Review of ‘Quantification of major particulate matter species from a single filter type using infrared spectroscopy – Application to a large-scale monitoring network’ by Debus et al:

The work by Debus et al. presents a methodology for analysing the aerosol composition using FT-IR analysis of PTFE samples, along with proof-of-concept measurements on samples collected on the USA IMPROVE network. As indicated by the authors, this approach offers a number of advantages over more traditional reference methods for analysing organic and inorganic aerosol composition, being faster and cheaper, making it attractive for large scale monitoring networks.

In my opinion, the authors have done a very good job describing the methodology, from the FT-IR analyses to statistical models used for calibration and prediction, and I was able to follow it even as someone who is non-expert in FT-IR. The reported results demonstrate that the FT-IR was able to predict total mass, organic (OC), elemental (EC) and total (TC) carbon, sulphate, and crustal elements (Si, Al, Ca, Ti, Fe) concentrations with a similar error to reference method. The exception was nitrate which demonstrated higher error, likely due to evaporation from the filter.

My one main comment is I would have liked to have seen more data on how the predicted composition by FT-IR compared to the reference data. My understanding is that 21 sites were used to build the statistical model required to extract the compositional data, and this was then applied to the remaining sites. Fig 4 is the main figure demonstrating how the predicted levels by FT-IR compared to reference for key species but appears to be for all samples and sites in 2015. What I was hoping to see (but may have missed) was similar data/figures demonstrating how the FT-IR predicted composition for individual sites, especially those sites not used to build the multilevel calibration model. This would help I think demonstrate that this method could be applied independently.

Overall, this paper is well written,  clearly presented and would be of interest to many in the community.

Minor comments

1. Line 337: Why was 2015 chosen for developing the model, when datasets from later years were available? Would using samples from later help with losses of semi-volatile species?
2. Figure 1: Is the composition data presented here measured by FT-IR or the reference methods?
3. Section 3.3.1: As there was good predictive capability for FT-IR for organic carbon, do the authors think that additional information on the organic aerosol could be extracted with FT-IR, perhaps related to the functional groups present?

Overall, this paper clearly presents a novel solution to increase the efficiency of IMPROVE measurements using FTIR. The results are comprehensive, as well as surprisingly good and consistent, with a few relatively small exceptions. In addition to the method comparison, there is a brief discussion characterizing the different sites with their regional similarities, and a slightly awkward overview map showing the CONUS concentrations, but with problematic nitrate.

The abstract omits that measurements of nitrate on the PTFE filter cannot adequately access particulate nitrate in the atmosphere. This is a pretty limiting feature and should be highlighted in the abstract, even if it has been reported previously, since it has significant impacts for this method.

The method seems sound in terms of separating training and testing sets, but there are a few aspects that should be shown in more detail:

1) the outlier results should be shown in supplement.

2) the comparison before the calibration to the limited sites (at least include in SI).

3) the statistics in the summary table should also be given including outliers and without biomass burning corrections, i.e. an untrained, uncorrected, full-dataset comparison.

If these things can be added, the paper would present a much more complete evaluation of the method for future potential users.

Specific Comments

Nitrate is one of the two most abundant inorganic anions that is quantified in the network (if not the most abundant component overall: see Fresno, Ebgert sites) yet determining nitrate from PTFE has a large uncertainty due to volatilization of nitrate on the filter. It was suggested that nylon filters that are analyzed by IC can be used as a reference method but how would this affect costs since this paper promotes FTIR+PTFE as a cost effective single-filter/single-technique. References to papers that successfully performed these alternative techniques for quantifying nitrate would be good to include (for quantifying ammonium nitrate: https://www.tandfonline.com/doi/full/10.1080/02786820701272038).