Both GC channels use Valco 10- and 12-port two-position valves (VICI, Houston, TX) to control flow switching and ECDs (Valco and Shimadzu) to detect specific trace species with high precision, with added dopant gas as needed (Phillips et al., 1979; Fehsenfeld et al., 1981). In its original configuration, one channel used OV-101 in packed columns to separate and measure CFC-12, halon-1211, and CFC-11 every 70 s, similar to the Lightweight Airborne Chromatograph Experiment (LACE) (Moore et al., 2003). After the initial Altair flights in 2005, these were replaced with Unibeads (pre-column, ∼ 1 m length) and molecular sieve 5A (main column, 0.7 m) to measure molecular hydrogen (H2), methane (CH4), and carbon monoxide (CO) every 140 s. A tiny flow of nitrous oxide (N2O) dopant (∼ 0.003 sccm) added to the ECD is required for adequate sensitivity. Flows and column temperatures varied over different missions; in ATom, with the chromatography optimized for both the troposphere and stratosphere, at a temperature of 94 ∘C the N2 carrier gas flow was 60 sccm, with 4 s of pre-emphasis at 100 sccm at the start of the injection, to rapidly bring the pre-column up to the same pressure as the main column. The pre-column was back-flushed after 25 s to remove any remaining compounds over the remainder of the 140 s time window.

The second channel uses a pre-column (0.6 m) and main column (1.8 m) of Porapak Q, followed by a post-column of 5A molecular sieve (originally 0.20 m, now 0.25 m) to measure sulfur hexafluoride (SF6) and N2O every 70 s; doping the nitrogen carrier gas with CO2 enhances the ECD response to N2O. The pre-columns and main columns were maintained at 91 ∘C and the post-column at ∼ 120 ∘C (for the shorter version) and 190 ∘C (longer version; changed in 2011). Carrier gas flows in ATom were 55 sccm, with the flow in the pre-column reversed after 13–14 s. Since the backflush switches occur early in the cycle, there is sufficient time for the pre-columns in each channel to be cleaned out, even with lower flow rates compared to the main flows. All the columns used were packed in 3.2 mm o.d. stainless-steel tubing, wound around a circular mandrel with heater cartridges and a resistance temperature detector (RTD), and packaged in insulated metal cans. The ECDs were packaged in similar cans but sealed and supplied with a ∼ 5 sccm purge flow to prevent oxidation and maintained at 330 or 350 ∘C with Omega temperature controllers.

Chromatograms are similar to those in Moore et al. (2003), Figs. 7c and 9. ECDs provide very high sensitivity but can have non-linearity, particularly for doped channels, where secondary ion–molecule reactions are used to detect trace species. UCATS was calibrated on the ground during each mission, with a set of standards spanning the range of expected atmospheric concentrations (typically 30 %–100 % of those in background tropospheric air) and occasionally including zero air to check baselines. From these experiments, calibration curves for each molecule are calculated, including an estimate of the error in the calibration. An example of a calibration curve for N2O is included in the Appendix (Fig. A2); all other curves were even closer to linear. In flight, a calibration standard from compressed background tropospheric air is injected every 6–10 min, and the peak heights of air samples and standards are analyzed with the calibration curves to generate a time series of mixing ratios for each molecule in sampled air.

Ambient air is drawn into UCATS from a side-facing or rear-facing inlet extending 25–30 cm from the skin of the aircraft (outside the aircraft boundary layer) through stainless-steel and Synflex tubing; sample flow tubing inside UCATS is stainless steel. Air is pressurized in the GC sample loops by an external two-stage KNF diaphragm pump (model UN726, with Teflon-coated heads and diaphragms) and maintained at 1225 hPa with an absolute pressure relief valve (Tavco, Inc.; Chatsworth, CA); excess air is dumped through the Tavco overflow. Air flows at approximately 80 sccm sequentially through the two sample loops (∼ 0.5 cc volume for each channel) and a flow meter and is controlled by solenoid valves and a pressure regulator set at 1080 hPa on the outlet. Every 70 or 140 s the contents of the sample loops are injected by the two-position Valco valves onto the pre-columns, providing a discontinuous ∼ 2 s snapshot measurement of ambient air.

The sample airflow is split just upstream of the GC pump to feed a tunable diode laser (TDL) hygrometer with its own pump (KNF, model NMP850) downstream of the absorption cell. The original hygrometer, a custom commercial sensor from MayComm, Inc., used infrared absorption at 1.37 µm with second harmonic detection to measure water vapor. Since water vapor number densities span 5 orders of magnitude from the surface to the stratosphere, the laser beam was split into two optical paths, a 13.4 cm “short path” for measurements from the surface to the mid-troposphere (40 000 to 500 parts per million (ppm)) and a 403 cm multi-pass “long path” for measurements from the mid-troposphere into the stratosphere (1000 to < 5 ppm). On the Altair missions, with a minimal payload, a small Vaisala probe was installed on the inlet for measurements of temperature, pressure, and relative humidity. This was not used subsequently, as the payloads on larger aircraft included dedicated instruments for meteorological measurements.

During ATom, the TDL hygrometer in UCATS was upgraded with a new model from Port City Instruments (Reno, NV), the successor to MayComm. It is similar in concept and uses a distributed feedback laser (DFB) to scan across two closely spaced absorption lines near 2.574 µm. Absorption at this wavelength is much stronger than in the original instrument, allowing higher sensitivity in the stratosphere. As before, the laser beam is split into two optical paths, with the short path (5.14 cm) for high values of tropospheric water (∼ 2000–40 000 ppm) using direct absorption. The long path (280.0 cm) is used with second harmonic detection for water vapor from 0–100 ppm, and intermediate values (100–5000 ppm) are measured using the long path with direct absorption. A second weak absorption line is also analyzed with direct absorption for water vapor mixing ratios above 1000 ppm; this is not being used at present. Both long- and short-path spectra are recorded simultaneously, with each scan taking approximately 200 ms. All four measurements of water vapor are calculated, and then each one is averaged together for ∼ 1 Hz output on a serial data line. All data are recorded by the UCATS computer, and the appropriate value for display and archiving is chosen based on the range of pressure and water vapor. Both instruments required extensive calibration using prepared water vapor standards and frost point hygrometers for accurate measurements. The new instrument allows higher-precision (±0.1 ppm) measurements of water vapor in the stratosphere compared to the original instrument, which was limited to ±1 ppm.

Ozone was measured by direct absorption (Beer–Lambert law) UV photometers from 2B Technologies (Boulder, CO), modified for high-altitude operation and mounted inside the UCATS package. The initial ozone instrument was a 2B model 205; modifications included a stronger pump (KNF, model UNMP-830), a small metal cylinder upstream of the pump to dampen pressure fluctuations that could degrade the measurement precision, O3 scrubbers with manganese dioxide (MnO2)-coated screens (Thermo Fisher), and pressure sensors with a range from 0 to over 1000 hPa (Honeywell ASDX series). Ambient air was brought to the ozone instruments from the inlet through a separate Teflon tube (6.35 mm o.d.), with the exhaust from the ozone and water instruments combined inside UCATS and released outside the aircraft. The model 205 is a dual-beam photometer, with the flow continuously split between unscrubbed (ambient) air into one cell and scrubbed (ozone-free) air into the other. Two photodiodes located at the end of 15 cm long absorption cells measure the intensity of 254 nm radiation emitted from a mercury lamp. Ozone concentrations are calculated from the ratio of measured intensities through the cells with scrubbed and unscrubbed air according to the Beer–Lambert law. The flow paths are switched by solenoids every 2 s, to allow alternating measurements of ambient and scrubbed air in each cell, with data averaged to 10 s on the original model. The instruments are checked against a NIST-traceable calibration system (Thermo Fisher, model 49i) on the ground before and after every mission. For ATom, a new and more sensitive 2B model 211 ozone photometer was added to UCATS in addition to the original model 205, with similar modifications as before and additional changes as described in Sect. 3.

The overall dimensions of UCATS were initially 41×46×26 cm, with a weight of 29 kg. To integrate the new water and ozone instruments for ATom, an additional section was added to the top, increasing the height from 26 to 33 cm and the weight to 33 kg. The external GC pump weighs an additional 5 kg, and fiber-wrapped aluminum bottles (SCI Composites; now Worthington Industries) for compressed nitrogen (N2) carrier gas (model 687) and dry, whole air calibration gas (model 209) for the GC, both filled to ∼ 13 000 kPa, together weigh approximately 7 kg. The total N2 flow (carrier gas, backflush, purge flows) is about 300 sccm. The N2 bottle needed to be filled every one or two flights, the calibration gas was filled once per deployment, and the small dopant bottles could last for over a year without refilling. For flights on passenger aircraft, such as the DC-8 for ATom, the N2 and air bottles can be replaced by larger gas cylinders as weight and space allow. UCATS is powered by 28 V DC, and the complete package draws 12 A at startup (∼ 350 W), decreasing to 150 W after the heaters warm up (∼ 30 min). The majority of the power is consumed by the column and ECD heaters; the GC pump and the TDL use about 1 A each, ozone less than 0.5 A, and other electronics about 1–2 A. Different voltages (+5, ±12, 15, and 24 V) are supplied by Vicor DC–DC converters. The internal wiring in the ECD cans is carefully adjusted to minimize electrical noise on the detector circuits; no other electromagnetic compatibility issues were observed. UCATS is controlled by an Ampro computer with the QNX operating system, and data are stored on flash memory for post-flight processing; quick-look, near-real-time data for ozone, water, N2O, SF6, and CH4 are also provided by a serial or Ethernet connection to the aircraft, for onboard use and telemetry to the ground. Data are analyzed post-flight with home-built software, including GC peak integration and quantitation routines, primarily using Igor software, and three separate data files (GC, ozone, and water, with different time intervals) are generated for archiving and dissemination.

Global Hawk flights during the GloPac mission covered a wide range of air masses in the stratosphere and provided an opportunity to demonstrate the capabilities of UCATS in the environment for which it was designed. Figure 2 shows a scatter plot of SF6 vs. N2O mole fractions for the flight of 23 April 2010 from Edwards Air Force Base, CA, to the western Arctic Ocean and back (∼ 35–85∘ N, 120–165∘ W) at altitudes from 16 to 20 km, with two profiles down to 13 km and back. N2O and SF6 are long-lived greenhouse gases emitted at Earth's surface and generally decline with altitude in the stratosphere (e.g., Plumb and Ko, 1992). For N2O this is primarily due to photochemical loss in the stratosphere, and for SF6 it is because older air entered the stratosphere at earlier times, when tropospheric SF6 mixing ratios were lower (Hall et al., 2011). As a result, N2O and SF6 are correlated in the stratosphere, with older air and air from higher altitudes having the lowest mixing ratios for both gases. This correlation can be seen in Fig. 2, where N2O declines strongly from its tropospheric value (∼ 320 ppb in 2010) as SF6 (tropospheric value ∼ 7 ppt in 2010) approaches 5.5 ppt. Data from ACATS-IV taken on the ER-2 aircraft almost 13 years earlier in the Arctic during the 1997 POLARIS mission are shown with the GloPac data for reference. POLARIS N2O and SF6 mixing ratios were adjusted upward for the tropospheric growth over the 13 years between missions (N2O increased from 312.5 to 322.9 ppb and SF6 from 3.9 to 7.0 ppt) by adding the difference in SF6 tropospheric values to the POLARIS data and multiplying the ratio of tropospheric values to the POLARIS data for N2O and other tracers that are photochemically destroyed in the stratosphere. This is mainly to bring the two data sets onto the same scale for easy visualization, though the similarity does reflect the relatively stable nature of stratospheric circulation and photochemistry. Average tropospheric values of these long-lived greenhouse gases were taken from the NOAA GML network (http://www.esrl.noaa.gov/gmd/ccgg/trends/global.html, last access: 29 September 2021). The measurement precision for UCATS is about ±0.05 ppt SF6 (1σ) and ±1.5 ppb N2O, similar to ACATS but with a data rate of every 70 s instead of 360 s. The slightly more gradual slope for POLARIS data is due to the fact that SF6 was increasing more slowly in the 1990s than in the years just before GloPac (Hall et al., 2011) and perhaps other more subtle differences in the trends. Similar plots of GloPac and (adjusted) POLARIS CH4 vs. N2O and H2 vs. CH4 data from the same flights (Fig. 3) show close correspondence between the two campaigns and tight, nearly linear correlations, as expected from the fact that these are all long-lived gases in the stratosphere. Overall, UCATS precision for CH4 and H2 was ±7–8 ppb (0.5 %) and ±5 ppb (1 %) respectively, equal to or slightly better than that of ACATS-IV and with a slightly faster data rate.

The START-08 and HIPPO missions were the first tropospheric campaigns for UCATS. On 12 January 2009, during the first HIPPO deployment, the GV sampled air in both the troposphere and stratosphere as it traveled from Anchorage, AK, north to near 80∘ N over the Arctic Ocean and back. The precision for UCATS N2O during most of the flight was near ±1 ppb in both the troposphere and stratosphere, calculated from flight segments with near-constant N2O near the start of the flight and comparisons with QCLS and PFP samples throughout the flight (Fig. 4a). A more quantitative comparison can be made by plotting UCATS and PFP data for the entire HIPPO 1 deployment against the higher-time-resolution QCLS data (Fig. 4b). Each UCATS GC measurement is a roughly 2 s average of the atmospheric composition along the flight path a few seconds before the air sample is injected and is plotted here against the corresponding 10 s average of QCLS data. Each PFP flask takes between 30 s and a few minutes to fill, depending on altitude, and a comparison with QCLS data is enabled by averaging the QCLS data over the sampling interval associated with each flask sample. QCLS data have been corrected here for the approximately 1 ppb offset with respect to the PFPs reported by Santoni et al. (2014) during HIPPO in 2009–2011. The UCATS vs. QCLS correlation allows an upper limit estimate of UCATS precision, assuming all the error is associated with the UCATS measurements, none from QCLS, and that effects related to atmospheric variability arising from timing mismatches are negligible. The resulting standard deviation (1σ precision) is ±1–2 ppb over the entire month of HIPPO flights, from the high Arctic through the tropics to the Southern Ocean and back. The slope of the fit is 0.91 ± 0.004; this difference has not been resolved. We note that the slope for the PFP data is 0.93 ± 0.02, though this is partially driven by the smaller slope for tropospheric (high N2O) data, as opposed to for UCATS, where the slope is also smaller in the troposphere but clearly reflects differences in the stratosphere (low N2O), where the dynamic range of N2O is large. The UCATS and PFP results agree closely over the more limited range of PFP N2O data, but because the measurements were not simultaneous, a quantitative comparison is not possible.

In ATom, with the GC system optimized for all conditions, UCATS produced precise and accurate data in the troposphere with both short- and long-term stability and without degradation in the humid tropics. This is demonstrated by N2O and SF6 time series for the DC-8 flight of 29 January 2017, from Palmdale, CA, to northern Alaska and back to Anchorage (∼ 35–70∘ N, 120–155∘ W), and scatter plots for the entire ATom-2 deployment (Fig. 5). Data from UCATS, PANTHER, QCLS (N2O only), and PFPs show excellent agreement for the time series (mean differences typically ±1–2 ppb N2O and ±0.05 ppt SF6). ATom QCLS N2O data show a similar offset relative to PFPs (taken as the reference instrument on board) as observed during HIPPO, attributed to the QCLS calibration procedure. As in Gonzalez et al. (2021), QCLS N2O data are corrected by subtracting the offset with respect to the PFP data (∼ 1.2 ppb). The 1σ precision of UCATS and PANTHER was ∼ ±1 ppb N2O and ±0.05 ppt SF6 from both the time series and the scatter plots (again assuming all the variability in the comparison with QCLS is associated with the GC data). As described above, SF6 and N2O are well correlated in the stratosphere, and the precision of SF6 from the lower right panel can be estimated for stratospheric data (lower values of both gases). In the troposphere, the plot reflects the strong latitudinal gradient in SF6, with lower SF6 in the Southern Hemisphere and higher SF6 in the Northern Hemisphere. This leads to a much steeper apparent slope, since N2O also has a latitudinal gradient, but weaker, also with lower values in the Southern Hemisphere. Transitions between the troposphere and the stratosphere lead to mixing lines between the two branches (from ∼ 310–330 ppb N2O). The only disagreement for N2O is at low values, where PANTHER and UCATS both measure about 3 ppb lower than the QCLS instrument, a deviation in the opposite direction compared to HIPPO. The tropical flight of 3 February 2017 (Fig. A3) illustrates the precision of N2O and SF6 where air masses sampled along the flight track varied slowly (because of its altitude range, the DC-8 is always in the troposphere at these latitudes). H2 measurements also showed good agreement between UCATS, PANTHER, and PFPs (Fig. 6), with nearby data points from the different instruments typically differing by about ±5 ppb (1 %) over the entire range of observed values. Values for precision and agreement of measurements from ATom and other missions are summarized in Table 2.

This section is primarily focused on the ATTREX mission, which was designed to probe the chemical composition of air over the tropical Pacific and transport into the stratosphere but applies to all UCATS stratospheric data. Because ozone mixing ratios peak in the stratosphere, the main requirements for a stratospheric ozone measurement are accuracy and stability, with sensitivity to low values usually less critical. In GloPac, the 2B model 205 in UCATS agreed within 1 % with the NOAA classic ozone instrument over the large observed range of ozone mixing ratios (Fig. A4). However, as discussed in Sect. 3, requirements are different in the TTL, where ozone mixing ratios are very low, often less than 30 ppb. Both measurement accuracy and precision are essential at these low values, and even errors of a few parts per billion in ozone (or small measurement biases in water vapor and other trace gases) can lead to different interpretations of the underlying atmospheric processes. The accuracy of the model 205 ozone instrument can be calculated similarly to Proffitt and McLaughlin (1983), where the most important uncertainties are the absorption cross section of ozone, the accuracy of cell temperature and pressure measurements, the absorption path length, and any nonlinearities in detector response. These add up to a few percent, but the initial calibration of the 2B instruments against a reference standard (by 2B) should correct for any slight inaccuracies. In all our calibration checks, the slope was within 1 % of unity and the offset less than 2 ppb (usually < 1 ppb) at ambient pressure (∼ 840 mbar in Boulder, CO, and 920 mbar in California) and room temperature.

As described in Sect. 3, two model 205 sensors were flown in UCATS during ATTREX-2 and 3, with data from the two instruments merged into a single data set with faster time resolution than the original instrument. A comparison of UCATS and NOAA-2 ozone data from ATTREX-2 (Fig. 7) shows that the slope is close to unity with a crossover point near 500 ppb. At low ozone (20–30 ppb), the UCATS data are on average lower by 3–4 ppb. Since the absorption cross sections are the same for both instruments and cell length is fixed (and measured to better than 1 % accuracy), the principal known sources of error are inaccuracies in measured cell temperature and pressure. The pressure sensor in the older 2B instrument was carefully calibrated over a range of pressures for many years (2010–2016) and was stable throughout that time. A small correction was made to account for the pressure drop from the cell to its outlet (where pressure is measured). This introduced about a 1 % increase in ozone at the highest altitudes but was negligible at lower altitudes. Temperature is measured on the cell body rather than in the airflow, but air temperature should have time to equilibrate inside UCATS before reaching the ozone instruments. (Flow to each 2B is ∼ 10 % of that for NOAA-2, which has been shown to measure temperature accurately after warming ambient air as it flows to the cell (Gao et al., 2012).) The offset between NOAA-2 and 2B data bears further examination. UV ozone photometers have been shown to produce offsets when transitioning from wet to dry conditions (Wilson and Birks, 2006), and that is certainly the case for the model 205, as discussed in the following section. However, except on initial ascent, air sampled in ATTREX was always extremely dry, and any artifact should become negligible within 1 h. Similar agreement between NOAA-2 and the original 2B instrument was obtained on ATTREX-1 and GloPac. Laboratory tests for measurement artifacts of the 2B under various conditions produced mostly negligible offsets and always less than 5 ppb.

During ATTREX-3, payload weight and balance issues prevented the NOAA-2 instrument from being flown on the Global Hawk. Coincident balloon-borne electrochemical concentration cell (ECC) ozonesonde launches from Guam provide a comparison for these flights. Data from the last 2 h of the 16–17 February 2014 Global Hawk flight (Fig. 8) most closely overlapped one of the balloon profiles in space and time (within 100 km and 1–2 h). The agreement between the ECC and the 2B instruments in the troposphere (< 16 km, where the balloon and aircraft were in closest proximity) is quite good and shows no significant bias in the UCATS data. A further check on UCATS ozone data is shown in Fig. A5 with ozone data from the GV aircraft (operating during the concurrent CONTRAST mission; Pan et al., 2017), the Global Hawk, and the Guam ozonesonde launch, which was timed to overlap with the return of the Global Hawk on 13 February. In summary, based on laboratory calibrations, tests, and in-flight comparisons, we assign a systematic uncertainty of less than 5 ppb to our model 205 ozone data in the TTL and lower stratosphere. The precision in the TTL ranged from ±5 to 10 ppb but can be improved by temporal averaging. The low values of ozone in the TTL demonstrate the importance of precise and sensitive ozone measurements in this region and the need to minimize or eliminate any systematic errors.

As described above, the model 205 in UCATS disagreed with the NOAA classic ozone instrument during HIPPO following transitions between wet and dry air. Most flights had only minor artifacts, but the issue was most pronounced in the tropics, with an example shown in Fig. 9 (top). At low altitudes there was generally good agreement (mean difference is 0.4 ppb, standard deviation is ±4.2 ppb for HIPPO 4), but as the GV aircraft climbed out of the very wet lower troposphere to higher altitudes, or descended back into the lower troposphere, changes in water retained in the scrubber likely affected the reflected light along the sides of the cell, causing anomalous ozone readings. Even though flows were greater than 1 L/min, the instrument took 15 min or more to recover. In ATom, with the newer model 211 instrument and moisture exchangers for both scrubbed and unscrubbed air, the agreement was much closer over a similar flight track, and there were no anomalous data segments as the DC-8 ascended and descended (Fig. 9, bottom). The one discrepancy is in the tropical marine boundary layer (MBL), where UCATS was typically a few parts per billion higher than the chemiluminescence instrument. This disagreement is outside the combined uncertainties of the two instruments and is not currently understood; UCATS showed no differences when calibrated with wet or dry air in the laboratory, and the effect of water vapor on the chemiluminescence instrument has recently been re-checked. There were no offsets in the high-latitude MBL (agreement within 1 ppb; see Fig. A7), so it is presumably related to the high humidity or something else present in the tropical MBL.

Scatter plots of UCATS ozone data against the corresponding instrument from NOAA CSL (Fig. 10) showed reasonable overall agreement in HIPPO, with the slope within 1 % of unity. But in HIPPO, there are many UCATS data points in the troposphere with significantly higher ozone than measured by the NOAA classic ozone instrument (e.g., near 50 ppb). The improvement between HIPPO (top) and ATom (bottom) is dramatic. This is partly due to the longer optical path length in the model 211, as well as other instrumental improvements, but the addition of the Nafion moisture exchangers makes a substantial difference as the aircraft transitions between wet and dry air masses. It should be noted that the NOAA CSL instruments being compared to here have completely different designs – the classic ozone instrument is a UV photometer like the 2B, while in ATom, ozone was detected by chemiluminescence, though it is fundamentally calibrated using an optical measurement. Both of these instruments have a precision in the troposphere of about ±0.5 ppb. The larger deviations occasionally observed in the ATom data are mostly due to timing mismatches during flight segments with sharp gradients in ozone, along with occasional outliers from all instruments (see Fig. A6). The older model 205 was also flown during ATom as a backup and for comparison with the model 211; the model 205 showed some of the same deviations between wet and dry air as in HIPPO, while the model 211 with Nafion moisture exchangers tracked the CSL instrument closely. For many applications, such as climatologies, chemical modeling, and transport studies in the troposphere, the precision of the model 211 (±0.5 ppb at sea level) as flown during ATom is more than adequate, given the good overall agreement with the chemiluminescence instrument. In the stratospheric parts of the ATom flights, the model 211 instrument had precision of about ±1 % and agreement within 2 % (not shown).

The original MayComm TDL instrument was used in UCATS from 2006 to the first ATom deployment in July–August 2016. Its uncertainty was the sum of 5 % + 1 ppm, based on laboratory calibrations with gravimetrically prepared standards and frost point hygrometers. The ±1 ppm precision limit made stratospheric measurements above ∼ 16 km (where water vapor is typically 2–8 ppm) somewhat qualitative compared to the troposphere and lower stratosphere (H2O > 10 ppm). In addition, there was a temperature effect on the electronics of the long-path channel (for low water vapor) such that the sensitivity dropped with increasing temperature inside UCATS, up to 30 % in extreme cases. This was addressed by adding a Peltier cooling circuit to the TDL electronics box during HIPPO, which kept it at 25 ∘C (except on occasional cold or very warm takeoffs), and also by calibrating and correcting for the temperature effect.

For ATom (starting with deployment 2, January–February 2017), we integrated a new, larger TDL instrument from Port City Instruments, the successor to MayComm. The longer wavelength (2.574 µm vs. 1.37 µm) utilizes stronger absorption lines for a precision of ±0.1 ppm or better in the stratosphere; it can also measure up to 40 000 ppm water vapor, higher than the maximum in the tropical MBL. Similar to the previous instrument, the large dynamic range was achieved by using two optical paths in the cell, strong and weak absorption lines, and different measurement techniques as described in Sect. 2.3. The data were found to have minimal or at least much less sensitivity to instrument temperature compared to the earlier version. For calibrations up to 200 ppm, we used ultra-pure air and gravimetrically prepared standards (Brewer et al., 2020). We also calibrated the instrument over the full range of water vapor mixing ratios and pressures found in the troposphere and lower stratosphere (near 0 to ∼ 30 000 ppm and ∼ 100–1000 mbar) with a bubbler and a frost point hygrometer (MBW, model 373LX). Air from the bubbler (or the standards) was passed through the TDL cell and then to the MBW, as well as to both instruments in parallel. Illustrative plots and in-flight comparisons for both instruments are shown in Fig. 11. Calibrations of the TDL are ongoing, as disagreements with the DLH instrument were observed in ATom, particularly at very high water vapor mixing ratios, > 20 000 ppm, where the TDL data seem to be anomalously high (not shown).