Functional Kriging for Spatiotemporal Modeling of Nitrogen Dioxide in a Middle Eastern Megacity
Round 1
Reviewer 1 Report
The article was significantly improved and now it is clearer. However, the abstract is only addressing the methodology. It should be referring also to the context, topic, implications of the results and conclusions. The introductory part of the article should also be more comprehensive and more references to existing literature should be added.
Author Response
Response to Reviewer 1 Comments
The article was significantly improved and now it is clearer.
We appreciate very much your precious time, in-depth reading, and invaluable comments, which improved our manuscript.
Point 1: However, the abstract is only addressing the methodology. It should be referring also to the context, topic, implications of the results, and conclusions.
Response 1: We have revised the abstract accordingly. Now it presents the context, topic, implications of the results, and conclusions instead of just heavily focusing on the modeling aspects.
Point 2: The introductory part of the article should also be more comprehensive and more references to existing literature should be added.
Response 2: We have expanded the introduction part to highlight the important issue that our study addressed (spatial modeling of long-term nitrogen dioxide intra-day concentrations) and to highlight the importance of having this type of model in health studies of long-term exposure to air pollutants.
Author Response File: 
Author Response.docx
Reviewer 2 Report
The authors use a functional data analysis approach to model long-term diurnal curves of nitrogen dioxide in the Middle Eastern megacity of Tehran. They employed smoothing techniques to estimate annual average NO2 diurnal curves constructed of hourly means in each monitoring station. Then they used the trace-variogram to find out the structure of spatial auto-correlation between these curves. And they performed ordinary kriging for functional data (OKFD) to provide spatial predictions. Moreover, They used functional analysis of variance (fANOVA) to study the effect of population density on the shape and values of NO2 diurnal curves. The estimated mean of long-term NO2 diurnal curve, calculated across all monitoring stations, had two distinct minima and maxima.
It seems the paper is a revised version of their previous version. I do not have much comments since they considered the previous report on the paper but I think the abstract should be shorten and the verbs should be in present tense. The abstract in meta data also should be same. Please submit the clean file for the next review.
Author Response
Response to Reviewer 2 Comments
It seems the paper is a revised version of their previous version. I do not have much comments since they considered the previous report on the paper but:
We appreciate very much your precious time, in-depth reading, and invaluable comments, which improved our manuscript.
Point 1: I think the abstract should be shorten and the verbs should be in present tense.
Response 1: We thank the reviewer for pointing this out. Actually, we have revised the abstract completely to present the context, topic, implications of the results, and conclusions instead of just heavily focusing on the modeling aspects.
Point 2: The abstract in meta data also should be same.
Response 2: Thank you for pointing this out. We have checked again to have exactly the same metadata (title, abstract, etc) as the final manuscript file in the submission system.
Point 3: Please submit the clean file for the next review.
Response 3: Thank you for your comments. We have provided both a clean file and a track-changes file for this round of the review.
Author Response File: Author Response.docx
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
The paper presents an interesting model to predict NO2 pollution based mainly on population density patterns. Nevertheless, we consider that some metrological concerns have to be addressed by the authors.
- THERE ARE IMPORTANT FACTORS THAT ARE NOT INCLUDED IN THE MODEL. NO2 is one of the min air pollutants that mainly results in cities from traffic flow intensity and from industrial activities. The population density is not directly connected; therefore, a more detailed analysis of the actual sources of pollution is necessary. Also, meteorological factors have to be taken into account when assessing the NO2 concentrations. Both meteorological conditions (at microscale) and differentiated sources of pollution are definitely important for the model as they are also responsible for “heterogeneity in values and shapes of NO2 diurnal curves” (p.5)
- THE CHEMISTRY OF AIR POLLUTION IS NOT COMPREHENSIVELY ADDRESSED. The relation of NO2 to other pollutants has to be addressed as for example NO2 has an important role in the forming of tropospheric ozone, in acid depositions or in the chemistry of PANs. The reactions and transformation have to be taken into account when discussing the concentrations of NO2 at a specific place and time. The chemistry of NO2 is essential for the current approach. The NOx (that also include NO) are more relevant in this sense. First NO is produced by traffic and then it converses to NO2. Therefore, it would have been necessary to consider the interval from the moment of maximum pollution to the peak of NO2 concentration. What is the impact of lagged variables on the fit of your explanation model? (as the impact the ability to predict the level o pollution) (see Kamińska, J. A., Jiménez, F., Lucena-Sánchez, E., Sciavicco, G., & Turek, T. (2020). Lag Variables in Nitrogen Oxide Concentration Modelling: A Case Study in Wrocław, Poland. Atmosphere, 11(12). https://doi.org/10.3390/atmos11121293)
- SPATIAL DIFFERENCES MATTER. Although the model refers to Spatio-temporal modelling the spatial differentiations are not clearly illustrated. Mapping the air quality differences within Teheran would be important. Using kriging and interpolating the data by using GIS would be important in order to give a higher utility to the approach.
Other issues to be addressed:
- Affirmations that are not entirely true: E.g. “air pollution (…) produced by biological and human activities” -not only biologic activities but also other processes produce pollution (volcanoes, sand storms, marine water spray etc.) …You could replace with “natural processes”;
- Some typing/proofreading issues have to be addressed- a careful check is necessary.
Reviewer 2 Report
Authors present results of their work on application of functional Kriging in spatio-temporal modeling of nitrogen dioxide distribution. The work is relatively low quality, relies on regular monitoring and simple concluding for time and space variations of results that can be done with knowledge of bachelor level degree. Kriging is – as Authors also state – used in many research fields since decades. The work is of routine reporting character rather than new achievement.
English in the entire manuscript needs serious revision. Editorial changes also must be introduced e.g. what is bigger idea behind this strange using of capital letters?
Line 14: write names of chemicals with small letters in the entire work
Line 22: giving only one minute values for highest/lowest content of NO2 is exaggeration, at least 30 min ranges should be given
Line 32: pollution is not produced
Line 35: energy-generation industry
Line 45: what do You mean by off-road equipment? And what about e.g. shipping marine industry?
Line 65: this paragraph is totally detached form previous paragraphs and subsequent as well.
Line 147: what were arbitrary numbers to confirm density?
Line 171: the diurnal concentrations of NO2 do not oscillate based on fig 3 – based on results presented in fig 3 one can state that the are …
I cannot see anything in fig 4. And 5.
Results presented in fig 6 slightly differ from data presented in abstract and results section above – here NO2 content is related to density of population and not traffic so much.
Conclusions are too general and don’t present critical evaluation of results.
Reviewer 3 Report
In the paper the authors used functional kriging to model long-term diurnal curves of Nitrogen Dioxide (NO_2) in Tehran. They assessed the effect of population density on values and the shape of the predicted pollutant curves. They used the trace-variogram to find the structure of spatial auto-correlations, and the Kriging with Functional Response to spatially predict long-term NO_2 diurnal curves. Then they employed the Functional Analysis of Variance to compare predicted diurnal curves between areas with low, moderate, and high population densities. They highlighted an increase in the NO_2 concentrations after 10 p.m., which coincided with the time that local traffic regulation let commuting of heavy trucks.
The paper is interesting, and mathematically, to the best of my knowledge, is correct. They need to read the paper carefully and improve some minor mistakes. For example, "and" in the end of the author lists, etc. Also to better readability of math expressions, they have to put each math expression in a different line, for example, in line 110.
Generally speaking, the structure of the presentation of the paper needs to be improved before publication.
Reviewer 4 Report
The authors show a statistical approach for understanding the NO2 concentrations in Teheran. This approach is based on the application of the functional kriging. The paper is quite interesting and the steps are well-described. There are some problems: why did the authors apply this functional kriging to data collected in 2014-2015? and not used data more recent? Why did they analyze only NO2? with this approach it should very easy to also describe other gaseous pollutants. Is the approach different if particulate matter was analyzed? This is an important issue because PM is very high in Iran, precisely in Isfahan (as well as SO2, see Movassaghi et al., FEB, 2008, 17, 787-792).. Finally, the main question: why did the authors use the functional kriging? This method "gives the best linear unbiased prediction (BLUP) at unsampled locations": so, Teheran is covered by 33 monitoring stations! where are sites upsampled? I would have expected to read NO2 profiles in areas close Teheran but with not monitoring stations.
