the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 19 Sep 2022
Predicted and Observed Changes in Summertime Biogenic and Total Organic Aerosol in the Southeast United States from 2001 to 2010
Abstract. Biogenic secondary organic aerosol (bSOA) is a major component of atmospheric particulate matter (PM2.5) in the southeast United States especially during the summer, when emissions of biogenic volatile organic compound (VOCs) are high and emissions from anthropogenic sources enhance the formation of secondary particulate matter. We evaluate the performance of PM2.5 organic aerosol predictions by a chemical transport model (PMCAMx) in response to significant changes in anthropogenic emissions during the summers of 2001 and 2010. Average predicted bSOA concentrations in the southeast US did not change appreciably from the summer of 2001 to the summer of 2010, while the anthropogenic SOA decreased by 45 %. As a result, the biogenic fraction of total OA increased from 0.46 in 2001 to 0.63 in 2010. Partitioning effects due to reduced anthropogenic OA from 2001 resulted in 0.4 µg m-3 less biogenic OA on average in the southeast US in the summer of 2010. This was offset by biogenic SOA increases due to higher biogenic vapor emissions in the warmer 2010 summer. Little noticeable difference was observed in OA prediction performance in the southeast US between the two summer simulation periods. The fractional error of OA predictions remained practically the same (0.41 and 0.44 at CSN sites and 0.40 to 0.41 at IMPROVE sites in the summers of 2001 and 2010 respectively). The fractional bias of OA predictions increased from 0.10 to 0.22 at CSN sites and decreased from 0 to -0.09 at IMPROVE sites between the two periods. Removing the NOx-dependence of SOA formation yields resulted in higher fractional error and fractional bias at both CSN and IMPROVE sites in both summer periods, demonstrating the efficacy of the current formulation of SOA yields. Our analysis suggests that the changes in biogenic OA in this forested relatively polluted region appear to be dominated by the partitioning effects and the NOx effects on SOA yields.
- Preprint
(2246 KB) - Metadata XML
-
Supplement
(220 KB) - BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on acp-2022-648', Anonymous Referee #1, 24 Oct 2022
Review of “Predicted and Observed Changes in Summertime Biogenic and Total Organic
Aerosol in the Southeast United States from 2001 to 2010” by Dinkelacker et al.
This manuscript describes application of the air quality model PMCAMx to the southeastern United States from 2001 to 2010. The authors describe changes in predicted organic aerosol over the time period in response to their simulated changes in anthropogenic emissions of SO2, NOx and NMVOCs, in addition to biogenic emissions of isoprene, monoterpenes, and sesquiterpenes. The authors find that NOx-dependence and partitioning have the largest influence on the anthropogenic controls on bSOA. It’s my opinion this conclusion is not robustly defended and I cannot recommend publication in the present form.
My two most pressing overarching comments:
- The discussion of RH dependence for SOA yield in the Introduction is lacking. Further, the lack of water pathways in the SOA model, and sole reliance on semi-volatile partitioning … and then finding that semi-volatile partitioning in the predominant controlling mechanism lacks context and the reasoning is flawed.
To my knowledge, Kamens was the first to show that changing RH in a laboratory smog chamber experiment does not necessarily change the amount of liquid water, and that water is the overriding controlling factor. The Kamens paper cited here has liquid water in the title.
For example, in the cited Nguyen paper, there was no seed in the smog chamber experiments. They found no dependence on RH over a wide range. A dramatic change in RH in those experiments did not change liquid water changed (except maybe on the walls?). Nguyen went on to find in later controlled smog chamber experiments with inorganic seeds, where liquid water in aerosols did change, that liquid water for bSOA formation had the predominant effect and in fact seemed to be necessary. Water mattered much more so than pH in those experiments. https://acp.copernicus.org/articles/14/3497/2014/
It’s odd that the water context is not discussed, in particular because the authors want to argue that partitioning is the dominant controlling mechanism, and because water pathways – neglected in this model application - are expected to contribute substantially to SOA in the region according to Carlton and Turpin: https://acp.copernicus.org/articles/13/10203/2013/acp-13-10203-2013.html
- The authors state in multiple locations that they compared their OA predictions to OA from the CSN and IMPROVE networks. This is highly problematic because neither measures OA, they measure OC. Did the authors mistakenly compare their OA predictions to OC measurements or did they use an OM:OC ratio? If they used a conversion factor what was it and was it constant? It is well established that this ratio is changing. Malm finds that the ratio is increasing. https://doi.org/10.1029/2019JD031480. Hand et al., http://vista.cira.colostate.edu/improve/wp-content/uploads/2019/03/Hand2019.pdf finds that RH problems in the laboratory are likely impacting mass measurements. Depending on the OM:OC ratio method employed, this could impact OA estimates. The authors provide no information on this, and it is difficult to surmise what potential impacts could be.
specific comments:
Line 86: could be primary
Application of MEGAN: did the land description change at all over the time period and if so, how was this reflected? If not, how is that choice justified?
Starting at Line 185: the authors state: “Anthropogenic VOC emissions decreased… which should cause a decrease in biogenic OA due to partitioning effects.” This is not necessarily true. Less anthropogenic VOCs will result in less competition for OH and other radicals. Biogenic VOCs that may have blown out of the domain in the older time period simulations, may have opportunity to react and form semi-volatile species in the newer time period simulation.
Citation: https://doi.org/10.5194/acp-2022-648-RC1 -
AC1: 'Response to Referee #1', Spyros Pandis, 12 Dec 2022
(1) This manuscript describes application of the air quality model PMCAMx to the southeastern United States from 2001 to 2010. The authors describe changes in predicted organic aerosol over the time period in response to their simulated changes in anthropogenic emissions of SO2, NOx and NMVOCs, in addition to biogenic emissions of isoprene, monoterpenes, and sesquiterpenes. The authors find that NOx-dependence and partitioning have the largest influence on the anthropogenic controls on bSOA. It’s my opinion this conclusion is not robustly defended and I cannot recommend publication in the present form.
We appreciate the effort devoted by the referee to review our manuscript. In this paper we do show that a chemical transport model that includes the effects of NOx on the gas-phase chemistry of an environment that is dominated by biogenic VOCs and the changes in the partitioning of semivolatile organic aerosol can explain reasonably well the observed changes in total organic aerosol in this area. The study investigates changes over a decade in which anthropogenic emissions of SO2, NOx, VOCs and organic aerosol took place. We do understand the argument of the referee that this does not necessarily prove that the simulated processes are the dominant ones. There can be scenarios in which two processes can have the same effect (e.g., lead to the oxidation of an organic molecule and transfer it to the particulate phase) and simulating just one of them is sufficient to reproduce observations. To address this issue we have rephrased this specific conclusion to explain that a model like PMCAMx simulating specific processes can reproduce the observed changes in this region that is rich in biogenic VOCs and SOA, without getting into the argument about the dominant processes.
Our responses and the corresponding changes (in regular font) follow the comments of the reviewer (in italics).
(2) The discussion of RH dependence for SOA yield in the Introduction is lacking. Further, the lack of water pathways in the SOA model, and sole reliance on semi-volatile partitioning … and then finding that semi-volatile partitioning in the predominant controlling mechanism lacks context and the reasoning is flawed. To my knowledge, Kamens was the first to show that changing RH in a laboratory smog chamber experiment does not necessarily change the amount of liquid water, and that water is the overriding controlling factor. The Kamens paper cited here has liquid water in the title. For example, in the cited Nguyen paper, there was no seed in the smog chamber experiments. They found no dependence on RH over a wide range. A dramatic change in RH in those experiments did not change liquid water changed (except maybe on the walls?). Nguyen went on to find in later controlled smog chamber experiments with inorganic seeds, where liquid water in aerosols did change, that liquid water for bSOA formation had the predominant effect and in fact seemed to be necessary. Water mattered much more so than pH in those experiments. It’s odd that the water context is not discussed, in particular because the authors want to argue that partitioning is the dominant controlling mechanism, and because water pathways – neglected in this model application - are expected to contribute substantially to SOA in the region according to Carlton and Turpin.
We have followed the suggestion of the referee and extended the discussion of the RH dependence on the SOA in the Introduction section. We have also added discussion of previous work regarding the formation of SOA in the aqueous phase. In both cases all references suggested by the reviewer are now cited. We then explain that the version of PMCAMx used in this work does not include these effects that have been observed in laboratory experiments and therefore any significant discrepancies between model predictions and observations could be due to the lack of simulation of these effects.
We also recognize that the explanation of the objectives of the present work may have not been clear enough in the original manuscript. There have been numerous laboratory studies suggesting that a process may affect SOA formation and therefore OA levels in the atmosphere. These include the water related processes discussed above, pH effects, catalytic effects, photo-degradation reactions, oligomerization reactions, etc. Our approach in this paper was to simplify (instead of complicating) the model used and test if this simpler model version can explain the observed changes. The main drivers of SOA formation in the CTM version used are precursor emissions, SOA yields (with empirically derived NOx-level dependence), NOx and VOC effects on gas-phase chemistry, semivolatile partitioning and chemical aging reactions as described by the VBS. We do recognize that there is always additional science that one could implemented in a CTM for an increased level of mechanistic accuracy. As stated in the manuscript, performance was consistent between the two periods even with significant changes in emissions. The findings suggest that the main drivers already accounted for in the model respond accordingly to those changes. We of course agree that there are processes affecting SOA formation in the atmosphere that are not accounted for in this version of the CTM - and there likely always will be. This discussion has been added to the revised manuscript.
(3) The authors state in multiple locations that they compared their OA predictions to OA from the CSN and IMPROVE networks. This is highly problematic because neither measures OA, they measure OC. Did the authors mistakenly compare their OA predictions to OC measurements or did they use an OM:OC ratio? If they used a conversion factor what was it and was it constant? It is well established that this ratio is changing. Malm finds that the ratio is increasing. Hand et al., finds that RH problems in the laboratory are likely impacting mass measurements. Depending on the OM:OC ratio method employed, this could impact OA estimates. The authors provide no information on this, and it is difficult to surmise what potential impacts could be.
We are well aware of difference between OC and OM and the need to covert the OC measurements to OM (or vice versa) for model evaluation. We do agree with the reviewer though that additional information is needed for this part of the analysis. We do explain in the revised manuscript the OM:OC ratios used for the conversion (they are different for the urban and the rural sites) and provide the corresponding references. It should be noted that the IMRPOVE network does report OA measurements now using the OM:OC values used in our work. We should also point out that significant effort was put into verifying the OC measurements, taking care to sort out inconsistencies in artifact corrections for historical data corresponding to different measurement codes for CSN OC. These corrections were applied appropriately to the corresponding measurement codes. A discussion of this issue has been added to the revised paper.
(4) Line 86: could be primary.
This is a valid point. We have rephrased the sentence to include the possibility that it may be primary anthropogenic OA too.
(5) Application of MEGAN: did the land description change at all over the time period and if so, how was this reflected? If not, how is that choice justified?
The differences in biogenic emissions predicted with MEGAN3 were driven by the differences in WRF predicted meteorology. There were only small changes in the land use change in the overall area over the relatively short period (a decade) between the two simulations. This is now explained in the revised paper.
(6) Starting at Line 185: the authors state: “Anthropogenic VOC emissions decreased… which should cause a decrease in biogenic OA due to partitioning effects.” This is not necessarily true. Less anthropogenic VOCs will result in less competition for OH and other radicals. Biogenic VOCs that may have blown out of the domain in the older time period simulations, may have opportunity to react and form semi-volatile species in the newer time period simulation.
This is a valid point. We have rephrased this section mentioning the various effects that the change in anthropogenic VOCs can have on biogenic SOA. The partitioning effect due to the reduction of anthropogenic SOA is only one of these effects. There are of course other competitive effects at play in this system and a lot of them (e.g., the effect on OH mentioned) are accounted for in the SOA and gas-phase chemistry of the CTM.
Citation: https://doi.org/10.5194/acp-2022-648-AC1
-
RC2: 'Comment on acp-2022-648', Anonymous Referee #2, 25 Oct 2022
Here the authors explore organic aerosol over the southeast United States, an area well known for especially high biogenic emissions. Using the chemical transport model PMCAMx, the authors simulate two summers, one in 2001 and the other in 2010, and compare conditions and results between them. The effects of various aspects of the overall SOA mechanism are discussed, and in some cases quantified. Manuscript text is clear and well composed, and figures are generally effective. However, while the topic itself is worthwhile and deserving of study, in my opinion this work is fundamentally incomplete, lacking a central question and failing to produce any novel conclusions. I do believe there are opportunities available to further develop this manuscript with additional modeling work and analyses, but in its current form I do not support its publication in ACP. Some specific big picture concerns follow.
- Years modeled: By choosing to examine only two total years, the authors limit their ability to draw meaningful conclusions regarding trends and variability. While its true that changing anthropogenic emission inventories in particular will show a strong signal when comparing between these years, interannual variability in underlying dynamics must be assumed to impact and confound those anthropogenic changes as well as those of meteorology-dependent biogenic emissions. To better understand trends in emissions and resulting OA concentrations, a more robust temporal domain (for example including the years between 2001 and 2010 as well) is recommended to help resolve some of these influences.
- Role and accuracy of meteorological variability: Meteorology and dynamics are inadequately addressed here in general, with mostly qualitative descriptions covering this extremely influential driver of differences in modeled output. Significantly more work here is necessary to better understand exactly how meteorology is affecting both biogenic emissions and precursor transport. Along these same lines, while significant attention is paid to observational comparisons of modeled aerosol, none is given to the WRF output driving emissions and transport. Considering their significance, and the novelty of the dynamics generated here to drive the CTM, this is a glaring omission and worthy of considerable evaluation
- Significance and novelty of conclusions: Perhaps most importantly, on the whole I really struggle to find a key takeaway message contained here. The model output is well presented, with clear maps and figures comparing the two examined years, but I see nothing surprising or helpful in terms of advancing the state of knowledge on the region or on the modeling of OA in general. The NOx-dependence of SOA yield is removed as a case study, but the reason and value for this is completely unclear to me. It seems trivially obvious to me that keeping only the low-NOx oxidation pathway would increase yields and overall concentrations, and I see no need to confirm this expected result. A major rethinking of what questions these simulations are intended to answer is necessary if this work is to make a meaningful contribution to the SOA modeling literature.
Smaller issues and questions:
- Lines 64-65: "Plenty of uncertainty still exists regarding the role of isoprene in SOA formation." This is a confusingly broad and poorly explained statement.
- Lines 135-145: The relevant bins and species included in the model are difficult to parse. For example, I couldn't tell for sure whether there was one species for monoterpenes and one for sesquiterpenes, or just a single species representing both together. A schematic or diagram would be very helpful here.
- Two days seems to me to be an unacceptably short spinup time, considering the lifetimes of relevant trace gases and precursors.
- The authors note that IMPROVE measurement comparisons were heavily influenced by fires on several days in the modeled domain. Does this mean that biomass burning in general is not included? If they are included, but simply lacked information on those specific fires, more background information on the inventory used is appropriate. If they are not included, this strikes me as a significant problem that should be addressed.
Citation: https://doi.org/10.5194/acp-2022-648-RC2 -
AC2: 'Response to Referee #2', Spyros Pandis, 12 Dec 2022
(1) Here the authors explore organic aerosol over the southeast United States, an area well known for especially high biogenic emissions. Using the chemical transport model PMCAMx, the authors simulate two summers, one in 2001 and the other in 2010, and compare conditions and results between them. The effects of various aspects of the overall SOA mechanism are discussed, and in some cases quantified. Manuscript text is clear and well composed, and figures are generally effective. However, while the topic itself is worthwhile and deserving of study, in my opinion this work is fundamentally incomplete, lacking a central question and failing to produce any novel conclusions. I do believe there are opportunities available to further develop this manuscript with additional modeling work and analyses, but in its current form I do not support its publication in ACP. Some specific big picture concerns follow.
We do appreciate the suggestions of the reviewer and we do agree that the central question of the paper was not made clear in the original paper. This question is: “can a chemical transport model including a complete description of gas-phase chemistry that is standard for CTMs, secondary organic formation based on the volatility basis set (semivolatile partitioning effects, NOx effects on SOA yields, temperature effects, aging reactions) and interactions between primary and secondary SOA based on partitioning theory reproduce in a satisfactory degree the observed OA changes in an environment that is dominated by biogenic emissions and in which significant changes of all anthropogenic emissions have taken place?”. There are of course a lot of other directions that could be pursued in this topic, but we believe that the question is intriguing enough and the insights gained by our work valuable enough for the paper to deserve publication after the recommended improvements.
Our responses and changes to the manuscript (in regular font) follow the comments of the referee (in italics).
(2) Years modeled: By choosing to examine only two total years, the authors limit their ability to draw meaningful conclusions regarding trends and variability. While it is true that changing anthropogenic emission inventories in particular will show a strong signal when comparing between these years, interannual variability in underlying dynamics must be assumed to impact and confound those anthropogenic changes as well as those of meteorology-dependent biogenic emissions. To better understand trends in emissions and resulting OA concentrations, a more robust temporal domain (for example including the years between 2001 and 2010 as well) is recommended to help resolve some of these influences.
More years is always better with this type of simulation work. Unfortunately, there are always tradeoffs that come with increasing the number of years that are simulates, the key ones being the development of emissions inventories, meteorology, and resources (computational and human time). As mentioned, the work is a follow up on the study by Skyllakou et al. (2021) which investigated changes in PM2.5 and its sources among 1990, 2001, and 2010. Seeking to explain bSOA changes in more detail, we did a deeper dive into the bSOA predictions and focused on the Southeastern US summer due to the prevalence of biogenics here. 1990 was an obvious additional year to use. Unfortunately, measurements during this time period are severely lacking leaving the corresponding analysis meaningless. We do agree that the investigation of the year to year variability of both OA observations and model predictions is a worthwhile scientific objective. We do believe though that it is beyond the scope of the current work. It could be addressed in future studies with the present or other CTMs. We do not believe that including more intermediate years would change our current conclusions. It would just open another set of issues (e.g., the ability of the meteorological model to simulate a specific summer or the uncertainty in reproducing wild fire emissions in another). A discussion of our choice of simulation years and the effects that it may have on the implications of our work, together with suggestions for future work have been added to the revised paper.
(3) Role and accuracy of meteorological variability: Meteorology and dynamics are inadequately addressed here in general, with mostly qualitative descriptions covering this extremely influential driver of differences in modeled output. Significantly more work here is necessary to better understand exactly how meteorology is affecting both biogenic emissions and precursor transport. Along these same lines, while significant attention is paid to observational comparisons of modeled aerosol, none is given to the WRF output driving emissions and transport. Considering their significance, and the novelty of the dynamics generated here to drive the CTM, this is a glaring omission and worthy of considerable evaluation.
We do agree with the reviewer about the importance of the meteorological variability between the examined years for OA in the region. To better address the corresponding effects we first include in the paper an analysis and discussion of the effects of meteorology on the emissions of biogenic VOCs. On the topic of transport, we have performed an additional simulation to quantify its role on the predicted OA levels. In this test we have used the 2001 emissions and 2010 meteorology. Negligible change in predicted bSOA was observed in this test. The WRF performance was similar to corresponding applications in other studies. A more detailed discussion of these issues has been added to the revised paper.
(4) Significance and novelty of conclusions: Perhaps most importantly, on the whole I really struggle to find a key takeaway message contained here. The model output is well presented, with clear maps and figures comparing the two examined years, but I see nothing surprising or helpful in terms of advancing the state of knowledge on the region or on the modeling of OA in general. The NOx-dependence of SOA yield is removed as a case study, but the reason and value for this is completely unclear to me. It seems trivially obvious to me that keeping only the low-NOx oxidation pathway would increase yields and overall concentrations, and I see no need to confirm this expected result. A major rethinking of what questions these simulations are intended to answer is necessary if this work is to make a meaningful contribution to the SOA modeling literature.
We do understand from the comments of both reviewers that the hypothesis tested by this work and its implications were not made entirely clear. Our central hypothesis can be phrased as: a chemical transport model including a complete description of gas-phase chemistry that is standard for CTMs, secondary organic formation based on the volatility basis set (semi-volatile partitioning effects, NOx effects on SOA yields, temperature effects, aging reactions) and interactions between primary and secondary SOA based on partitioning theory can reproduce in a satisfactory degree the observed OA changes in an environment that is dominated by biogenic emissions and in which significant changes of all anthropogenic emissions have taken place. The simulations performed provide significant support to the hypothesis.
To quantify the effects that the major processes affecting bSOA concentrations that are currently in the model, we formulated sensitivity tests, including the NOx dependence of the SOA yields mentioned in this comment. These sensitivity tests are meant to clearly illustrate how these levers are pulled in different scenarios, and their contributions to the bSOA prediction. The study is meant to show that these key levers alone produce consistent predictions of bSOA in the Southeast US in response to changes in both anthropogenic and biogenic emissions, and the sensitivity tests are meant to illustrate how the key processes do their job in the model. Even if the results of a single test may seem trivial to some, we aim for transparency with regards to the SOA formation mechanisms in the model and ultimately seek to identify if additional detail is needed.
To address these issues we have rewritten parts of both the Introduction of the paper and the discussion in the Conclusions in an effort to address the comment of the reviewer.
(5) Lines 64-65: "Plenty of uncertainty still exists regarding the role of isoprene in SOA formation." This is a confusingly broad and poorly explained statement.
This is a valid point. We have rewritten the statements with additional detail to increase its clarity.
(6) Lines 135-145: The relevant bins and species included in the model are difficult to parse. For example, I couldn't tell for sure whether there was one species for monoterpenes and one for sesquiterpenes, or just a single species representing both together. A schematic or diagram would be very helpful here.
There is a single lumped species representing monoterpenes and a second lumped species representing lumped sesquiterpenes (and of course isoprene is its own chemical species in the model as well). There are several bins then for the SOA products (both in the gas and particle phase) which include the products of the oxidation of the aforementioned VOCs. A more detailed description of this aspect of the model has been added to the manuscript.
(7) Two days seems to me to be an unacceptably short spin-up time, considering the lifetimes of relevant trace gases and precursors.
The spin-up time in regional models is usually determined by the residence time of the various pollutants inside the modeling domain and not by their lifetimes. Given the size of our domain and the average wind speed during the simulation period the two days were sufficient to “wash out” most of the initial conditions and to allow the emissions and meteorology to dominate the predicted concentrations. We have tested this for the Eastern US using the Particle Source Apportionment Technology (PSAT) in which the effect of the initial conditions in each simulation is explicitly simulated. This point is now discussed in the paper and the corresponding references are provided.
(8) The authors note that IMPROVE measurement comparisons were heavily influenced by fires on several days in the modeled domain. Does this mean that biomass burning in general is not included? If they are included, but simply lacked information on those specific fires, more background information on the inventory used is appropriate. If they are not included, this strikes me as a significant problem that should be addressed.
Biomass burning is of course included in the simulations. However, there was a major fire near a particular sampling site during the 2001 simulated period, that affected dramatically the corresponding OA observations and also the period average. Our analysis showed that the model failed to reproduce the extremely high OA levels observed. This could be due to an underestimation of the emissions or an error in wind direction (the plume can miss the sampling site in the model). Given that this issue was not related to bSOA, we did not include that specific data point in the evaluation. We now explain in more detail this point to avoid misunderstandings.
Citation: https://doi.org/10.5194/acp-2022-648-AC2
Status: closed
-
RC1: 'Comment on acp-2022-648', Anonymous Referee #1, 24 Oct 2022
Review of “Predicted and Observed Changes in Summertime Biogenic and Total Organic
Aerosol in the Southeast United States from 2001 to 2010” by Dinkelacker et al.
This manuscript describes application of the air quality model PMCAMx to the southeastern United States from 2001 to 2010. The authors describe changes in predicted organic aerosol over the time period in response to their simulated changes in anthropogenic emissions of SO2, NOx and NMVOCs, in addition to biogenic emissions of isoprene, monoterpenes, and sesquiterpenes. The authors find that NOx-dependence and partitioning have the largest influence on the anthropogenic controls on bSOA. It’s my opinion this conclusion is not robustly defended and I cannot recommend publication in the present form.
My two most pressing overarching comments:
- The discussion of RH dependence for SOA yield in the Introduction is lacking. Further, the lack of water pathways in the SOA model, and sole reliance on semi-volatile partitioning … and then finding that semi-volatile partitioning in the predominant controlling mechanism lacks context and the reasoning is flawed.
To my knowledge, Kamens was the first to show that changing RH in a laboratory smog chamber experiment does not necessarily change the amount of liquid water, and that water is the overriding controlling factor. The Kamens paper cited here has liquid water in the title.
For example, in the cited Nguyen paper, there was no seed in the smog chamber experiments. They found no dependence on RH over a wide range. A dramatic change in RH in those experiments did not change liquid water changed (except maybe on the walls?). Nguyen went on to find in later controlled smog chamber experiments with inorganic seeds, where liquid water in aerosols did change, that liquid water for bSOA formation had the predominant effect and in fact seemed to be necessary. Water mattered much more so than pH in those experiments. https://acp.copernicus.org/articles/14/3497/2014/
It’s odd that the water context is not discussed, in particular because the authors want to argue that partitioning is the dominant controlling mechanism, and because water pathways – neglected in this model application - are expected to contribute substantially to SOA in the region according to Carlton and Turpin: https://acp.copernicus.org/articles/13/10203/2013/acp-13-10203-2013.html
- The authors state in multiple locations that they compared their OA predictions to OA from the CSN and IMPROVE networks. This is highly problematic because neither measures OA, they measure OC. Did the authors mistakenly compare their OA predictions to OC measurements or did they use an OM:OC ratio? If they used a conversion factor what was it and was it constant? It is well established that this ratio is changing. Malm finds that the ratio is increasing. https://doi.org/10.1029/2019JD031480. Hand et al., http://vista.cira.colostate.edu/improve/wp-content/uploads/2019/03/Hand2019.pdf finds that RH problems in the laboratory are likely impacting mass measurements. Depending on the OM:OC ratio method employed, this could impact OA estimates. The authors provide no information on this, and it is difficult to surmise what potential impacts could be.
specific comments:
Line 86: could be primary
Application of MEGAN: did the land description change at all over the time period and if so, how was this reflected? If not, how is that choice justified?
Starting at Line 185: the authors state: “Anthropogenic VOC emissions decreased… which should cause a decrease in biogenic OA due to partitioning effects.” This is not necessarily true. Less anthropogenic VOCs will result in less competition for OH and other radicals. Biogenic VOCs that may have blown out of the domain in the older time period simulations, may have opportunity to react and form semi-volatile species in the newer time period simulation.
Citation: https://doi.org/10.5194/acp-2022-648-RC1 -
AC1: 'Response to Referee #1', Spyros Pandis, 12 Dec 2022
(1) This manuscript describes application of the air quality model PMCAMx to the southeastern United States from 2001 to 2010. The authors describe changes in predicted organic aerosol over the time period in response to their simulated changes in anthropogenic emissions of SO2, NOx and NMVOCs, in addition to biogenic emissions of isoprene, monoterpenes, and sesquiterpenes. The authors find that NOx-dependence and partitioning have the largest influence on the anthropogenic controls on bSOA. It’s my opinion this conclusion is not robustly defended and I cannot recommend publication in the present form.
We appreciate the effort devoted by the referee to review our manuscript. In this paper we do show that a chemical transport model that includes the effects of NOx on the gas-phase chemistry of an environment that is dominated by biogenic VOCs and the changes in the partitioning of semivolatile organic aerosol can explain reasonably well the observed changes in total organic aerosol in this area. The study investigates changes over a decade in which anthropogenic emissions of SO2, NOx, VOCs and organic aerosol took place. We do understand the argument of the referee that this does not necessarily prove that the simulated processes are the dominant ones. There can be scenarios in which two processes can have the same effect (e.g., lead to the oxidation of an organic molecule and transfer it to the particulate phase) and simulating just one of them is sufficient to reproduce observations. To address this issue we have rephrased this specific conclusion to explain that a model like PMCAMx simulating specific processes can reproduce the observed changes in this region that is rich in biogenic VOCs and SOA, without getting into the argument about the dominant processes.
Our responses and the corresponding changes (in regular font) follow the comments of the reviewer (in italics).
(2) The discussion of RH dependence for SOA yield in the Introduction is lacking. Further, the lack of water pathways in the SOA model, and sole reliance on semi-volatile partitioning … and then finding that semi-volatile partitioning in the predominant controlling mechanism lacks context and the reasoning is flawed. To my knowledge, Kamens was the first to show that changing RH in a laboratory smog chamber experiment does not necessarily change the amount of liquid water, and that water is the overriding controlling factor. The Kamens paper cited here has liquid water in the title. For example, in the cited Nguyen paper, there was no seed in the smog chamber experiments. They found no dependence on RH over a wide range. A dramatic change in RH in those experiments did not change liquid water changed (except maybe on the walls?). Nguyen went on to find in later controlled smog chamber experiments with inorganic seeds, where liquid water in aerosols did change, that liquid water for bSOA formation had the predominant effect and in fact seemed to be necessary. Water mattered much more so than pH in those experiments. It’s odd that the water context is not discussed, in particular because the authors want to argue that partitioning is the dominant controlling mechanism, and because water pathways – neglected in this model application - are expected to contribute substantially to SOA in the region according to Carlton and Turpin.
We have followed the suggestion of the referee and extended the discussion of the RH dependence on the SOA in the Introduction section. We have also added discussion of previous work regarding the formation of SOA in the aqueous phase. In both cases all references suggested by the reviewer are now cited. We then explain that the version of PMCAMx used in this work does not include these effects that have been observed in laboratory experiments and therefore any significant discrepancies between model predictions and observations could be due to the lack of simulation of these effects.
We also recognize that the explanation of the objectives of the present work may have not been clear enough in the original manuscript. There have been numerous laboratory studies suggesting that a process may affect SOA formation and therefore OA levels in the atmosphere. These include the water related processes discussed above, pH effects, catalytic effects, photo-degradation reactions, oligomerization reactions, etc. Our approach in this paper was to simplify (instead of complicating) the model used and test if this simpler model version can explain the observed changes. The main drivers of SOA formation in the CTM version used are precursor emissions, SOA yields (with empirically derived NOx-level dependence), NOx and VOC effects on gas-phase chemistry, semivolatile partitioning and chemical aging reactions as described by the VBS. We do recognize that there is always additional science that one could implemented in a CTM for an increased level of mechanistic accuracy. As stated in the manuscript, performance was consistent between the two periods even with significant changes in emissions. The findings suggest that the main drivers already accounted for in the model respond accordingly to those changes. We of course agree that there are processes affecting SOA formation in the atmosphere that are not accounted for in this version of the CTM - and there likely always will be. This discussion has been added to the revised manuscript.
(3) The authors state in multiple locations that they compared their OA predictions to OA from the CSN and IMPROVE networks. This is highly problematic because neither measures OA, they measure OC. Did the authors mistakenly compare their OA predictions to OC measurements or did they use an OM:OC ratio? If they used a conversion factor what was it and was it constant? It is well established that this ratio is changing. Malm finds that the ratio is increasing. Hand et al., finds that RH problems in the laboratory are likely impacting mass measurements. Depending on the OM:OC ratio method employed, this could impact OA estimates. The authors provide no information on this, and it is difficult to surmise what potential impacts could be.
We are well aware of difference between OC and OM and the need to covert the OC measurements to OM (or vice versa) for model evaluation. We do agree with the reviewer though that additional information is needed for this part of the analysis. We do explain in the revised manuscript the OM:OC ratios used for the conversion (they are different for the urban and the rural sites) and provide the corresponding references. It should be noted that the IMRPOVE network does report OA measurements now using the OM:OC values used in our work. We should also point out that significant effort was put into verifying the OC measurements, taking care to sort out inconsistencies in artifact corrections for historical data corresponding to different measurement codes for CSN OC. These corrections were applied appropriately to the corresponding measurement codes. A discussion of this issue has been added to the revised paper.
(4) Line 86: could be primary.
This is a valid point. We have rephrased the sentence to include the possibility that it may be primary anthropogenic OA too.
(5) Application of MEGAN: did the land description change at all over the time period and if so, how was this reflected? If not, how is that choice justified?
The differences in biogenic emissions predicted with MEGAN3 were driven by the differences in WRF predicted meteorology. There were only small changes in the land use change in the overall area over the relatively short period (a decade) between the two simulations. This is now explained in the revised paper.
(6) Starting at Line 185: the authors state: “Anthropogenic VOC emissions decreased… which should cause a decrease in biogenic OA due to partitioning effects.” This is not necessarily true. Less anthropogenic VOCs will result in less competition for OH and other radicals. Biogenic VOCs that may have blown out of the domain in the older time period simulations, may have opportunity to react and form semi-volatile species in the newer time period simulation.
This is a valid point. We have rephrased this section mentioning the various effects that the change in anthropogenic VOCs can have on biogenic SOA. The partitioning effect due to the reduction of anthropogenic SOA is only one of these effects. There are of course other competitive effects at play in this system and a lot of them (e.g., the effect on OH mentioned) are accounted for in the SOA and gas-phase chemistry of the CTM.
Citation: https://doi.org/10.5194/acp-2022-648-AC1
-
RC2: 'Comment on acp-2022-648', Anonymous Referee #2, 25 Oct 2022
Here the authors explore organic aerosol over the southeast United States, an area well known for especially high biogenic emissions. Using the chemical transport model PMCAMx, the authors simulate two summers, one in 2001 and the other in 2010, and compare conditions and results between them. The effects of various aspects of the overall SOA mechanism are discussed, and in some cases quantified. Manuscript text is clear and well composed, and figures are generally effective. However, while the topic itself is worthwhile and deserving of study, in my opinion this work is fundamentally incomplete, lacking a central question and failing to produce any novel conclusions. I do believe there are opportunities available to further develop this manuscript with additional modeling work and analyses, but in its current form I do not support its publication in ACP. Some specific big picture concerns follow.
- Years modeled: By choosing to examine only two total years, the authors limit their ability to draw meaningful conclusions regarding trends and variability. While its true that changing anthropogenic emission inventories in particular will show a strong signal when comparing between these years, interannual variability in underlying dynamics must be assumed to impact and confound those anthropogenic changes as well as those of meteorology-dependent biogenic emissions. To better understand trends in emissions and resulting OA concentrations, a more robust temporal domain (for example including the years between 2001 and 2010 as well) is recommended to help resolve some of these influences.
- Role and accuracy of meteorological variability: Meteorology and dynamics are inadequately addressed here in general, with mostly qualitative descriptions covering this extremely influential driver of differences in modeled output. Significantly more work here is necessary to better understand exactly how meteorology is affecting both biogenic emissions and precursor transport. Along these same lines, while significant attention is paid to observational comparisons of modeled aerosol, none is given to the WRF output driving emissions and transport. Considering their significance, and the novelty of the dynamics generated here to drive the CTM, this is a glaring omission and worthy of considerable evaluation
- Significance and novelty of conclusions: Perhaps most importantly, on the whole I really struggle to find a key takeaway message contained here. The model output is well presented, with clear maps and figures comparing the two examined years, but I see nothing surprising or helpful in terms of advancing the state of knowledge on the region or on the modeling of OA in general. The NOx-dependence of SOA yield is removed as a case study, but the reason and value for this is completely unclear to me. It seems trivially obvious to me that keeping only the low-NOx oxidation pathway would increase yields and overall concentrations, and I see no need to confirm this expected result. A major rethinking of what questions these simulations are intended to answer is necessary if this work is to make a meaningful contribution to the SOA modeling literature.
Smaller issues and questions:
- Lines 64-65: "Plenty of uncertainty still exists regarding the role of isoprene in SOA formation." This is a confusingly broad and poorly explained statement.
- Lines 135-145: The relevant bins and species included in the model are difficult to parse. For example, I couldn't tell for sure whether there was one species for monoterpenes and one for sesquiterpenes, or just a single species representing both together. A schematic or diagram would be very helpful here.
- Two days seems to me to be an unacceptably short spinup time, considering the lifetimes of relevant trace gases and precursors.
- The authors note that IMPROVE measurement comparisons were heavily influenced by fires on several days in the modeled domain. Does this mean that biomass burning in general is not included? If they are included, but simply lacked information on those specific fires, more background information on the inventory used is appropriate. If they are not included, this strikes me as a significant problem that should be addressed.
Citation: https://doi.org/10.5194/acp-2022-648-RC2 -
AC2: 'Response to Referee #2', Spyros Pandis, 12 Dec 2022
(1) Here the authors explore organic aerosol over the southeast United States, an area well known for especially high biogenic emissions. Using the chemical transport model PMCAMx, the authors simulate two summers, one in 2001 and the other in 2010, and compare conditions and results between them. The effects of various aspects of the overall SOA mechanism are discussed, and in some cases quantified. Manuscript text is clear and well composed, and figures are generally effective. However, while the topic itself is worthwhile and deserving of study, in my opinion this work is fundamentally incomplete, lacking a central question and failing to produce any novel conclusions. I do believe there are opportunities available to further develop this manuscript with additional modeling work and analyses, but in its current form I do not support its publication in ACP. Some specific big picture concerns follow.
We do appreciate the suggestions of the reviewer and we do agree that the central question of the paper was not made clear in the original paper. This question is: “can a chemical transport model including a complete description of gas-phase chemistry that is standard for CTMs, secondary organic formation based on the volatility basis set (semivolatile partitioning effects, NOx effects on SOA yields, temperature effects, aging reactions) and interactions between primary and secondary SOA based on partitioning theory reproduce in a satisfactory degree the observed OA changes in an environment that is dominated by biogenic emissions and in which significant changes of all anthropogenic emissions have taken place?”. There are of course a lot of other directions that could be pursued in this topic, but we believe that the question is intriguing enough and the insights gained by our work valuable enough for the paper to deserve publication after the recommended improvements.
Our responses and changes to the manuscript (in regular font) follow the comments of the referee (in italics).
(2) Years modeled: By choosing to examine only two total years, the authors limit their ability to draw meaningful conclusions regarding trends and variability. While it is true that changing anthropogenic emission inventories in particular will show a strong signal when comparing between these years, interannual variability in underlying dynamics must be assumed to impact and confound those anthropogenic changes as well as those of meteorology-dependent biogenic emissions. To better understand trends in emissions and resulting OA concentrations, a more robust temporal domain (for example including the years between 2001 and 2010 as well) is recommended to help resolve some of these influences.
More years is always better with this type of simulation work. Unfortunately, there are always tradeoffs that come with increasing the number of years that are simulates, the key ones being the development of emissions inventories, meteorology, and resources (computational and human time). As mentioned, the work is a follow up on the study by Skyllakou et al. (2021) which investigated changes in PM2.5 and its sources among 1990, 2001, and 2010. Seeking to explain bSOA changes in more detail, we did a deeper dive into the bSOA predictions and focused on the Southeastern US summer due to the prevalence of biogenics here. 1990 was an obvious additional year to use. Unfortunately, measurements during this time period are severely lacking leaving the corresponding analysis meaningless. We do agree that the investigation of the year to year variability of both OA observations and model predictions is a worthwhile scientific objective. We do believe though that it is beyond the scope of the current work. It could be addressed in future studies with the present or other CTMs. We do not believe that including more intermediate years would change our current conclusions. It would just open another set of issues (e.g., the ability of the meteorological model to simulate a specific summer or the uncertainty in reproducing wild fire emissions in another). A discussion of our choice of simulation years and the effects that it may have on the implications of our work, together with suggestions for future work have been added to the revised paper.
(3) Role and accuracy of meteorological variability: Meteorology and dynamics are inadequately addressed here in general, with mostly qualitative descriptions covering this extremely influential driver of differences in modeled output. Significantly more work here is necessary to better understand exactly how meteorology is affecting both biogenic emissions and precursor transport. Along these same lines, while significant attention is paid to observational comparisons of modeled aerosol, none is given to the WRF output driving emissions and transport. Considering their significance, and the novelty of the dynamics generated here to drive the CTM, this is a glaring omission and worthy of considerable evaluation.
We do agree with the reviewer about the importance of the meteorological variability between the examined years for OA in the region. To better address the corresponding effects we first include in the paper an analysis and discussion of the effects of meteorology on the emissions of biogenic VOCs. On the topic of transport, we have performed an additional simulation to quantify its role on the predicted OA levels. In this test we have used the 2001 emissions and 2010 meteorology. Negligible change in predicted bSOA was observed in this test. The WRF performance was similar to corresponding applications in other studies. A more detailed discussion of these issues has been added to the revised paper.
(4) Significance and novelty of conclusions: Perhaps most importantly, on the whole I really struggle to find a key takeaway message contained here. The model output is well presented, with clear maps and figures comparing the two examined years, but I see nothing surprising or helpful in terms of advancing the state of knowledge on the region or on the modeling of OA in general. The NOx-dependence of SOA yield is removed as a case study, but the reason and value for this is completely unclear to me. It seems trivially obvious to me that keeping only the low-NOx oxidation pathway would increase yields and overall concentrations, and I see no need to confirm this expected result. A major rethinking of what questions these simulations are intended to answer is necessary if this work is to make a meaningful contribution to the SOA modeling literature.
We do understand from the comments of both reviewers that the hypothesis tested by this work and its implications were not made entirely clear. Our central hypothesis can be phrased as: a chemical transport model including a complete description of gas-phase chemistry that is standard for CTMs, secondary organic formation based on the volatility basis set (semi-volatile partitioning effects, NOx effects on SOA yields, temperature effects, aging reactions) and interactions between primary and secondary SOA based on partitioning theory can reproduce in a satisfactory degree the observed OA changes in an environment that is dominated by biogenic emissions and in which significant changes of all anthropogenic emissions have taken place. The simulations performed provide significant support to the hypothesis.
To quantify the effects that the major processes affecting bSOA concentrations that are currently in the model, we formulated sensitivity tests, including the NOx dependence of the SOA yields mentioned in this comment. These sensitivity tests are meant to clearly illustrate how these levers are pulled in different scenarios, and their contributions to the bSOA prediction. The study is meant to show that these key levers alone produce consistent predictions of bSOA in the Southeast US in response to changes in both anthropogenic and biogenic emissions, and the sensitivity tests are meant to illustrate how the key processes do their job in the model. Even if the results of a single test may seem trivial to some, we aim for transparency with regards to the SOA formation mechanisms in the model and ultimately seek to identify if additional detail is needed.
To address these issues we have rewritten parts of both the Introduction of the paper and the discussion in the Conclusions in an effort to address the comment of the reviewer.
(5) Lines 64-65: "Plenty of uncertainty still exists regarding the role of isoprene in SOA formation." This is a confusingly broad and poorly explained statement.
This is a valid point. We have rewritten the statements with additional detail to increase its clarity.
(6) Lines 135-145: The relevant bins and species included in the model are difficult to parse. For example, I couldn't tell for sure whether there was one species for monoterpenes and one for sesquiterpenes, or just a single species representing both together. A schematic or diagram would be very helpful here.
There is a single lumped species representing monoterpenes and a second lumped species representing lumped sesquiterpenes (and of course isoprene is its own chemical species in the model as well). There are several bins then for the SOA products (both in the gas and particle phase) which include the products of the oxidation of the aforementioned VOCs. A more detailed description of this aspect of the model has been added to the manuscript.
(7) Two days seems to me to be an unacceptably short spin-up time, considering the lifetimes of relevant trace gases and precursors.
The spin-up time in regional models is usually determined by the residence time of the various pollutants inside the modeling domain and not by their lifetimes. Given the size of our domain and the average wind speed during the simulation period the two days were sufficient to “wash out” most of the initial conditions and to allow the emissions and meteorology to dominate the predicted concentrations. We have tested this for the Eastern US using the Particle Source Apportionment Technology (PSAT) in which the effect of the initial conditions in each simulation is explicitly simulated. This point is now discussed in the paper and the corresponding references are provided.
(8) The authors note that IMPROVE measurement comparisons were heavily influenced by fires on several days in the modeled domain. Does this mean that biomass burning in general is not included? If they are included, but simply lacked information on those specific fires, more background information on the inventory used is appropriate. If they are not included, this strikes me as a significant problem that should be addressed.
Biomass burning is of course included in the simulations. However, there was a major fire near a particular sampling site during the 2001 simulated period, that affected dramatically the corresponding OA observations and also the period average. Our analysis showed that the model failed to reproduce the extremely high OA levels observed. This could be due to an underestimation of the emissions or an error in wind direction (the plume can miss the sampling site in the model). Given that this issue was not related to bSOA, we did not include that specific data point in the evaluation. We now explain in more detail this point to avoid misunderstandings.
Citation: https://doi.org/10.5194/acp-2022-648-AC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 923 | 191 | 43 | 1,157 | 73 | 35 | 59 |
- HTML: 923
- PDF: 191
- XML: 43
- Total: 1,157
- Supplement: 73
- BibTeX: 35
- EndNote: 59
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1