Review for Yeti

Chan et al have presented a manuscript which introduces the Yeti model. Yeti is a collection of Python scripts to estimate vehicular emissions at street level. The model process the volume of traffic at different hours. The emission factors comes from the model HBEFA. HBEFA provides emission factors (g/km) depending on traffic situation, which represent a configuration type of road, level of congestion, environment and vehicular technology. To the best of my knowledge, there are not many open source models that estimate emissions using traffic situation emission factors. Hence, Yeti represent a valuable effort and an advance in this area. Nevertheless, the manuscript in its present form need to be modified in its current form due to major and minor issues:

Major issues:

The manuscript in not well written using too much ambiguity when describing concepts and explaining the model. Furthermore, there are no explanation to basic concepts such as traffic situation, or mentioning the PHEM model, which apparently generates the emission factors inside HBEFA. Also, the phrases in general are too long. The manuscript also presents lack of comparison with other estimates such as the % of cold start and evaporative emissions with other areas or cities. The expectation is that different estimates will converge to similar results, rather than produce equal  values.

Minor issues

Long and ambiguous phrases on lines (lines 34-37, 54-58 and more). Also, there are paragraphs consisting in only two sentences. Each paragraph should have at least three sentences.

Line 80. The authors mention that models WRF Chem and OpenFOAM which could be integrated with emissions models. Nevertheless, I would recommend the author to also mention the models MUNICH and CityChem, which have methods to estimate the dispersion of street emissions.

Sectiton 2, I do not see definition of traffic situation.

The definition of the emission factors as “epsilon” symbols is not defined.

Regarding equation 2, it is not clear for me the temporal resolution of the data. Later, the author mention that the traffic counts represent certain hourly intervals during the day. Then, what are the units of the emissions of equation 2.

Cold start emissions, in equation 3 have the parameter N which represents hourly traffic counts. Is N different with the parameter in Eq 2 that represent traffic counts?

The methodology for cold starts expressed in Equation 2, assumes that these emissions only occures in collectors and access roads. I understand that the authors are assuming that most of cold start emissions are in these type of roads, which are associated with higher residential areas where there are more cars. If this is the rationing, I think it is plausible and reasonable considering the usual limited input data when performing emissions inventories. However, this should be stated explicitly, then, future estimates can improve the current limitation. For instance, an possible future improvement could be associate type of roads with residential density and land use. In conclusion, authors need to be explicit and direct.

Line 130: “fuels and various volatile fluids”, what are various volatile fluids?.

Section 2.1.3, I liked it very much that you used an approach to porject diurnal evaporative emission factors to hourly resolution.

Running Losses evaporative emissions, equation 7, parameter xl. Are we assuming the value of 0.3 for xl?

Equation 8. As the non-exhaust emission factors represent both, ressuspension and wear emissions, it still not clear form me, is it the sum or the average?

Line 272, what is categorical? The type of vehicle?  Or a fraction that is associated with a vehicle category?

Section 3.3, do we have a different vehicular composition by hour?

Where are defined the level of services (LOS)?

What is the 5tpurpose of comparing emission factors from HBEFA 3.3 and 4.1?

Table 6, It is better to inform totals rather than averages. This is because the spatial distribution of traffic flow and emissions is not normal. There are few roads with most of emissions.

Table 6, it is also necessary to compare emission factors.

Line 321, t lower NOx emissions for 2020: “This is possible due to the introduction of diesel passenger vehicles with generally lower reported emission factors”. This statement implies that we need to a comparison of the emission factors is needed. Furthermore, Carslaw et al (2011) found that diesel vehicles with newer emissions standards emit higher emissions on real world. We also have dieselgate. Then again, a comparison of the emission factors is needed.

How important are cold start and evaproative emissions ?

Is it possible to include other methodologies in Yeti rather than solely depending on HBEFA? For instance, include emission factors and functions to consider  wear PM emissions separated from ressuspension?

I think it would be better if Figurse 3 and  4 represented the whole city ather than specific roads.

Lines 361-363, I think it is necessary to include a plot of the volumes for the city, with an appropriate color scheme to easlily identify the roads with highest volume.

Lines 390-394, as we can expect different vehicular by type of road, I do not think it is appropiate to compare emissions based on few roads.

Lines 400-404, what are aggregatted emission factors?

Lines 414-415, which pollutant?

Line 420, what is the purpose of providing normalized street emissions (total by road / length of road)? I strongly believe it should be better to inform simply the totals.

Identify the areas with more surface emission fluxes is important. This can be done by gridding the street emissions into a spatial polygon grid. The process must be mass conservative, emaning that the emissions inside each pixel, should be the same as the emissions of the road inside that pixel. Here the emissions can be represented as fluxes (mass  / area / time) or as emissions maps (mass / time). For reference, the EDGAR emission data can be obtained as emission fluxes (NetCDF) or mass/time (.txt). A standard recommendation for emission fluxes is mass/ hour / km2 with spatial distancing of 1 km.

@article{carslaw2011recent,  title={Recent evidence concerning higher NOx emissions from passenger cars and light duty vehicles},  author={Carslaw, David C and Beevers, Sean D and Tate, James E and Westmoreland, Emily J and Williams, Martin L},  journal={Atmospheric Environment},  volume={45},  number={39},  pages={7053--7063},  year={2011},  publisher={Elsevier}}

General comments:

You should define the targets for your model somewhere at the beginning. Then the reader gets an impression, what the targets for model development are. Shall emissions calculated by Yeti be used in air quality models? I assume so, then a high resolution makes sense. If one just wants to know the total emissions in Berlin, low resolution is sufficient and more cost efficient.

General remark: it seems you did a very good job in programming the tool. However, the paper reads in many passages rather like a detailed user guide than a scientific paper. Together with the annexes and the supplementing information the text is extremely long for a description, how HBEFA emission factors have been multiplied with the number of vehicles passing street sections. If this detailed description was intended, it is ok. My conclusion is, that I would find such a paper interesting only, if I would be just in the position to develop a similar software (fortunately, we have similar software ready in place since more than 20 yearsJ). I hope you had the target group of readers in mind, when writing the paper.

Some detailed comments and suggestions for improvements

Line 28 ff: your definition of top down and bottom um is not common sense. I see top down to use overall traffic activities (e.g. km/day in Berlin per vehicle category) with average emission factors. Bottom up uses traffic flow on street segment levels. Emission factors may be used or detailed vehicle simulation tools.

Why should the top down be “less sensitive to atmospheric transport processes and boundary conditions”? The calculated emissions do not depend on atmospheric transport. Please explain or adjust text.

Line 98: how is the emission factor e selected from HBEFA? Traffic situation and level of service are needed as input for each road section. Did you perform a manual attribution of traffic situations for all roads in Berlin?

Line 127: is there any analysis behind setting the cold start share to 0.3? Please explain.

In addition, the parking time is important for cold start extra emissions (zero extra emissions at short park time, full extra emissions after ca 6 to 8 hours). Please describe how you included parking time. Did you use the 8 hour parking values only?

Are deterioration factors for all exhaust components and ambient temperature corrections for hot NOx emissions from diesel cars considered? Same NOx emissions for winter and summer reported in chapter 4.1 suggest, that no temperature correction is included. Please specify possible simplifications.

Line 138, Equation 4: is the unit for the day to hour redistribution factor provided in B2correct? Changing g/day into g/h suggests a unit of [day/hour] you state day^-1

B2: please provide the units for “emission factor” and check units for “hourly emissions”, which should be [g/h] and not [g].

Line 235 ff: this reads rather like a user guide for Yeti than a scientific paper. Knowing that road category is named RoadCat in Yeti is not relevant for the majority of readers I guess. How you attribute the road category and hourly levels of service to all streets in Berlin on the other hand is more important but not described. You describe in chapter 3.1.3 input data, but do not mention any source for traffic flows, road categories etc.

Thank you very much for this interesting study. The manuscript presents a bottom-up modeling framework based on HBEFA emission factors to estimate emissions at the city level. In this case, as an example, the framework is applied to the city of Berlin in order to assess the emission results in comparison to previously published data for the city.

General comments to the text:

• The level of detail that the framework needs for the input information and the resolution it can achieve are not clearly defined.
• Currently the text reads very specific on file formats and model running processes. Suggestions to focus on the application of the framework and demonstrate the applicability are made.
• Some additional analyses are suggested to demonstrate the accuracy of the framework for all the modelled pollutants.

Main comments to the text:

The study claims that high temporal and spatial resolutions are needed (Page 1 line 24-27). Please specify what is meant by high spatial and temporal resolution in this study and justify how this aligns with existing literature (potentially with microscale emission modelling). For the spatial resolution, it would help to add the average length of the roads and the minimum and maximum lengths used in this study. For the temporal resolution, are the 1-hour intervals enough, have those been aggregated from more disaggregated information? Is that enough to capture traffic flow fluctuations that influence emission levels? Please specify and introduce in the text.

A clear scope of the proposed framework is needed. In the current manuscript, the Yeti framework seems to be mostly based on the use of HBEFA (an already existing model) so that the novelty and/or additional knowledge added is not clear to me. A clear description of the added value of Yeti would help to understand better the utility, capabilities, interactions with existing knowledge and potential limitations of the proposed framework.

Specific comments:

1. Page 2 line 29-30: Please specify which databases are meant when referring to “integration of vehicle-level emission factors estimated from existing databases”.
2. Page 2 line 49-51: Are COPERT, MOVES and HBEFA, considered in this study as high resolution emission models?
3. Page 2 line 54: Please be specific on the details and resolution the bottom-up approach can provide. In this case, specifying the resolution that can be achieved and the specific details that can be provided would help to understand better the capabilities and requirements of the bottom-up approach.
4. Page 3 line 65-72: Is Yeti a traffic emission inventory or a tool to calculate traffic emission inventories based on existing models and data? Please revise the text accordingly, in the current version it is not clear to me. Also, how flexible is the Yeti approach with the input information? Could it use second-by-second information and also with yearly averages? What are the limitations related to input information resolution. Since this framework seems to aim to work with varying levels of detail, did you analyze how the framework performed when using different levels of resolution for the input data?
5. Page 3 line 71: Please explain the concept of “process symmetry”, is not interpretable from the current text.
6. Page 3 line 77: What is meant by vehicle sectors and which specific ones have been investigated? Stating at this point a clear scope of the study would help to understand better the proposed framework capabilities and potential limitations.
7. Page 3 line 76-80: Please clearly state the goals. The current text seems more as a description of what is done but makes not clear to me what are the goals to answer.
8. Page 3 line 83: how long are the road segments in average? And what are the maximum and minimum length those could have for the proposed framework to estimate emission accurately?
9. Page 3 line 84: where is the geometrical attribute of grade for each road segment coming from? Please specify in the text.
10. Page 3 line 84: please provide a definition for road segment, and road level. Also, along the text road link is mentioned. Please unify and/or provide difinitions.
11. Page 3 line 88: Please provide a definition for “vehicle subsegments”.
12. Page 5 line 120: Please provide a justification on why selecting only the temperature is the better option instead of selecting any of the other two options. Have you perform an analysis of the variables that have a stronger effect on cold start emissions or is this based on literature? If the first, please explain. If the second, please add the corresponding references. What is the associated error for the cold exhaust emissions calculation attributed to this selection?
13. Page 5 line 123-125: How are the cold start counts inferred from the hourly traffic count and the road type so that cold start events are identified? Please add explanation in the text.
14. Page 5 line 126-127: Why is the dimensionless factor set to 0.3? Please explain in the text the rationale behind selecting this specific number.
15. Page 5 line 135-136: Please be specific with the terminology, I guess you are referring to emission factors from evaporative emissions
16. Page 5 line 143-144: what is the share of fuel-injected vehicles Yeti assumes in the vehicle fleet? Is that a case-specific share? Please specify in the text.
17. Page 6 line 149: please be specific. Despite this framework seems to be mostly based on HBEFA, this manuscript should read as a standalone text containing all the essential information to follow the story. In this specific case, how many are the engine stops considered in HBEFA or how are those classified, and how does this translate to this specific study? Please, revise the manuscript for similar expressions.
18. Page 6 line 151-152: What is referred to by direct data? Did you perform measurements? Please explain. Also, what percent of engine stops are coming, in this study, from direct data and what percent is estimated? Did you perform a comparison on how well the proposed estimation works compared to the “direct data” of number of engine stops?
19. For each of the emission processes (non-exhaust, evaporative, cold and hot exhaust), please specify which pollutants are estimated in Yeti.
20. Page 6 line 170: It seems that for all the other emission processes HBEFA 3.3 was used and only for non-exhaust PM HBEFA 4.1 was used. What are the implications of using two versions of the model to obtain different sets of emission factors for the different emission processes? Could the framework be adapted to be updated with current and coming versions of HBEFA to calculate emission inventories with the most updated emission factors available? The last is partially raised in the Summary at the end of the manuscript, but it seems that a specific update would be required with each update of HBEFA. An estimation on how fast Yeti can adapt to updates of the emission model would be useful.
21. Page 7 line 177: The term road links is used here. Different terminology is used throughout the text, previously road level, and road segments where mentioned. Please introduce a definition for each of them in the text.
22. Section 2.2 seems to be very specific on the modules used, but I would be interested in knowing how those influence the emission calculation. For example, it is mentioned in lines 178-179 that Yeti can operate on a subset of the traffic network. Does that also imply that if for a specific subset very detailed traffic activity input information is available more detailed emission outcomes could be obtained? Can Yeti consider that in some way? How small can the network subset be? And on the other hand what is the bigger network that would still be an option to run with Yeti. For the last, it would help if the authors provide a processing time and data size quantification for this specific study in Berlin.
23. Section 2.2.1 Data organization, section 2.2.2 User-specified configuration and section 2.2.3 Execution flow: this sections are lacking the specific relationship to the study done in Berlin. It kind of reads as a user manual of Yeti. For example in line 195 it would help to know the specific execution parameters that were used in this study if relevant to explain some of the outcomes. Please revise, I think this is a good opportunity to show the applicability of Yeti and not place the focus only on how the framework internally works.
24. Page 7 line 190: several new concepts are introduced in this line (emission strategy, day type, meteorological profile), please provide a definition for them.
25. Page 9 lines 241-243 in reference to Table 2: some symbols from table 2 need additional explanation such as IDTS.
26. Page 9 Section 3.1.1. HBEFA emission factors and field data: please clearly state which ones are the field data specified in the title of this section and how those were obtained and used in Yeti. A note on how that could be replicated in other cities and which would be the needed information to obtain would be helpful.
27. Page 9 section 3.1.3. This section has a lot of details on how the files look like and the directories, but is lacking an explanation of the content of table 3. This also applies to previous section 3.1.1. Please, it would help to add an interpretation of the information presented in the tables to properly understand them. If not relevant for to explain the results in the main paper, maybe this part could be moved to the supplementary material.
28. Page 10 line 268: Why would you have road segments with zero length? And what is the error associated to ignore the segments that have no indicated traffic direction? Are those a significant amount of your segments?
29. Page 10 line 274: please be specific with the highest resolution that can be achieved. How long are the individual road segments, and what would be the highest temporal resolution?
30. Page 10 line 280: please add the essential information to understand how the data recompilation was done. What period of time was covered with the measurements? What kind of roads were measured? How big was the sample size obtained and how were those postprocessed to be obtain a full dataset to use in Yeti?
31. Page 10 293-294: Why was finally this attribution selected and how is this attribution of LOS IV vs LOS V affecting the results for this specific example?
32. Page 12 line 356: results for the other pollutants (CO and PM) would be also interesting to see in the supplementary material. That would help to understand the applicability of the proposed framework, explain relationships between pollutants and accuracy of the outcomes for all the modelled pollutants in this study.
33. Page 13 section 4.1 (line 365): In figure 4, NOx emissions for summer and winter are the same. Can Yeti consider differences in traffic volumes and traffic composition for different seasons or even lower temporal resolutions (month, week, days,…)?
34. Page 13 Section 4.2. related to the NOx results, was fuel type also considered to analyze the outcomes? what is the average vehicle fuel distribution (% of diesel, gasoline,…) in the example roads for the different vehicle types (passenger cars, bus, motorcycle…)? Together with current tables A1 and A2, this additional table would help to understand better the outcomes presented in tables 10-11.
35. Page 15 Summary section: please revise to add notes on the scope and potential limitations of the framework raised previously.
36. Figure 5: add to the caption the average length of the segments.

Other minor comments:

Page 3, line 63: Revise English in the phrase “but the cost and the level configurability play a significant role”

Page 3, line 66: revise the expression “road network traffic conditions and emission factors are large”. This sentence is not clear to me.

Page 4 line 109: typo, “LODs” seems it should be “LOS”

Page 8 line 226: “from the” is repeated in “from the from the City of Berlin (Diegmann et al, 2020)”

Page 12 line 342: Revise the sentence “an increase in aggregate evaporative increase in evaporative diurnal HC emissions…”

Page 13 line 383 and line 398: busses

Page 19 line 518: 𝛽 is included in table 1 not table 2.