This paper falls within the scope of AMT and should be published with minor revisions. The authors present a novel instrument of a 4 channel GCMS that can fit in the back of a Prius. The authors clearly lay out the instrument information as well as key calibration work to test  the validity of the instrument and ability to use all 4 channels together. The paper is laid out well and shows how this new instrument fits in with a history of mobile instruments and what instrumental gap it fills.

Below are some minor comments or points of confusion for this reviewer:

General comments:

The use of the “Prius wagon vehicle” is awkward language. Is there a better way to re-phrase this? Maybe some combination of a Prius wagon, a hybrid vehicle?

The ordering of figures and tables in the supplement doesn’t match the main text. It also doesn't seem specifically done in relation to text in the supplement either. 

Section specific:

Lines 593-597: Why did you not try overlapping sampling? You mention they are non-overlapping and that (line 593) “This loss of sensitivity could be mitigated by increasing the mass spectrometer dwell time with selected ion monitoring or by overlapping sample collections” with only (line 597) “However, a longer sampling time and volume might increase artifacts from breakthrough and ozone interferences.” listed as to why you didn’t try this. Was it tried but not shared? Where does this idea that there could be more interferences come from? Wouldn't that have been a good way to check the different channel sensitivities? Especially that the different correction factors were valid? 

Line 265: do you mean you did the Cook’s distance test multiple times until the threshold was met? Is that valid? “Measurements which had a Cook’s distance of greater than four were removed, recalculating this parameter without the excluded data until all points met this threshold.” The discussion here is unclear.

Figure 6 caption: 100 ppb Ozone relative to zero air, word ozone is missing

Lines 396-397: The volume math and logic was difficult to follow as written. How did you get from 5 min @ 30 mL/min = 150 to 1200 mL in 15 min, was it the same 30 mL/min rate? Or something different?

Section 3.5.1 is hard to follow in parts, in particular the minimum injected concentration. Was that done for only A1 and A2? Because it was more sensitive? It was also hard to tell at first the horizontal line through the symbol is what is represented and not a separate line. It looks more like an error bar through the symbol. It doesn’t seem like the ‘minimum injected concentration’ is referenced in the main body text. Perhaps that would help with any confusion.

Figure 9: Formatting looks odd but probably not the author’s fault, larger font on the legend would be nice

Section 3.7: How fast was the Prius going? What distance was covered per sample? An example for context would be nice. 

This manuscript describes a new compact, multi-channel gas chromatograph coupled to mass spectrometer designed to provide continuous data sampling with significantly shorter sampling periods (typical 10 min) the customary 1-hour with FID-based AutoGC systems. Since the system is also highly mobile and relatively low-power, it will fill a need for high-speciated and quantitated measurements of ambient air in urban and indoor environments. Authors demonstrate superior technical and analytical skill with the material presented. The complexity of the instrument leads to considerable work required to resolve differences between different channels, which the authors present in detail. This is a good manuscript and should be published after minor revisions to address the issues presented below.

General comments

Line 166. “the two signals is simple and sufficient” The analysis shown in the supplemental (e.g. Figure S3) does not strike me as simple, especially if confronted with a peak from an unidentified compound. I do wonder how well this correction would work when a large signal on one channel is measured simultaneously with a small signal on the other. Please provide some error or uncertainty analysis for this correction.

Section 2.7 Liquid calibration standards. I didn’t see any discussion of instrument zeros or blanks to challenge the system memory after measuring an ambient or calibration sample. Please provide some characterization of instrument zeros (apologies if this was somewhere in the supplemental and I just missed it).

Line 260. The discussion of compounded-dependent difference in detector response is important, as the ability to intercompare measurements between channels for uncalibrated species hinges upon this. This analysis relies upon ambient data, assuming that signals are relatively constant within a 30-min window. This section would benefit if the authors could use their calibrated compounds to validate this assumption (or identify optimal time periods for this analysis), along with the utility of using the quantitation ion as the independent variable. Does the quant ion correlate with volatility? Regardless, this section requires some uncertainty analysis.

Discussion in sections 3.3 and 3.4 are well presented. One aspect of the instrument that I would have liked to have seen here is some analysis or discussion of the aging of the sample traps, specifically regarding artifact and breakthrough volumes. How often must the sample traps be replaced during ambient use? If the authors can bring in information from Wernis et al. 2021, that would be fine here.

I enjoyed the data discussion in Section 3.6, which showed a nice use of this new instrument (although see comment below regarding Figure 10). Was there any co-sampling by high-time response instrument, e.g. CIMS, CO/CO2) that could be compared with the GC data to further demonstrate the value of the continuous data measured by the GC?

In Section 4.1, the authors discuss future improvements for the instrument, informed by the lessons learned from their first deployments. I do wonder If the stated goal of further increases sampling rate, by decreasing sample volume collected, or extending the volatility range of the GC via higher column temperature, is the best path considering the relatively high LODs for this instrument.

Section 4.2. Limitations. During this discussion of the effort required by the authors to merge the data collected on RGA A and RGA B, I had hoped that they would discuss whether the large difference in sensitivity here was inherent to the design of the paired RGAs, or just a happenstance of their detector. Did they discuss this issue with the detector manufacturer? Have they considered other means to cross-calibrate (e.g. adding a valve to systematically switch the detector that the column effluent was directed towards)?

Specific comments

Line 89. Specifying the size of the helium cylinder isn’t needed. You could state the regulator delivery pressure if that is relevant.

Line 92. “helical vibration isolators” I believe the standard name for these is “wire-rope vibration isolators”.

Line 125. Can you describe the GC sample pump used?

Line 126. “collector is purged”. Please specify the direction of purge (forward- or reverse-purged). Also, can you provide the purge flow rate rather than the flush gas pressure?

Line 128. “maintained at 30 °C” Please provide a range (1ơ) for this temperature.

Line 151. “by transfer lines” Please specify the material of these lines, and if they are temperature-regulated before entering the vacuum chamber.

Line 155. “total scan time of 553 ms” Does this provide a sufficient number of data points to describe the chromatographic peaks with FWHM < 3 sec?

Line 160. “ionization sources” Please provide a brief description of this – filament-type? Emission current?

Line 164. “A plate separates” What kind of plate? Does it have any voltage potential to reduce crosstalk?

Line 164. “above a threshold” Can you put some number to this threshold, preferably in ambient mixing ratio.

Line 226. “A seven-point calibration curve was measured with a replicate at all concentration levels” While commendable, this strikes me as particularly labor-intensive and impractical for an extended deployment (e.g. continuous operation). Can you describe the time required for this (roughly 2.5 hours, I think)?

Line 234. “the daily correction factor.” How much variance exists with this correction factor? Can you provide some range for this, or a time series?

Line 248. “both linear and power fits were investigated here.” Please provide some rationale (i.e. physical explanation) for why the instrument would have a power function sensitivity curve. Consider using only a linear calibration curve without this explanation.

Line 274. “The inlet tubing (1/8” diameter, 2.2 m length, PFA)”. Is that 1/8” diameter refer to ID or OD?

Line 309. “Carboxylic acids: Acetic, propanoic, and benzoic acids were all detected during the indoor field study.” Has any work been done with this compound class to determine if they are measured quantitatively? Carboxylic acids are notoriously difficult to pass through GC systems measuring ambient air.

Line 328. “less than 3 s for most compounds with retention times greater than 900.” With the detector operating at nominal 2 Hz (line 156), are enough data points collected per peak to sufficiently define the peak shape?

Line 334. “3.2 Precision of GC temperature control and retention time.” While technically impressive, this discussion strikes me as something better suited for the supplemental. I’d suggest moving it there.

Line 432. “the limit of quantification (LOQ) was calculated” I’m curious why the authors used the standard deviation of the fitted line intercept here, but used standard deviation of the baseline for LOD. Could they provide a brief rationale (or reference) for this in the text?

Line 439. Figure 8. I found this figure very difficult to pull information from, especially with the small y-axis log-scaled. Could the authors consider converting this figure into a table for an easier understanding of the underlying data?

Line 457. “though a power fit is more representative at lower concentrations for some compounds” Again, please provide some physical basis for the power fit.

Line 478. Figure 9. I found the image quality of Figure 9 to be poor, likely of lower resolution compared to other figures in the manuscript. Can the authors revise with a higher resolution version?

Figure 522. Figure 10. This is a very nice time series, but I was frustrated that no species were shown in units of mixing ratio. Can the authors provide a dual axes or stacked axes figure, presenting as many species as possible with units of mixing ratio rather than relative abundance? I feel using only relative abundance here diminishes the value of the work done here.

Line 542. Figure 11. Can markers be added to color trace to indicate start and stop location of each sample? The discontinuous data shown here is counter to the operational goal of the instrument (relatively high-time series continuous measurements). If this was due to instrumental issues (e.g. sample pump failure), consider making some brief note to that effect in the figure caption, which may be as far as some readers will go to parse what is shown here.