Testing a New “Decrypted” Algorithm for Plantower Sensors Measuring PM2.5: Comparison with an Alternative Algorithm
Round 1
Reviewer 1 Report
The topic of the research performed in this paper is at the front of the current research efforts.
The results presented have the potential for publication in this journal, but several improvements are necessary, namely:
1. Earlier experimental efforts employing other measurement techniques must be briefly discussed, such as:
"An observational study of the atmospheric ultra-fine particle dynamics." Atmospheric Environment 59 (2012): 312-319.
"Using satellite data to estimate particulate air quality in a subtropical city: an evaluation of accuracy and sampling issues." Remote Sensing Letters 6.5 (2015): 370-379.
2. A very brief discussion on the contribution of this research to the success of the Sustainable Development Goals-UN 2030 agenda would be of readers benefit. See for example:
"Remote Sensing Letters contribution to the success of the Sustainable Development Goals-UN 2030 agenda." Remote Sensing Letters 11.8 (2020): 715-719. (https://doi.org/10.1080/2150704X.2020.1753338)
"Atmospheric and ecosystem big data providing key contributions in reaching United Nations’ sustainable development goals." Big Earth Data 5.3 (2021): 277-305.
By summarising the paper has the potential to be published, after taking into account the minor revisions mentioned above.
Minor editing is necessary
Author Response
Please see file
Author Response File: 
Author Response.docx
Reviewer 2 Report
Overview and General Impression
The presented manuscript (MS) is dedicated to the description of the performed in the study “decryption” of concrete-type low-cost sensor’s (LCS) algorithm for converting the aerosol particle number readings to particle number concentrations as well as of the comparison with an alternative algorithm.
With an acceptable ratio of price/performance, the LCS is an attractive possibility for independent researchers or groups of researchers to build their monitoring sites and even networks. However, the technical capabilities of this equipment should be evaluated in advance, in order to have an impression that it can meet the requirements. Hence the examined LCS Plantower mounted in PurpleAir monitor is essentially optical counter, it is crucial to apply proper conversation method (algorithm, from particle number to mass concentration) in order to obtain accurate estimation of the concentration. These algorithms are based on some assumptions (very constraining indeed) and are linear relations.
Most general, the work is well conceptualized and structured and, respectively, clearly written and illustrated. Alongside the straightforward and rigorous presentation of the results, I could outline as a certain strength the consideration for the performed tests of many datasets with relative big number of records in each of them.
The main merit of the work is, according to my opinion, the practical importance of the key messages from an end-user perspective which are clearly stated in the ‘Conclusion’. In the MS is expressed (but in the same time well argued!) strong criticism to the “proprietary” algorithms.
I agree completely with this opinion. The presence of such “black boxes” could be prerequisite for misleading results/interpretations with far reaching negative consequences. Recent results from other authors/other parts of the world are in overall concordance with this conclusion. The recent publication (https://doi.org/10.3390/s23146541) reports, although no so explicitly, for principal problems caused by “proprietary” algorithms/firmware.
The MS fits well in the thematic scope of ‘Algorithms’ and has a high potential to attract readers.
During the review, I have not detected any general flaws or principal caveats. Thus, I have not included in this text a ‘Major remarks’ section. I have only some tinny remarks.
Remarks:
- r9: PM 2.5 → ‘5’ as subscript.
- r18: ‘LOD’ in abstract appears cryptic – write it as limit of detection.
- r65: Add full stop after exclusively
- Tables 2 & 3: The units (of mass concentration/mass) of the coefficients are missing.
- r465: Insert blank after the cubic meter in the mass concentration unit.
Author Response
Please see file
Author Response File: Author Response.docx
Reviewer 3 Report
In this paper the author used data obtained from low-cost PM2.5 sensors to compare different available algorithms. The available algorithms have been previously compared but, in this study, comprehensive analysis is performed using data from two different sites measured in the period 2021-23. In addition to comparison, accuracy, and precision of the algorithms (sensors) have been discussed. It is stated that new CF_ATM algorithm has no physical basis and from that point its usage is questionable.
The main findings are interesting and might be of wider interest. However, from my point of view this is rather technical paper which is not carefully prepared. There are a lot of technical details and typing errors which should be considerably improved so the readers can easily follow the text. For example, introduction section is too long with many unnecessary details and descriptions of general low-cost sensors history and explanations (also in supplementary material). There is some redundant information, for example Figure 1 is redundant, the data is already presented in Table 3, thus it should be removed. The quality of almost all figures is not well and should be improved (font size, legends…). There are many typing errors (for example Line 9: PM2.5, subscript, Line 84: 1 g cm-3, ,inconsistency in indexing, bold sentences, colored tables etc.
Although related to very interesting and important research area, air quality and potential of low-cost monitors usage, I believe this paper should be rewritten focusing on scientific part – testing the results obtained. It is clear that some kind of reverse engineering has been done to provide information on algorithms used that are not available. This is the main problem with measurement devices - if the methodology is not completely described and available for additional testing such instruments cannot be used for scientific research.
For instance, the following paragraph
"Plantower had made no notice about the change, but when contacted by PurpleAir they did admit that a change had occurred. PurpleAir made the decision not to accept the “new” instruments, which could be distinguished from the “old” instruments by the tests that PurpleAir runs on all monitors before releasing them for sale. After some time, no further “new” instruments were received by PurpleAir. However, it is unclear whether some of these “new” instruments may still be available to the 10 or so companies that use Plantower sensors"
suggested that the usage of low-cost sensors is not reliable and they can be only used as black-boxes. If so, what is the scientific contribution of comparison different algorithms? The only valid tests could be intercomparison and validation with respect to referent methodology (devices).
I strongly recommend to rewrite the manuscript focusing on data analysis only.
Author Response
Pleaase see file
Author Response File: Author Response.docx
Round 2
Reviewer 3 Report
The paper is moderately improved, some technical issues are corrected, several explanations are added and previous comment addressed. I am not against its publishing
