ and CH₄ were observed to be generally in good agreement, and the N₂O concentration was very stable (Fig. 6). Importantly, it was observed for CO₂ that after 24 h the measurements were slightly recalibrated and shifted in relation to the reference method. This indicates the need for repeated calibration at least every 24,̉ especially for CO₂. However, even without recalibration, the maximum difference between the GC measurement and the reference value was below 3 % for both CO₂ and CH₄. The largest variations in the GC results were observed for N₂O, especially when comparing them to the very stable Picarro measurements. This is the most challenging analysis, since N₂O ambient concentrations are lowest and hardest to measure correctly. The maximum difference between the GC measurement and the reference value for N₂O was around 8 %, and the standard deviation of GC measurements was 15 ppb, which represents a 4.4 % error. This is quite high when compared with Picarro statistics where the standard deviation over 55 h of measurement was 0.24 ppb, which represents less than 0.1 % error. However, for typical N₂O measurements of unknown sample with GC techniques, a 5 % error is a satisfactory result, typically given as an accepted GC measurement error in research studies (Arnold et al., 2001; Harvey et al., 2020). The precision obtained for ambient air measurements is similar for CO₂ and CH₄ compared with classical FID measurements, with an error of around 2 % (Loftfield et al., 1997), but is lower when compared with ECD measurements, for which a 1.2 % error has been reported (Loftfield et al., 1997). Usually systems that enable the simultaneous measurement of CH₄ and CO₂ are very accurate, but the main limitation is the upper detection limit. For example, Wang and Wang (2003) achieved a CH₄ precision error in a range from 3.37 % (ambient concentration) to 0.05 % (60 ppm), or for CO₂ from 0.66 % (ambient concentration) to 0.04 % (4000 ppm). In the present system, higher concentrations of CH₄ can be measured by the TCD simply using a RT-Msieve 5A column. The chromatographic system can be improved by the addition of a second gas valve aimed at transferring CO₂ to the TCD after separation using the Carboxen column. This would allow further improvement of this chromatographic system for the maximum range of measured GHG concentrations.