Please identify the top weaknesses (up to 3) of the Rolsky and Kelkar (2021) paper

1) The conclusions are often very speculative, such as the ability of the PVA to mobilize heavy metals or pharmaceuticals, and cause their enrichment in the food chain.
2) The extrapolation is very generic, as WWTP plant operations vary tremendously across the US, and this is not captured in the current Figure 3 panels.
3) There seems to be a math problem, 34 billion gallons daily across 16,000 WWTPs yields 2.12 million gallons daily per facility. The text claim 2.21 billions daily per facility.

Predictions of a model are only as good as its parameters

1) Data used to estimate anaerobic and aerobic biodegradation in digesters and aeration tanks, respectively, are not very useful (blends tested, often not clear what polymer was tested, test periods too short (e.g 100 hours when testing anaerobic degradation), no continuously-fed activated sludge test (experiment), no regulatory tests, etc. Recently excellent data obtained with OECD tests became available (Byrne et al 2021).
2) Little understanding of wastewater treatment systems hampers data interpretation. For instance the following stated in the paper is not correct: "¨The average hydraulic retention time (HRT) in the ASP is approximately 18–24 h, and the sludge retention time (SRT) is 12–15 days [42]. Due to the hydrophilic nature of PVA, the majority of PVA is expected to be in the water phase, in which the HRT would subsequently play a larger role in its degradation.¨ The HRT is only indirectly important in activated sludge systems because the HRT influences 1) the load and thereby the F/M ratio and most importantly the SRT and 2) the operation of the settling tanks which might result in varying sludge concentrations and thereby the SRT. Primarily SRT and growth rate (doubling time) of competent microorganisms (those degrading PVA) determine the removal of a specific substance water-soluble and water-insoluble alike. When the doubling time of the competent microorganisms is higher than the SRT no degradation will occur. In other qords, the competetnt microorganisms will be washed out of the reactor. Schonberger et al does contain most likely the most useful data as for instance monitoring data of a full-scale plant. This is little valued by the authors.
The rate of biodegradation is inversely correlated to the degree of adsorption (Chielini et al 2006). This is not clearly explained whereas this might explain some of the test results showing little degradation.

1. The model contains several sources of potential bias/confounding. An online survey with 527 respondents is used to estimate LDP use across the entire US, which is not sufficiently robust; I estimate at least 10,000-20,000 respondents are needed for any decent statistical power. Online surveys could have selection bias towards a demographic more likely to use LDPs. The authors assume each WWTP in the US has the same capacity. While such a simplification is understandable, it will likely lead to overestimation of effects in populous states (which are likely to have a higher proportion of very large WWTPs) and underestimate effects in rural states (which are likely to have a relatively high number of relatively small WWTPs). The different sources of bias/confounding could reinforce total bias/confounding in some states, depending on the combination between population density and socioeconomic make-up.

2. There is high variability in the dry weight recorded for LDPs, especially laundry pods (1.0 +/- 0.6 grams). Pods from three laundry and two dish pod brands were evaluated, which seems like a very small sample size (especially considering the high variability in recorded dry weight). While the PVA output estimates contain an error margin, this margin is much smaller than the error margin for the dried pods. It is not clear how the error margin for PVA output estimates is calculated, but it appears that the error margin in the dry weight determination was disregarded or underestimated.

3. The authors state that adaptation of activated sludge requires a long lag-time and heavy influx from textile industries. The last part of that statement is quite perplexing, I truly wonder what kind of research is required to conclude that heavy influx from textile industries is an absolute requirement to obtain adapted micro-organisms. More importantly though, the statement of the authors about long lag-times is not in agreement with the generally accepted phenomenon that continuous influx at any relevant concentration would lead to adaptation of activated sludge. This concept is actually supported by the text in Chapter 4 of the manuscript itself, where 10 % of total PVA is captured in return activated sludge, which means that even with semi-continuous influx the sludge will be continuously exposed to relevant concentrations of PVA.

1. The most important weakness of the article is the section on "Online Survey". The weakness is sufficiently serious that it jeopardizes the usefulness of the results from that section and how it is integrated into the other methods. The sampling of the population surveyed is not described, and there could have been selection biases introduced. The overall response rate was not discussed. The survey questions (in the supplemental section) were not piloted or vetted. Critically, there is no evidence that the authors sought institutional review board (IRB) approval for the study. It is surprising that the journal editors allowed the manuscript review to proceed with the following designation "Institutional Review Board Statement: Not applicable" as stated on page 13. Studies with human participants have to be reviewed and approved, even if it is designated as "exempt".

2. The second weakness is that the study was funded by "Blueland" which is a company that sells products that are not packaged with PVA, in direct competition with companies that utilize PVA packaging of detergents. In fact, Blueland advertises their products specifically on this point. We do not learn this fact until the end on page 13, and a reader will have to go online to find out what "Blueland" is and why the company is funding this research even though they claim not to use PVA. Shockingly, on page 13, the authors declare two seemingly contradictory statements:

Acknowledgments: This work was supported in part by Blueland. Their sponsorship and critical input greatly assisted with the study. We thank Shireen Dooling for her assistance with graphic design.

Conflicts of Interest: The funders had no role in the design of the study; collection, analyses, or interpretation of data; writing of the manuscript; or decision to publish the results.

The reader is left to wonder what constituted "critical input" under the acknowledgments section.

3. A third weakness is the section on filtration, and comparison with pharmaceutical compounds with known log Kow values. The prediction of PVA behavior is not compelling. The accompanying Table is not well described (for example, a reader has to go find out what "RE (%)" means). The weakness in this section is an example of general weakness in the execution of the study and the conclusion of approximate proportions instead of conducting sensitivity analysis and better, more quantitative, clarification of uncertainties and assumptions.

1 - The first weakness is the parametrization of the model and the way in which in some cases the literature is interpreted. As an example: On page 7 - lines 6 and 7 - reference 36 is used to support the claim of PVA being 'likely' eliminated from the aqueous phase. Reference 36 has, however, nothing to do with PVA.
2 - Probably the top weakness is presented in sub-paragraph 3.2.1. where it is stated that adaptation of PVA-degrading organisms is likely to occur only in WWTPs receiving a heavy influx from textile industry. It is fully ignored that actually all WWTPs - except industrial ones - are receiving continuous inputs of large quantities of PVA. It is thus highly likely that adaptation of the microbial community to PVA degradation, has occurred in domestic WWTPs across the USA. The extent of biodegradation used to feed the model, might thus be an underestimation of the actual amount of PVA biodegraded in a WWTP.
3 - In general terms, the model suffers from validation on the basis of measured PVA levels in influent and effluent of WWTPs. Especially effluent concentrations are of relevance in this respect.

1. Lack of concrete evidence in supporting the research results especially on the amount and degree of degradation of PVA in treated waster water. Some of the data from the method cannot be found in the Results & Discussion section. For example, the data/graph/figure on the GIS and Mapping cannot be found. The outcome resulted from the Equations (7 & 9) might be inaccurate and not representative since the authors adopted the outdated data (i.e. 2015 state-wise population numbers from the USGS report) for calculation. Hence, the overall estimate obtained might be overestimated or underestimated. The responses obtained from the online survey seems vague. The authors mentioned "60% of the responses were from the top 20 designated market areas (DMAs)", however the authors did not specify what and why are these DMAs.

2. Reviewer did not find much novelty in this publication. This publication hardly advanced the understanding of PVA degradation and emissions in the USA as most of the data were derived from existing literature.

3. The method use in this study is quite brief and unclear. For example, Section 2.4. How the GIS and Mapping were performed? What are the parameters involve for the GIS and Mapping? What do you mean by collecting data from outside sources? Please be more specific. It was not clear how authors derived and processed some of their data. For example, the data source of US states and environmental emissions of PVA was not given and the numerical values of the fractions of PVA in laundry and dish pods in equation (8) were not given. Very unclear description of the research methodology. Each method and equation were not properly defined and explain. For example, how the authors calculate the mass balance in Fig 2? How the environmental release in Fig 4 was measured?

1. Consumption estimates appear to be based on a modest online survey (527 respondents); no data are given on how representative this sample was of the US population (just a sex ratio and age range). I wondered why industry production data were not used to estimate PVA use in these products – perhaps these are not readily available? They would provide a much better estimate of the amount of PVA used in the production of laundry and detergent pods in the USA.
2. There is limited attempt to account for the considerable uncertainty inherent in the gross extrapolation to reach a national estimate of PVA leakage. The basis of various error terms is poorly explained; indeed, even the meaning of the “±” term associated with most estimates is not explained. It is specified as being standard deviation in relation to the average retention in sand filters (Table 1), but not for any other error terms. I assume the authors mean them to be ‘plausible bounds’ for estimates, but it is likely that a more robust assessment would yield a much greater uncertainty around the final values, given the lack of confidence in parameters used throughout the model. That said, lack of certainty in the estimate should not detract from the broad message of the paper – the current best estimate is that a substantial amount of PVA is entering the environment in the USA alone. It is true that the amount inferred by Rolsky and Kelkar (2021) might overestimate actual leakage, but it could equally be a substantial underestimate.
3. The spatial analyses (state-by-state use and release patterns) appear to be based entirely on differences in human population size and waste water treatment practices (i.e. relative proportion of untreated vs treated waste water). It is not clear how the per capita state-by-state emissions are estimated (Fig. S2). The signal range shown in Fig. S2 (3.7-5.2 mtu/yr/100,000 people) is trivial relative to the uncertainty in other input parameters, and had I been asked to review the paper I would have suggested dropping this figure.

1) Figure 2 represents the mass balance of PVA in a default WWTP. Some basic errors are detected that can produce important changes in the estimations.
The most important error appears at the bottom outlet of the secondary clarifier. If 48 is the total output of concentrated sludge, the PVA distribution should be based on the flow rate, but it is not. The return activated sludge (RAS) flow is usually around 50% of influent flow, but the waste activated sludge (WAS) flow should normally be lower than 1% of the influent flow. This would lead to about 1/51 of the PVA in the WAS versus 50/51 in the RAS, because the PVA content will be based only on the respective flows, since the composition of the RAS and WAS is the same. These values are very different with respect the reported values (10 and 38).
Even if the sludge can adsorb and concentrate PVA on its surface, that gives the possibility for the activated sludge to degrade more PVA in the aeration tank. Hence if the 20% removal is maintained, the absolute removal will be higher because the concentration in the reactor will be higher (dissolved + adsorbed in the sludge).
This error also implies that the PVA finally in the effluent of the anerobic reactor will be lower and will depend mainly on the fraction separated in the primary clarifier.
Furthermore, the assumptions made on page 10 appear to be arbitrary and based on non-relevant works, as for example reference [50] that describes anaerobic treatment at 37 ºC reporting 20% degradation of PVA, but this value is used as reference for the aerobic conditions of an activated sludge system.
2) The equation for wastewater generation (Equation 2) considers 20% losses of wastewater in volume. This estimation appears to be based on data from reference [27], but this reference reports data from India, which does not necessarily share similar values to the USA. Moreover, the calculations of treated/untreated wastewater are incorrect. Wastewater generated (equation 2) is supposed to be 80% of the water supplied for public, domestic and industry use. Then, overall losses should always be the rest (20%). However, the authors estimate the wastewater treated based on the number of WWTPs per state and an average size USA-WWTP (equation 3). The size of cities of each state is not considered, which will have impact on the size of the WWTP. Then, using equation (4) for the calculation of untreated wastewater is only reflecting the error that is generated by considering an equal size for all WWTPs. A very clear example, considering that Los Angeles WWTPs are of the average size will imply the estimation of a huge amount of untreated wastewater being generated. Then, the information reported in Figure 3 appear to be useless.
Moreover, if 34 BGD of wastewater are treated daily by 16,000 US WWTPs, the average flowrate treated per plant should be 0.0021 BGD/WWTP, and not 2.12 BGD/WWTP, as reported in equation (2). In any case, the use of an average size per WWTP is an inaccurate generalization.

3) P7. Section 3.2.1. The works reported in this section do not appear to be representative of what is happening in an urban WWTP. There are no data on full-scale WWTPs with PVA input and output. The removal performance is a very rough estimate. It is my understanding that this information is not available and therefore the estimates are not accurate. I have tried to find examples of this type of operation in the literature, but they do not appear to be available. Long-term pilot or full-scale operation under relevant environmental conditions would be necessary to have reliable data. Long-term operation with significant input of PVA would be necessary to enrich the microbial community in PVA-degrading microorganisms. Otherwise, testing biodegradability in non-acclimated biomass would give unreliable results on what is the possible PVA removal in a full-scale WWTP. In any case, accurate estimations of the real emission of WWTPs in USA would require measurement of PVA concentrations in the effluent, developing reliable analytical methods if the concentrations are very low.
Two additional limitations: i) information of industrial use of PVA is not included, and its possible emissions by industrial WWTPs are not considered. The composition of industrial effluents can be very different to that of domestic wastewater. Ii) Wastewater considered to be treated in WWTPs is calculated based on water supply (Equation 1). If separate stormwater collectors are not used, it is expected that some of the wastewater treated will be mixed with rainwater before entering the WWTP, which could dilute PVA concentration.