the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 14 Mar 2025
Application of PRIM for understanding patterns in carbon dioxide model-observation differences
Abstract. Reducing uncertainties in regional carbon balances requires a better understanding of CO2 transport in synoptic weather systems. Here, we apply the Patient Rule Induction Method (PRIM) to airborne observations of of potential temperature, wind speed, water vapor mixing ratio, and CO2 dry mol fraction gathered during the Atmospheric Carbon and Transport (ACT)-America Summer 2016 and Winter 2017 campaigns. ACT observations were targeted at expert-designated cases of fair weather and near-frontal warm and cold sector air at atmospheric boundary-layer, lower-, and higher free tropospheric levels (ABL, LFT, and HFT, respectively).
We investigate atmospheric characteristics cases of these pre-defined cases and associated CO2 model-observation-differences in the mesoscale WRF-Chem model. PRIM results separate winter- and summertime observations as well as observations from ABL, LFT, and HFT with enrichment factors of 4–20 inside the PRIM box compared to the entire dataset but cannot distinguish between near-frontal warm and cold sector observations in the higher free troposphere. Using the PRIM constrained atmospheric parameter space, we find that large magnitude model observation differences preferentially associated with times when atmospheric conditions are less typical. This association suggests that that PRIM could provide a useful tool for isolating atmospheric conditions with large-magnitude and non-Gaussian CO2-residuals for targeted transport model evaluation and to potentially improve inversion results during synoptically active periods.
- Preprint
(1000 KB) - Metadata XML
-
Supplement
(371 KB) - BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on egusphere-2025-341', Anonymous Referee #1, 31 Mar 2025
Main comment
This study investigates synoptic weather system related to CO2 transport by applying PRIM to airborne observations from the ACT-America summer 2016 and winter 2017 campaigns, and aims at quantifying the WRF-Chem model uncertainties associated with specific atmospheric characteristics to improve atmospheric inversion processes.
The present study appears to be relevant, novel and well within the scope of ACP. However, I believe that some improvements could be made. They are summarized here, and they are further expanded in the detailed/line-by-line comments:
- The manuscript is overall well-structured, but the method subsection that illustrates the PRIM (2.3) should be reworked to be clearer for also readers that are unfamiliar with this method. Furthermore, the section that describes the atmospheric conditions in ACT (3.1) is very short, and I suggest either to expand it (e.g. by highlighting facts that are relevant to the results later presented in the paper) or by integrating it in another subsection.
- The manuscript has some typos, and there are phrases throughout the text that could be improved in clarity
- This study’s results could be made more robust by adding another data analysis method, as some of the results could be due to PRIM’s features, rather than being actual properties of the considered atmospheric characteristics. A conventional, simple clustering method would suffice, and this could also be included in the SI. In my opinion this will make the results presented more robust in the case the two methods agree (and if they do not then the cause could be guessed). While I don't think that this aspect is critical (which is the reason for suggesting minor revision), I strongly believe that this aspect should be investigated further.
Detailed/line-by-line comments
Line 15: The introduction is very simple and clear, covers the issue at hand effectively and states the study's goal effectively. I do suggest only a couple of changes in this section.
Line 56-57: Explanation of what OCO-2 v9 MIP is should be included here.
Line 72: The method subsections 2.1 and 2.2 seem good, but 2.3 needs quite some adjustments in my opinion, as many concepts are not explained clearly.
Line 87: “To facilitate our analysis, we exclusively use data from level-leg flight segments”. Why would this facilitate the analysis?
Line 101-102: “with unusually high [CO2], indicative of CO2 point sources, ([CO2]>430ppm)”. Could this be justified better? How are you sure that you are not canceling any actual feature when excluding these kind of outliers?
Line 111: As far as I know, residuals are usually defined as "observed minus modelled", rather than the other way around
Line 119: Please review this subsection carefully, as terms here appear to be used interchangeably and could result in some confusion.
Line 121-123: “Simple rules … higher than usual frequency.” This sentence is not clear. What are "simple rules about input variables"? What does it mean that a "designated variable of interest occurs at a higher than usual frequency"?
Line 123-125: “PRIM rules … (Hadka et al., 2015)” This sentence seems a little bit off here, and I suggest to move it later in the section.
Line 126-130: “The PRIM … inside the box.” This is not very clear - what does higher mean value mans? What is a target?
Line 143-149: This example makes things more clear than the explanation what was provided before, so I'd try to restructure the section to present this earlier in order to explain what PRIM is - I also suggest using this toy example to clarify definitions (e.g. writing here in parentheses what is a target/variable of interest/frequency, and so on).
Line 156: Is the threshold 0.75 arbitrary? While sounding reasonable, this choice should be justified.
Line 171-172: One concern of mine here is related to how the ACT dataset is handled. Shouldn't the PRIM classification capabilities be tested on an external dataset? i.e. determine the parameter subspace for each classes on a subset of the ACT dataset ("calibration/training dataset"), and then apply those rules to the rest of the data and see how they match with the expert designations ("validation dataset" – whose results should be displayed in Figure 3 and referred in this section).
Line 268-272: I'd rather say that PRIM's inability is in line with the hypothesis that upper tropospheric air represents background conditions. The way it was phrased makes it seem like PRIM’s inability discern frontal warm and cold sectors is a proof for upper tropospheric air representing background conditions (which cannot be proven this way and would need more evidences).
Typos/unclear sentences
Line 2: “of of”
Line 10-12: “Using the PRIM … are less typical.” Unclear sentence
Line 12: “that that”
Line 60-62: “model … (Gerken et al., 2021).” Unclear sentence
Line 65: Useful tool?
Line 156-166: “There is notably are much larger”?
Line 191: “1% 1%”
Line 250: “to to”
Line 258-260: “While PRIM … by PRIM” Unclear sentence.
Figures and tables
Figure 1: y-axis labels overlap
Figure 4: I think it would be better if y-axis limits were consistent throughout ABL-LFT-HFT
Figure S1: figure resolution should be higher
Figure 5 and figure 6: a plot like the one in figure 6 feels more useful than the one figure 5 (which could be as well SI in my opinion). Is there a reason why this is just ABL and not LFT and HFT?
Citation: https://doi.org/10.5194/egusphere-2025-341-RC1 - AC1: 'Reply on RC1', Tobias Gerken, 17 Jul 2025
-
RC2: 'Comment on egusphere-2025-341', Anonymous Referee #2, 06 May 2025
This MS used PRIM method to evaluate the model-observation difference during synoptic weather systems, and identified the influencing factors. Honestly speaking, I am good at modeling and observating the GHG concentration and also inversing approach, not a expert of PRIM method, but as a normal reader, it seems vary hard to understand the logical expression and follow authors' main idea. The difference between observation and simulations can be treatd mainly caused by both atmospheric transport model and prior CO2 flux, where the atmospheric inversion approach usually give an known uncertainty extent to transport model and attribute the difference of observed and simulated GHG concentration (or enhancement) to spatial-temporal bias in CO2 flux. Compared with use of PRIM method, direct comparasions between spatial-temporal difference between model and simulations can better give us potential reasons and locations of such difference. In general, the discription of PRIM method is not well displayed or explained, which (I think) is also vary hard for other audience to follow this MS. the readers may want to know what's the main reasons between observations and simulations during different synoptic weather systems or contributors (values) behind them. However, in the 5. conclusion section, no values are displayed that was concluded in this MS, the authors mainly describe the background, meanning of this study and the advantage of PRIM method, which seems unnormal as a research paper.
there are also some Logical jump that make me confused i.e. line 8 "PRIM results separate winter- and summertime observations as well as observations from ABL, LFT, and HFT with enrichment factors of 4–20", what does with enrichment factors of mean? please explain it at the first time when it appear, and I also can not find how does 4-20 come from in the main text, from the numbers in Table 2, it should be 4.0-20.5 or 4-21.
instead of using description on lines 120-155, the diagram figure that explain how PRIM method works is strongly suggested.
typo as "This association suggests that that PRIM"
Citation: https://doi.org/10.5194/egusphere-2025-341-RC2 - AC2: 'Reply on RC2', Tobias Gerken, 17 Jul 2025
Status: closed
-
RC1: 'Comment on egusphere-2025-341', Anonymous Referee #1, 31 Mar 2025
Main comment
This study investigates synoptic weather system related to CO2 transport by applying PRIM to airborne observations from the ACT-America summer 2016 and winter 2017 campaigns, and aims at quantifying the WRF-Chem model uncertainties associated with specific atmospheric characteristics to improve atmospheric inversion processes.
The present study appears to be relevant, novel and well within the scope of ACP. However, I believe that some improvements could be made. They are summarized here, and they are further expanded in the detailed/line-by-line comments:
- The manuscript is overall well-structured, but the method subsection that illustrates the PRIM (2.3) should be reworked to be clearer for also readers that are unfamiliar with this method. Furthermore, the section that describes the atmospheric conditions in ACT (3.1) is very short, and I suggest either to expand it (e.g. by highlighting facts that are relevant to the results later presented in the paper) or by integrating it in another subsection.
- The manuscript has some typos, and there are phrases throughout the text that could be improved in clarity
- This study’s results could be made more robust by adding another data analysis method, as some of the results could be due to PRIM’s features, rather than being actual properties of the considered atmospheric characteristics. A conventional, simple clustering method would suffice, and this could also be included in the SI. In my opinion this will make the results presented more robust in the case the two methods agree (and if they do not then the cause could be guessed). While I don't think that this aspect is critical (which is the reason for suggesting minor revision), I strongly believe that this aspect should be investigated further.
Detailed/line-by-line comments
Line 15: The introduction is very simple and clear, covers the issue at hand effectively and states the study's goal effectively. I do suggest only a couple of changes in this section.
Line 56-57: Explanation of what OCO-2 v9 MIP is should be included here.
Line 72: The method subsections 2.1 and 2.2 seem good, but 2.3 needs quite some adjustments in my opinion, as many concepts are not explained clearly.
Line 87: “To facilitate our analysis, we exclusively use data from level-leg flight segments”. Why would this facilitate the analysis?
Line 101-102: “with unusually high [CO2], indicative of CO2 point sources, ([CO2]>430ppm)”. Could this be justified better? How are you sure that you are not canceling any actual feature when excluding these kind of outliers?
Line 111: As far as I know, residuals are usually defined as "observed minus modelled", rather than the other way around
Line 119: Please review this subsection carefully, as terms here appear to be used interchangeably and could result in some confusion.
Line 121-123: “Simple rules … higher than usual frequency.” This sentence is not clear. What are "simple rules about input variables"? What does it mean that a "designated variable of interest occurs at a higher than usual frequency"?
Line 123-125: “PRIM rules … (Hadka et al., 2015)” This sentence seems a little bit off here, and I suggest to move it later in the section.
Line 126-130: “The PRIM … inside the box.” This is not very clear - what does higher mean value mans? What is a target?
Line 143-149: This example makes things more clear than the explanation what was provided before, so I'd try to restructure the section to present this earlier in order to explain what PRIM is - I also suggest using this toy example to clarify definitions (e.g. writing here in parentheses what is a target/variable of interest/frequency, and so on).
Line 156: Is the threshold 0.75 arbitrary? While sounding reasonable, this choice should be justified.
Line 171-172: One concern of mine here is related to how the ACT dataset is handled. Shouldn't the PRIM classification capabilities be tested on an external dataset? i.e. determine the parameter subspace for each classes on a subset of the ACT dataset ("calibration/training dataset"), and then apply those rules to the rest of the data and see how they match with the expert designations ("validation dataset" – whose results should be displayed in Figure 3 and referred in this section).
Line 268-272: I'd rather say that PRIM's inability is in line with the hypothesis that upper tropospheric air represents background conditions. The way it was phrased makes it seem like PRIM’s inability discern frontal warm and cold sectors is a proof for upper tropospheric air representing background conditions (which cannot be proven this way and would need more evidences).
Typos/unclear sentences
Line 2: “of of”
Line 10-12: “Using the PRIM … are less typical.” Unclear sentence
Line 12: “that that”
Line 60-62: “model … (Gerken et al., 2021).” Unclear sentence
Line 65: Useful tool?
Line 156-166: “There is notably are much larger”?
Line 191: “1% 1%”
Line 250: “to to”
Line 258-260: “While PRIM … by PRIM” Unclear sentence.
Figures and tables
Figure 1: y-axis labels overlap
Figure 4: I think it would be better if y-axis limits were consistent throughout ABL-LFT-HFT
Figure S1: figure resolution should be higher
Figure 5 and figure 6: a plot like the one in figure 6 feels more useful than the one figure 5 (which could be as well SI in my opinion). Is there a reason why this is just ABL and not LFT and HFT?
Citation: https://doi.org/10.5194/egusphere-2025-341-RC1 - AC1: 'Reply on RC1', Tobias Gerken, 17 Jul 2025
-
RC2: 'Comment on egusphere-2025-341', Anonymous Referee #2, 06 May 2025
This MS used PRIM method to evaluate the model-observation difference during synoptic weather systems, and identified the influencing factors. Honestly speaking, I am good at modeling and observating the GHG concentration and also inversing approach, not a expert of PRIM method, but as a normal reader, it seems vary hard to understand the logical expression and follow authors' main idea. The difference between observation and simulations can be treatd mainly caused by both atmospheric transport model and prior CO2 flux, where the atmospheric inversion approach usually give an known uncertainty extent to transport model and attribute the difference of observed and simulated GHG concentration (or enhancement) to spatial-temporal bias in CO2 flux. Compared with use of PRIM method, direct comparasions between spatial-temporal difference between model and simulations can better give us potential reasons and locations of such difference. In general, the discription of PRIM method is not well displayed or explained, which (I think) is also vary hard for other audience to follow this MS. the readers may want to know what's the main reasons between observations and simulations during different synoptic weather systems or contributors (values) behind them. However, in the 5. conclusion section, no values are displayed that was concluded in this MS, the authors mainly describe the background, meanning of this study and the advantage of PRIM method, which seems unnormal as a research paper.
there are also some Logical jump that make me confused i.e. line 8 "PRIM results separate winter- and summertime observations as well as observations from ABL, LFT, and HFT with enrichment factors of 4–20", what does with enrichment factors of mean? please explain it at the first time when it appear, and I also can not find how does 4-20 come from in the main text, from the numbers in Table 2, it should be 4.0-20.5 or 4-21.
instead of using description on lines 120-155, the diagram figure that explain how PRIM method works is strongly suggested.
typo as "This association suggests that that PRIM"
Citation: https://doi.org/10.5194/egusphere-2025-341-RC2 - AC2: 'Reply on RC2', Tobias Gerken, 17 Jul 2025
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 380 | 44 | 23 | 447 | 30 | 14 | 26 |
- HTML: 380
- PDF: 44
- XML: 23
- Total: 447
- Supplement: 30
- BibTeX: 14
- EndNote: 26
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1