Summary: The authors estimate the maximum observability of persistent linear contrails in current geostationary satellite imagery by applying a contrail detection algorithm similar to Mannstein et al. (1999) to synthetic thermal infrared images of contrails within an otherwise clear sky over a homogeneous background. By varying several parameters related to contrail observability singly, they determine how contrail optical properties and sensor resolution can affect the contrail observability. Although only 46 percent of the contrails produced in CoCiP model simulations are expected to be detectable in current 2-km imagery, the detectable contrails would represent nearly 80 percent of the net contrail radiative forcing. The authors also find that the lifetime-integrated LW radiative forcing increases nearly linearly with the observable lifetime of the simulated contrails, implying that to the first order, contrail longevity in satellite imagery is a function of the optical depth of the contrails.

General comments: The paper covers an interesting and important aspect of the satellite remote sensing of contrails. Accurate estimates of the detectability of contrails from geostationary satellites are needed, and this study provides a straightforward way to establish the upper bounds of contrail observability. Although the study’s methodology and conclusions appear to be sound, the paper is weakened by its poor organization. The reader is presented with many details throughout the paper that are often presented in a confusing manner. The manuscript would be strengthened by rearranging the exposition in a more comprehensible manner. The authors present some excellent results so I recommend publication after a complete reorganization of the text.

Specific comments:

Line 106 “synthetic” is misspelled.

Line 305:  I suggest for better writing style, do not begin a sentence with a number. 

Line 316: “diurnal” is misspelled.

Introduction: All of the discussion in the manuscript pertains to linear contrails only, but this is not acknowledged in the paper.

It is not clear why the first 10 - 30 min of the contrail lifetime when contrails are not observed would affect the overall contrail cirrus radiative forcing significantly. If a contrail is detected, the extra forcing from the pre-observable contrail could be approximated and added to the overall lifetime radiative forcing estimate. The contrails are thin and narrow at this time so the overall radiative impact is small, as suggested by the results in Fig 9(c) and 9(d).

Section 2: This section contains many details about the contrails, but nothing about the surface conditions.

Do the surface conditions change? What surface temperature is used in the calculations? Also, there is no information here about the viewing zenith angles and solar zenith angles used in the calculations.

The discussion of the background cirrus layer is out of place and would be better located in section 4.1.

Why is the effective radius of the background cirrus layer only 5 microns? Most cirrus layers have much larger mean particle sizes. For example, Wang et al (2019) and Yi et al. (2017) report a global mean of ice cloud effective radius around 30 microns. An effective radius of 5 microns is odd for a midlatitude cirrus layer and a more realistic value may affect the observability results in section 4.1. 

Figure 5: This figure is difficult for the reader to follow. Many details are missing or unclear, both in the figure caption and in the accompanying text. The observability test results for 2 km wide contrails are presented as  three separate subplots, but the results are condensed into one subplot for the other two contrail widths. The caption also describes parts (d) and (e) as histograms, rather than plots of contrail observability. Parts (b) and (c) are not clearly explained. What do the values in part (b) represent? The fraction of CoCIP contrails of a particular effective radius and IWP that would be detected? What do the values in part (c) represent? The fraction of contrail segments with a positive or negative contrail radiative forcing (which is positive (LW forcing?) and which is negative (SW forcing?)? Why do 2-km and 10-km wide contrails produce only positive forcing, but the 0.2-km wide contrails produce both positive and negative forcing? The results appear to indicate that only contrails with net positive forcing would be detectable by GOES imagery, except a few (?) optically thick and very narrow contrails. This surprising result would be an important outcome of this research, and should be discussed more.

The discussion in lines 271 through 277 about the adjusted observability threshold and the ‘cause of unobservability’ is confusing and doesn’t add much information. Perhaps it can be removed?

Section 3.3

Line 294: population 2 (Fig 8). The authors have several references like this one, in which figures and results not yet presented in the text are mentioned, confusing the reader and weakening the exposition of the research. Please remove these references or place them at more appropriate locations in the text.

Lines 309 through 310: What is the ratio between contrails with strong (what constitutes strong forcing?) LW forcing compared to those with strong SW forcing? This section once again hints that only unusually thick yet narrow contrails have an overall negative (SW greater than LW) forcing. Is this because only small solar zenith angles are used in the radiative transfer calculations? I would expect SW forcing to dominate when the sun is low.

Section 4.1

Lines 314 through 315: The descriptions of Pobs-derived threshold uncertainty and seasonal and diurnal variability should be placed here, not in sections 2.2 and 2.3. The populations 1 and 2 are labeled inconsistently throughout the paper (population 2 versus CoCiP population, instantaneous population versus population 1). There is also a baseline population, which appears to be different from populations 1 and 2? Please use only one set of descriptors for the contrail populations, otherwise it is very confusing.

Lines 323 through 329: The Mannstein et al (1999) algorithm was developed for one AVHRR satellite, which has different resolution, sensor sensitivity, and noise characteristics than the GOES imagers described here, and the threshold “value identified by Mannstein et al (1999)” cannot be applied to other AVHRR imagers, let alone the GOES imagers. The detection algorithm can be adjusted to detect more contrails, but given the differences between the AVHRR and GOES imagery performance, the numbers presented here are speculative, especially without any information about false positive detections.

Lines 339 through 340: “Wider contrails with similar optical thickness are detected, because less-optically-thick parts of the contrail exist which remain detectable.” I have no idea what the authors are trying to say here.

Section 4.2

Fig 9 is very complicated but the description in the text is not always helpful. As a result, it is not clear, for example, what a ‘post-observable’ contrail segment is, what LW RF/length and what LW RF/length/segment mean. The comment on lines 372-373 that narrow and optically thin contrails must be “avoided “ in contrail mitigation trails is so vague as to be meaningless. Should the production of narrow and optically thin contrails be avoided, or is it the counting of such contrails in an assessment of the mitigation trial that should be avoided? Revise the figure caption and accompanying text so that the results of the figure are more understandable.

Line 375 - The pre-observable population in the forcing-weighted case (Fig 9(c) and 9(d)) appear to be only a few percent of the fraction of contrails for the 2-km resolution case, and much less for the 0.5-km resolution case, How can the authors claim they play an important radiative role (especially for the 0.5-km case)?

Lines 382-383: “As a check for consistency with the analysis of Fig. 7, the proportion of observable contrails has been integrated with the corresponding total.” The corresponding total of what? I’m assuming we are looking at all of the population 2 contrails here (once again called something different [time-evolving contrails] in the Fig 10 caption)?

Line 397: To confirm, does the (21 +73 -11) min symbol mean the average first detection was at 21 min, while the minimum time for the first detection was 21 - 11 = 10 min, and the maximum time for the first detection was 21 + 73 = 94 min?  

Fig 11: This plot seems to imply that the longest-lived contrails, that is the ones detectable in the GOES imagery the longest, are simply the most optically thick? This is another important result that needs to be highlighted in the conclusions.

Section 5.2  The author’s discussion here seems contradictory at times. The authors state that detection probability is not just a function of contrail optical depth, but earlier in the paper they characterized the cause of unobservability in terms of either “too narrow” or “too optically thin”. An argument can be made from the results in Fig 5(b) that the effective radii of most contrail segments fall between 2 and 10 microns, and that from Fig 5(a) between 2 and 10 microns the detection probability threshold follows the line of constant optical thickness (I estimate the thickness to be about 0.05-0.06). I don’t disagree that particle size does matter, but to the first order, optical thickness appears to be the most important factor determining contrail observability for most of the CoCiP contrails used here.

References

Mannstein, H., Meyer, R., and Wendling, P.: Operational Detection of Contrails from NOAA-AVHRR-data, International Journal of Remote Sensing, 20, 1641–1660, https://doi.org/10.1080/014311699212650, 1999.

Wang, Y., Yang, P., Hioki, S., King, M. D., Baum, B. A., Di Girolamo, L., & Fu, D. (2019). Ice cloud optical thickness, effective radius, and ice water path inferred from fused MISR and MODIS measurements based on a pixel-level optimal ice particle roughness model. Journal of Geophysical Research: Atmospheres, 124, 12126–12140. https://doi.org/10.1029/2019JD030457 

Yi, B., A. D. Rapp, P. Yang, B. A. Baum, and M. D. King (2017), A comparison of Aqua MODIS ice and liquid water cloud physical and optical properties between collection 6 and collection 5.1: Pixel-to-pixel comparisons, J. Geophys. Res. Atmos., 122, 4528–4549, doi:10.1002/2016JD025586.

General comments

In their manuscript “Factors limiting contrail detection in satellite imagery”, the authors investigate the principal limits of contrail observability using current generation geostationary satellite imagers under ideal conditions. A global dataset of CoCiP-modelled contrails was used to derive typical contrail properties at different stages of their lifecycle. Based on these properties, radiative transfer calculations were carried out to create synthetic satellite images corresponding to different contrails and different satellite imager resolutions. Finally, a simple contrail detection algorithm focusing mainly on line-detection was applied to those images to test the contrail observability. Using this setup, various sensitivity studies with respect to the contrail properties were conducted, indicating the main parameters, and the contrail observability at different stages of the contrail lifecycle was derived.

Both the study and the manuscript are well structured and written. Motivation, study approach results, conclusions and limitations were mostly described clearly. The study gives important insights in the limitations of contrail observability which are useful, e.g., for the satellite-based study of contrails, the validation of contrail models as well as the evaluation of live trials focusing on operational contrail avoidance.

Thus, the results are very important and the manuscript is recommended for publication after minor revisions.

Specific comments

L 1: Refine definition, e.g., “Contrails (ice clouds, originally line-shaped, initiated by aircraft exhaust) …”

L 4: Stress the high temporal resolution of geostationary imagers, “time-resolved” can also mean once a day.

L 10: Maybe replace “instruments” by “satellite imagers”

L 12-14: Which resolution are you considering? I think if you continue to increase the resolution, you should eventually be able to observe all contrails. Could you specify your assumptions here?

L 17: Maybe replace “observation requirements” by “observation conditions”

L 39: Explain why “human labellers” are considered here (i.e., mention human-labeled training data)

L 41: Compare also Geraedts et al. (10.1088/2515-7620/ad11ab, 2024), stating “Most flight segments start matching contrails about half an hour after formation, with the mean time until first observation being 41 minutes”

L 43-48: Maybe restructure section instead of jumping from Kärcher et al. (2009) to other papers and then back to Kärcher et al. (2009).

L 43: Which model/simulation has been considered? Maybe note that there might be significant uncertainties in the contrail modelling as well.

L 48-49: Consider reformulating the statement “depends not only on the optical thickness of a contrail, but also the microphysical properties”, as the optical thickness depends also on the microphysical properties, so these are not distinct properties.

L 43-54: Check whether you can restructure this section to make it easier to follow. E.g., you mention first “Kärcher et al. (2009) established that simulated and observed distributions of contrail optical thicknesses differ, and can be reconciled using optical-thickness-dependent detection efficiencies.” Then CNNs are mentioned. And then some sentences later you introduce the concept again, writing “If these satellite-observed contrails are to be used to evaluate model simulations on contrails, it is essential that the properties of the observing system are taken into account”

L 56: Maybe better: “CoCiP produces predictions which generally align with observations regarding the order of magnitude and principle age-dependencies for micro- and macrophysical properties”

L 56-60: Four sentences starting with “CoCiP…”. Try to reformulate due to style reasons.

L 59-60: Note that CoCiP does not check for persistence, see Schumann (2012): “An explicit criterion for persistency as a function of supersaturation is not necessary in CoCiP. In case of very low temperatures, short-lived contrails may form from the emitted water vapor even in totally dry air.” I think the pycontrails implementation works similarly.

L 66: Specify “tactical avoidance application”

L 73-74: Suggest to reformulate: “Contrail detectability is tested in otherwise-clear-sky synthetic satellite images by applying a contrail detection algorithm.”

L 74: Where does this “baseline contrail population” come from? Is it modelled?

L 83: What means “radiative importance conclusions”?

Fig. 1: Maybe reformulate: “Schematic of the process for deriving the contrail detection efficiency  by application of a contrail detection algorithm to synthetic satellite observations of a single contrail, using a specific imager and contrail detection algorithm, and a pre-calculated radiative transfer lookup table.”

Fig. 2: What denotes the title for the three panels (0.5 km, 1 km, 2km)? Are the simulated calibration error and the NEdT given somewhere in the plot? Which brightness temperature channel is shown?

L 86: You maybe want to check with Schumann et al. (10.1175/JAMC-D-11-0242.1, 2012) and Wolf et al. (10.5194/acp-23-14003-2023, 2023), both describing comprehensive radiative transfer calculations for contrails. You might want to mention here or elsewhere in the manuscript how your assumptions agree or disagree with those studies.

L 93-101: What is the setting for the ice water path IWP_0? What parameter space is covered for r_eff? Are slanted observations considered, i.e., non-zero viewing zenith angles?

L 145: Maybe only a language issue, but what is the “Mannstein et al. (1999) style detector”?

L 175: I suggest to say that it is “assumed” that the CoCiP-based statistics aligns with reality. More validation is necessary at this point.

L 181: Suggest to speak of “contrail segments” instead of waypoints.

L 196: Please clarify which “globally-consistent thresholds” are meant.

L 199: Maybe replace “clear” by “apparent”?

L 211: I assume some of these properties (e.g., width and depth) might be related? So a limitation of this approach is that some of the configurations you consider are not realistic?

L 217: For IWP, no upper detection limit is shown for the 2km resolution in Fig. 3b. Do you expect this limit to be at even higher IWPs? Why aren’t the limits for both imager resolutions at the same value?

L 223: It seems like the observable range in Fig. 3c can maybe related to the imager resolution? Maybe from 1/5 to 5 times the imager resolution? A consequence would be that for the newest generation of imagers some kind of downsampling is necessary to detect broader contrails, right?

Fig. 7: The “seasonal and diurnal variability” complete coincides with the “combined uncertainty”, right? However, I think the “seasonal and diurnal variability” is hardly visible right now. I suggest to plot this differently.

L 354: Why was the proportion of LW forcing and not the net forcing considered? The latter is the crucial quantity.

L 445: The Mannstein algorithm checks only for linearity and is applicable to detect contrails in the early phase of the lifecycle, where this linearity is present. In principle, detectability might be increased by combining different detection methods or apply tracking procedures to observe contrails also beyond their linear stage, e.g., Vazquez-Navarro et al. (2010, 10.5194/amt-3-1089-2010), Vazquez-Navarro et al. (2015, 10.5194/acp-15-8739-2015).

In the paper you considered imager resolutions of 0.5 and 2 km. MSG/SEVIRI used in previous studies had resolution of 3 km at nadir. Can you make any comments on its performance?

Technical corrections

L 53: “micro-“

L 207: “width” mentioned twice.

Fig. 4: “contrails as width” ?

L 285: “on” doubled

Fig. 9: “Perisisting” --> “Persisting”

L 336: Incomplete sentence?

L 389: “was also found is plotted” ?

L 418: “dominated by soot while current fuels” ?