the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
INCHEM-Py v1.2: A community box model for indoor air chemistry
Toby J. Carter
Helen L. Davies
Ellen Harding-Smith
Elliott C. Crocker
Georgia Beel
Zixu Wang
Nicola Carslaw
Abstract. The Indoor CHEMical model in Python, INCHEM-Py, is an open-source and accessible box model for the simulation of the indoor atmosphere, and is a refactor and significant development of the INdoor Detailed Chemical Model (INDCM). INCHEM-Py creates and solves a system of coupled ordinary differential equations that include gas-phase chemistry, surface deposition, indoor/outdoor air change, indoor photolysis processes and gas-to-particle partitioning for three common terpenes. It is optimised for ease of installation and simple modification for inexperienced users, while also providing unfettered access to customise the physical and chemical processes for more advanced users. A detailed user manual is included with the model and updated with each version release. In this paper, INCHEM-Py v1.2 is introduced, the modelled processes are described in detail, with benchmarking between simulated data and published experimental results presented, alongside discussion of the parameters and assumptions used. It is shown that INCHEM-Py achieves excellent agreement with measurements from two experimental campaigns which investigate the effects of people and different surfaces on the concentrations of different indoor air pollutants. In addition, INCHEM-Py shows closer agreement to experimental data than INDCM. This is due to the increased functionality of INCHEM-Py to model additional processes, such as deposition-induced surface emissions. Published community use-cases of INCHEM-Py are also presented to show the variety of applications for which this model is valuable to further our understanding of indoor air chemistry.
- Preprint
(4994 KB) - Metadata XML
-
Supplement
(2386 KB) - BibTeX
- EndNote
David R. Shaw et al.
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2023-1328', Anonymous Referee #1, 31 Jul 2023
General Comments
In this model description paper, Shaw et al. present the INCHEM-Py box model. My interpretation of the paper’s purpose is to provide a user of the model with a reference for the key aspects of the model: processes, parameters, assumptions, verification and examples of use. The novelty and need for such a model is without question. The paper reads very well. However, I cannot currently recommend publication without major revisions as some basic aspects of a GMD description paper have not been met. Below I explain these revisions alongside suggestions for minor and technical revisions. The most significant revision relates to the difference (as defined by GMD) between verification and evaluation. Some key aspects of the model are evaluated against observations, however there is currently no verification to show that the mathematics is operating as expected (more detail below).
Specific Comments
Paragraph beginning line 38 – the justification of needing a model because comprehensive measurements is difficult or lacking is misleading – the model is only as good as the measurements used to evaluate it – and the model offers more than what new measurements could supply – hindcasts, forecasts and hypothetical scenarios. Please reconsider the relationship between model and measurement and the resulting motivation for the model.
Line 52 – witth the motivation for this study being stated as human health – there should be more comment on how the lack of microbial representation affects the limit of model application
Introduction in general – should outdoor effect of indoor emissions be a motivation for the model? There is currentlty no mention of alternative available models and why this model is novel, and there needs to be.
Paragraph beginning line 71 – if INCHEM-Py can only be used with the MCM plus the ‘indoor chemistry’ (with user-defined supplementary reactions), this needs to be stated explicitly, since this is an important facet (e.g. users with alternative chemical schemes (that could take the place of MCM plus ‘indoor chemistry’) cannot apply them).
Line 81 – here please reference the reader to the section of the paper detailing outdoor concentration representation
Line 82 – how is the assumption of irreversible deposition justified, particularly in light of observations of reversible partitioning of some chemicals with surfaces: doi.org/10.1021/acs.est.0c00966
Line 85 – I got lost with the reference to the ‘new surface mechanisms’, as it isn’t clear what this is.
Line 87 – Particle reactions need to be introduced earlier in this paragraph prior to their mentioning in the total number of reactions.
Line 90 and around – Please mention how stoichiometries other than 1 are dealt with.
Line 99 – I am confused about how gas-particle partitioning has the same form as eq. 2 (i.e. the implication is that partitioning is reaction-based, is this correct?), please expand.
Line 106 – please make clearer whether the Jacobian is calculated by INCHEM-Py code or the code of the ODE solver (or something else)
Text preceding eq. (8) – it is not clear what the distinction/relationship between the Hayman, Saunders and Jenkin references are here, and therefore unclear what the provencance of l, m and n is.
Line 131 – why is the transmission factor weighted by photolysis properties of species? Isn’t the transmission factor just a wavelength-dependent fraction of the penetrating light?
Line 133 – are the quantum yields and cross-sections of Kowal et al. (2017) consistent with the values used in for indoor photolysis from natural light? If not, how is this justified?
Section 2.3 – please discuss if/how model users can effect light variations due to indoor settings facing different directions and having varying degrees of natural illumiation (e.g. angle of sunlight incident on windows is such that volume of room illuminated can vary). In addition, it is not clear whether photolysis affects the outdoor concentrations of ingressing species in the same way that it affects indoor species. The comment on line 142 about outdoor OH concentrations suggest that outdoor concentrations are somehow fixed, so please clarify – i.e. are outdoor species subject to the same natural light intensity as indoor species? (By the way, I see that this is dealt with in section 2.3, therefore please reference 2.3 in the relevant section(s) of 2.2)
Section 2.3 – please provide the reader with a note on how they may be limited by outdoor concentrations that are taken from European city summers – if they may be substantially limited please provide (where this is possible) a note on how they may contribute their own outdoor concentrations. Note, that I see in 2.6 the reader is told they may adjust the outdoor concentration file, but this information would be very useful in 2.3.
Line 180 and line 157 – it is not clear exactly which species are modelled to ingress, is it just NO, NO2 and O3, or are OH, HO2 and CH3O2 also ingressing? If OH does ingress, then the statement about lifetime on line 180 appears to be conflicting.
Line 199 – If Kp is the quotient of two rates, shouldn’t it be dimensionless? If I misunderstand, please provide the units of kon and koff.
Eq. (10) – please provide a reference for this equation
Section 2.4 – is confusing: neither here, nor in Carslaw et al. 2012 is a complete calculation for how the model estimates change to species gas- and particle-phase concentrations due to gas-particle partitioning – a complete calculation would show how the change is calculated in molecules/m3 (or molecules/m3/s if solved in ODE form) (the concentration unit used by the model). In addition – how often is this calculation made in the model and where (i.e. alongside chemical reaction ODEs in the solver, or separately)? Finally, please explain why Kp is called partitioning coefficient here but is called partitioning constant in Pankow (1994).
Section 2.8 – is very focussed on oxidant deposition and the resulting yields of organics. However, the reader would benefit from knowing whether organics also deposit to surfaces, and from an indication of whether it is this simulated deposition of organics that provides the organic reactant that generates aldehyde emission from surfaces – or is it the fabric of the surface itself that reacts with oxidants, (or something else)?
Section 2 – what is the benefit and limitation of using a box model for indoor chemistry – compared to other types of model? What is the time interval used for solving ODE equations over and can this be defined by the user?
Section 3.1 – please remind the reader in this section what the relevant difference between INCHEM-Py and INDCM is. Please expand on why spatial represenation is important when comparing against observation.
Overall
- It is unclear what makes this a community model rather than a model.
- Needs some indication of processing time (and processor used) for representative runs
- What is the benefit (and any limitation) of using Python?
- Where does the user access the model?
- A figure demonstrating the code workflow would be helpful so that users know the general structure and order of processing in the model, e.g., which processes are solved inside the ODE solver – and which outside (e.g. gas-particle partitioning is based on equilibrium, so is this solved separately to the ODEs?).
- A section in which the core of the model (gas-phase chemistry) is verified against a benchmark like AtChem2 is needed.
- Currently missing this part of a description paper (https://www.geoscientific-model-development.net/about/manuscript_types.html#item1): The model webpage URL, the hardware and software requirements and the license information should be given in the text. If papers are describing subsequent development to a paper already published in GMD, authors should request them to be electronically linked to the previous version(s) in a special issue, and an overview webpage will be created.
Technical Corrections
Line 29, comma between windows and doors
Line 63 and elsewhere – Shouldn’t letter C should be in math font to be consistent with the equation?
Sections around all equations – please consistently use math format for letters used in equations.
Equation (1) – to be consistent with the use of sigma and j for reactions, surface losses should be summed oved multiple potential surfaces
Line 68 and elsewhere – should ‘air change rate’ read ‘air exchange rate’ to be consistent with indoor research convention? Perhaps the convention has changed recently (in which case please state in rebuttal), but I notice that exchange is used in https://doi.org/10.1021/es301350x
Line 90 – when referrring to something in the gas phase, shouldn’t a hyphen be used: gas-phase reactions rather than gas phase reactions?
Line 132 – full stop missing
Line 177 – I think that loosely and tightly should be swapped to fit the stated air exchange rates
Line 178 – affects rather than effects
Line 182 – space needed after comma
Table 1 – should there be /s included in units?
Line 329 – a rogue ‘8.’
Line 361 – ‘an’ change to ‘a’
Line 669 - doi of Shaw 2023b looks funny
Citation: https://doi.org/10.5194/egusphere-2023-1328-RC1 -
RC2: 'Comment on egusphere-2023-1328', Anonymous Referee #2, 30 Aug 2023
General Comments
The manuscript authored by Shaw et al. describes an advanced indoor air quality model, which shows improved performance over the previous version(s) in matching experimental results. The manuscript reads well. The cited literature is proper and accredited. The mathematical expressions are correctly defined and used. The model is flexible, and future users are expected to be able to reproduce the results discussed with enough familiarity with the framework. Nevertheless, I cannot recommend its publication unless the manuscript is improved with some major revisions. Some assumptions, especially those associated with airborne particles, are not clearly outlined. The manuscript involves insufficient discussions on how the interaction between essential sources and sinks leads to model predictions per its mathematical framework. There is no information on the errors related to the experimental data used to evaluate the predictions, and the claims on the acceptability of model-experiment comparison results are not substantiated enough. While several limitations (see my specific comments outlined below) arise from the model's assumptions and parameterization, the authors are missing a thoughtful discussion on model limitations. Finally, the manuscript lacks sufficient elaboration on some equations and procedures (see my specific comments outlined below). For a model with a stand-alone user manual, it is not straightforward to draw a line for the limit of details expressed within the manuscript versus the information obtained via consulting the manual. In my opinion, the current version of the manuscript is over-referring the reader to the manual and other information uploaded into the online repository.ÂSpecific Comments
Line 34: The statement "SOA are ultrafine particles..." is inaccurate. As an OA component, SOA is a fraction of PM with various sizes, not necessarily ultra-fine and breathable particles.The paragraph starting with line 54: It will be more helpful to rationalize why you consider the processes included in the model framework essential in contrast with those not considered.Â
The paragraph starting with line 65: You chose to mention the units for the Equation's left-hand side on the previous paragraph. I would also suggest mentioning units for the parameters on the right-hand side too.Â
Line 86: This is almost the first time you mention the particles. I suggest more clarification. Have you considered different sizes of particles? Does your model consider particle sources and sinks like dust in-tracking, deposition, and resuspension? If not, how can user-defined dynamics be integrated into the model?
Equation (3): Why only first-order reactions? Are first-order kinetics accurate per the MCM mechanisms? Nevertheless, the equation format is confusing. I am unsure how the summation is expanded in practice. For instance, if two reactants are A and B, would the reaction rate be given by k*(CA + CB) or kA*CA + kB*CB?
Equation (4): The Rij could be positive or negative depending on whether species i is a reactant or a product. This distinction is not reflected in Equation (3). That expression, as the sum of non-negative multiplications, seems to be always positive.Â
Equation (5): This is unnecessarily detailed. Why would you mention the Jacobian equation when you are not providing more context on how it is utilized in the computation framework? Equation (5) is just the mathematical definition of the Jacobian. Assuming the reader needs this level of information, one expects to receive more mathematical details. If you choose to provide a high-level summary of the solution process, there is no need to post an equation for the Jacobian.
Equation (6): Explain the coefficients and constants used in Equation (6). I can only understand the logic behind the numbers -23.45 and 365.25. Note that you have made many explanations for clearer equations in previous sections.
The discussion around Figure 2: Box models are prone to severe errors upon characterizing indoor photochemical phenomena. The OH concentrations featured in Figure 2 are more reconcilable with near-window areas instead of bulk indoor space (See Lakey et al., https://doi.org/10.1038/s42004-021-00548-5). I can hardly relate to indoor OH concentrations beyond 10^5 molecules per cubic centimeters from the order of magnitude point of view. Since you have dedicated a section to photochemistry, I think the lack of dimensionality and its implications on photochemistry need to be further discussed. This comment exemplifies my broader point above that the manuscript lacks a thoughtful discussion on model limitations and critical assumptions.
Lines 143 and 144: I can hardly reconcile your interpretations for the two figure panels. The inter-seasonal gap between the peaks in Figure 2(b) is larger than the inter-latitude gap in Figure 2(a). However, you discuss differences in the latter while commenting on the former as "comparable."
Lines 163 and 164: These cities do not represent well the latitude range in your model.Â
Lines 176 and 177: The air change rate does not only depend on airtightness. Even airtight buildings can have air change rates as high as 5 per hour upon window opening or operating high-speed ventilation fans.Â
Lines 180 to 192: These two paragraphs would be more informative if they followed an alternative organization. I would suggest first mentioning the reactions. Then, argue how ACR changes will contribute to OH increase or decrease per the competing phenomena associated with reactions. Then, comment on how their interplay leads to the ACR effect per the extreme case of 0.2 and higher values.
Lines 185 and 186: You should also mention that enhanced ACR contributed to removing OH from the indoor domain. However, the elevated production effect seems more intensified.
General comment on Section 2.2 and Section 2.3: It is excellent to evaluate the model against published experimental data. However, the reader of modeling research also wants to see information on model verification. What does the model tell you about interactions between rate processes? What are the fundamental mechanisms driving concentrations for typical conditions? Does the mathematical behavior of model predictions follow the established physical principles? For instance, how would you interpret the local minima and maxima in Figure 4?
Line 194: Why only these species? You have species more prone to sorption to particles like octanal in Table A.1. The parameterization does not seem to limit you in extending the list of partitioning species, as the parameters in Equation (10) are available for many other chemicals.Â
Equations (9) and (10): As you have chosen to present some equations from Pankow (1994), the set needs to be logically comprehensive. You need to provide expressions from kon and koff (at least within your appendices) to guide the reader on how Equation (10) is deduced from Equation (9).
General comment section 2.4: Most of the discussion in this section is dependent on particle size. It seems that this factor was not included in your framework. If not, this assumption must be explicitly mentioned and discussed.
Figure 5: I am confused by the results. Regardless of how the temperature is evaluated, the model should perform similarly for a given temperature. The left panel shows that the temperature estimations are almost identical for the three methods at t~13 h. Why do the predicted OH values differ simultaneously in the right panel?
Lines 230 to 232: You are not providing your reader with enough context to compare bimolecular rate constants. I suggest calculating the decay timescales (for example, in seconds) for the rate constant to assist the reader in comparing the values. How do these reactive sinks compare with ventilation and partitioning sinks?
Lines 280 and 281: Some context, preferably listing the parametrization for each surface, is needed to help the reader understand their distinctness. For instance, glass and metal are often considered in the same category in indoor modeling. Regard it this way: if we want to consider a new surface, do we need to add another surface category, or are there guides on how your proposed suit could represent new surfaces?
Lines 288 and 289: There seem to be several reactions entangled here. Are you assuming that their yields are independent? How are you combining the yields to have a resultant one?
The discussion around Figure 8: Your model is definitely doing a better job in predicting ozone degradation compared to Kruza and Carslaw. Per comparison with experimental data, how would you define a good agreement? Usually, errors within data variability ranges signal satisfactory performance. The experimental data shown in Fig.8 have no information on data variability (e.g., error bands or error bars). I could not find any information on experimental uncertainty when I checked the cited reference itself, although you pointed out uncertainties in line 353 without providing any quantitative information. Couldn't you find a study with repeated measurements? With no information on data precision, it is hard to accept your claim about satisfactory model-observation parity. For example, 4-OPA experimental and model data differ by a factor of two. Is that difference still within the experimental error and dismissable? Moreover, the model performance in predicting 4-OPA does not seem to trump Kruza's model. Regarding 4-OPA concentration, Figure 8 shows that both your predictions and Kruza's are biased. Yours at the beginning and theirs at some time later.
The discussion around Figure 9: Are the outdoor concentrations used to create this figure comparable with the conditions associated with Table A.1 values?
The text within parentheses in line 352: I suggest only referring to the manuscript sections instead of a Python file in your code repository.
Lines 360 and 361: Is this difference significant enough to be discussed? You have dismissed more considerable differences between the model and observations at t~16.9 hours as cases of good model-observation agreement.
General comment on discussions around Figure 8 and Figure 9: There could be objections to this strategy of adjusting emission rates to reproduce initial observations. There is a hidden assumption with this methodology that your model grasps all the applicable physicochemical phenomena, which is not the case. How alternative emission rates would impact your estimations? I would suggest some sensitivity analysis to assess model performance during other conditions.
General comment on Appendix A: i)Some of the updates between versions 1.1 and 1.2 are noticeable (e.g., acrolein changes by more than one order of magnitude), while some are minute (e.g., the updated value for methane is different from the older estimation by less than 1%). What is your criterion for implementing an update? I believe a sensitivity analysis could be helpful. How much are your results sensitive to changes in outdoor concentrations of a species? Do these changes matter in light of the sensitivity estimates?
ii) Most of these references cited in the table precede in publication time v1.1 and v1.2. So, which version did they serve? v1.2? More clarification is appreciated.
iii) Outdoor concentrations are pretty dependent on location and time. The references you are citing here are different from each other regarding those parameters. For instance, Uchiyama et al.'s numbers pertain to Japan, whereas Bari and Kindzierski estimations are for Calgary, Canada. It would help if you cautioned the reader about the spatial and temporal variability of outdoor changes. You also need to add another column to this table to explicitly signal the location and time attributes of the cited references to prevent the reader from misunderstanding that the values are globally applicable.General comment on Appendix B: This text could be eliminated and transferred to the online repositories. The reader has little idea how these variables are handled in your source code. So, what is the point of putting it here as if it is just coming out of nowhere?
General comment for the whole manuscript: I understand that the main source code consists of several functions and sub-programs that cannot be directly discussed in the manuscript. However, it will be helpful to have a figure, at least in an appendix section, demonstrating the general workflow of operating the model. A high-level flow chart referencing the code's functions/variables will work great in that capacity.Â
Technical Corrections:Â
The complete list of typos and punctuation errors is longer than what follows. However, I guess you can easily tackle those upon another round of proofreading after implementing the revisions.Lines 18 and 19: The cited link is invalid. You can use this one: https://www.who.int/publications/i/item/9789289002134.
Line 28: A comma is missing before "and."
Line 29: Missing commas. It should be "windows, doors, and cracks."
Line 132: Missing period after the word "range."
Figure 7: The limonene arrows seem more red than orange.
Figure 8: Your choices of color are not effective here. Grey and black are too similar. Such a bright yellow is not suitable for a white background.
Citation: https://doi.org/10.5194/egusphere-2023-1328-RC2
David R. Shaw et al.
Data sets
INCHEM-Py v1.2: A community box model for indoor air chemistry paper data David R. Shaw, Toby J. Carter, Helen L. Davies, Ellen Harding-Smith, Elliott C. Crocker, Georgia Beel, Zixu Wang, and Nicola Carslaw https://doi.org/10.15124/849aa8cb-701d-4468-b5da-070a9fb21801
Model code and software
INCHEM-Py v1.2: A community box model for indoor air chemistry David R. Shaw, Toby J. Carter, Helen L. Davies, Ellen Harding-Smith, Elliott C. Crocker, Georgia Beel, Zixu Wang, and Nicola Carslaw https://doi.org/10.5281/zenodo.8046598
INCHEM-Py, github David R. Shaw, Toby J. Carter, Helen L. Davies, Ellen Harding-Smith, Elliott C. Crocker, Georgia Beel, Zixu Wang, and Nicola Carslaw https://github.com/DrDaveShaw/INCHEM-Py
David R. Shaw et al.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 292 | 94 | 10 | 396 | 26 | 10 | 8 |
- HTML: 292
- PDF: 94
- XML: 10
- Total: 396
- Supplement: 26
- BibTeX: 10
- EndNote: 8
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
