the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 19 Sep 2024
Construction and Application of a Pollen Emissions Model based on Phenology and Random Forests
Abstract. In recent years, the intensification of global climate change and environmental pollution has led to a marked increase in pollen-induced allergic diseases. This study leverages 16 years of continuous pollen monitoring data, alongside meteorological factors and plant functional type data, to construct a pollen emissions model using phenology and random forests (RF). This model is then employed to simulate the emission characteristics of three primary types of autumn pollen (Artemisia, Chenopod, and total pollen concentration), elucidating the emission patterns throughout the seasonal cycle in Beijing. Phenology and RF precisely simulate the start and end day of year of pollen, as well as the annual pollen production. There are significant spatiotemporal differences among the three types of pollen. On average, pollen dispersal begins around August 10, peaks around August 30, and concludes by September 25, with a dispersal period lasting approximately 45 days. Furthermore, the relationship between pollen emissions and meteorological factors is investigated, revealing that temperature, relative humidity (RH), and sunshine hours (SSH) significantly influence annual pollen emissions. Specifically, temperature and RH exhibit a strong positive correlation with annual pollen emissions, while SSH shows a negative correlation. Different pollen types display varied responses to meteorological factors. Finally, the constructed pollen emissions model is integrated into RegCM and validated using pollen observation data, confirming its reliability in predicting pollen concentrations. This study not only enhances the understanding of pollen release mechanisms but also provides scientific evidence for the selection and planting of urban greening plants.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(2750 KB)
-
Supplement
(580 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(2750 KB) - Metadata XML
-
Supplement
(580 KB) - BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-2000', Anonymous Referee #1, 02 Nov 2024
This work by Li et al., has successfully constructed a pollen emission model based on phenology and random forest algorithm based on the 16 years of pollen observation data observed in Beijing. Moreover, the impact of temperature and relative humidity on the pollen emission has been reported. Furtherly, this pollen emission model was integrated into the regional climate model to validate the reliability of this model. The work presents a useful pollen prediction model for future prediction.
General points:
- why the data was stopped at 2021? Could this model be used to predict those for year 2022, 2023?
- there are too many figures and tables in this manuscript. I would suggest the authors to move some of them into the supporting information.
- previous literatures on pollen prediction model should be introduced in the introduction. How about the advancement of the current model when compared to those in the literatures?
Minor points:
Line 31: RegCM. The first time it appeared, the full name should be appeared first. Similar for sDOY, eDOY in lines 134-135.
Lines 58-62: references should be provided.
Line 95, Line 103: what is the difference between numerical model and pollen emission model? It would be better to clearly explain it here.
Line 123: how was this formula obtained? Was this a novel formula or from literatures?
Line 147: what is the meaning of Rs1, Rs2, Rssig?
Line 182: Has the RF algorithm been used for pollen emission simulation in the previous literatures?
Section 3.1.1: was temperature the only variable in this autumn phenology model?
Lines 548-551: please provide more details or discussions regarding the linkage between pollen amount and climate change.
Lines 617-625: it seems that basic information regarding the RegCM should be in the introduction section. It might be better for the authors to concise the current introduction and move such information into the introduction also.
Section 3.3.1: some of them should be in the introduction section, and some may be better be in the method section.
Citation: https://doi.org/10.5194/egusphere-2024-2000-RC1 -
AC1: 'Reply on RC1', Jiangtao Li, 13 Nov 2024
Thank you for your detailed and insightful comments on our manuscript. We have addressed all of your concerns and questions in the attached PDF. In the document, your review comments are displayed in black text, and our responses and corresponding revisions are highlighted in red. Once again, we sincerely appreciate your valuable feedback.
-
RC2: 'Comment on egusphere-2024-2000', Anonymous Referee #3, 20 Nov 2024
Predicting allergenic pollen concentrations is important for urban public health management in the context of climate change. This study attempted to construct a pollen emission model using phenology and random forests, which was further integrated into RegCM to predict pollen concentrations driven by meteorological forecasts. Although the topic is important, several concerns prevent me from recommending this paper for publication in ACP. First, the novelty of the method is insufficient. The authors claim that the use of phenology and RF is a major novelty of this study, but previous studies such as Lo et al. (2018) in North America, Shokouhi et al. (2024) in Europe, and Huete et al. (2019) in southeastern Australia have all considered phenology and used RF models for pollen forecasting. Therefore, the novelty of this study is questionable. Second, plant functional type is an important information to determine the source of pollen emission, especially the partitioning of grasses into cool season C3 and warm season C4 components. For an urban study such as this one, the spatial resolution of the PFT data should be sufficient to resolve the spatial heteorogeneity of the PFT (C3/C4 partitioning). Instead of searching for a fine-scale local PFT database and validating it against in-situ measurements, the authors used a coarse-resolution (~5 km) database from NCAR's CLM, originally developed for global simulations. Uncertainties associated with the coarse-resolution PFT data are not rigorously considered or critically discussed in the error propagation. Finally, the quality of the figures presented is generally poor, often lacking critical information for the reader to appreciate the figure (e.g. Fig. 11, the reason for using two identical y-axes but with different scales is not mentioned in the figure caption, very difficult to understand). As such, I think the quality and novelty of this study does not meet the standard of ACP.
Citation: https://doi.org/10.5194/egusphere-2024-2000-RC2 -
AC2: 'Reply on RC2', Jiangtao Li, 02 Dec 2024
We sincerely appreciate the time and effort you have dedicated to reviewing our manuscript and providing valuable feedback. We are deeply sorry that our initial submission did not meet your expectations in certain areas. Your insightful comments are of great importance to us, and we have carefully considered each point raised. Below, we provide a detailed response to your concerns:
（1）On the novelty of our methodology
We value your comments regarding the novelty of our study and have reviewed the references you mentioned (Lo et al., 2018; Shokouhi et al., 2024; Huete et al., 2019). While acknowledging the contributions of these studies, we believe our work has distinct differences:
Lo et al. (2018) developed pollen calendars for North America using data from 31 National Allergy Bureau stations, focusing on analyzing the start, end, and duration of pollen seasons and annual pollen integrals for various taxa. In contrast, our study centers on developing a phenological model specifically tailored to grass pollen in the autumn season. Using temperature and sunshine hours (SSH) as inputs, our model can accurately predict the start and end dates of pollen seasons not only for the current year but also for future scenarios when meteorological forecasts are available. This approach moves beyond data analysis of existing records to establish a predictive framework for grass pollen phenology.
Shokouhi et al. (2024) and Huete et al. (2019) employed machine learning algorithms to directly simulate pollen concentrations, yielding notable results. In our study, however, the random forest (RF) model was applied solely to simulate annual pollen production due to its effectiveness in handling the numerous factors influencing this variable. Unlike these studies, RF represents only one component of our broader modeling framework, rather than the primary focus of our methodology.
In conclusion, we believe our study presents clear novelty and fundamental differences compared to previous research. By combining phenological model tailored for grass pollen with predictive capabilities and integrating it into a broader modeling framework, our work goes beyond the methods used in earlier studies. This approach not only advances our understanding of pollen dynamics but also provides tools for future forecasting under changing climate conditions, demonstrating its distinctiveness and innovation.
（2）On the spatial resolution of plant functional type (PFT) data
We greatly appreciate your insightful comments regarding the importance of PFT data and its resolution requirements. We fully agree that accurate characterization of pollen emission sources requires careful consideration of PFT spatial heterogeneity, particularly the partitioning of grasses into C3 and C4 components. Although the 5-km resolution PFT dataset provided by NCAR CLM may not be highly detailed, we believe it is sufficient for the scope of this study, given that Beijing, as a major metropolitan area in China, spans approximately 16,411 km². Moreover, for the purpose of this study, which focuses on developing and testing a pollen emission potential model, we believe that this resolution is sufficient to address the core research objectives.
Additionally, we understand that concerns about insufficient spatial heterogeneity in PFT data, combined with issues regarding figure quality, may stem from the spatial resolution of our final pollen emission model output. In our study, the model was applied at a 0.1° (~10 km) resolution across the Beijing area. We recognize that this resolution might appear coarse, but this decision was made to balance the trade-off between computational efficiency and model complexity. Simulating pollen emissions at higher resolutions would significantly increase computational demands, especially when using interpolated meteorological data and random forest algorithms for regional modeling.
That said, we agree with your suggestion to consider finer-scale PFT data and higher-resolution pollen emission modeling in future studies. Such efforts will allow us to better capture the spatial variability of pollen sources and improve the accuracy of our predictions.
（3）On the quality of figures
We sincerely apologize for the shortcomings in the quality and clarity of the figures in our original submission. We understand that figures are a crucial element for effectively communicating results. In response to your feedback, we have thoroughly reviewed and revised all figure captions to enhance their clarity and informativeness. Specifically, for the figure you mentioned (Figure 11, now revised as Figure 7), we have redesigned the layout and provided a detailed caption explaining the rationale for using two y-axes with different scales. These changes aim to eliminate any potential confusion for readers. We have also conducted a comprehensive review of all other figures to ensure they meet high standards of clarity and presentation.
Once again, we deeply appreciate your constructive feedback, which has been instrumental in helping us improve our work. We hope that the revised manuscript will better address your concerns and meet the expectations of ACP. If you have any additional comments or suggestions, we would be most grateful to receive them.
Thank you for your understanding and consideration.
Citation: https://doi.org/10.5194/egusphere-2024-2000-AC2
-
AC2: 'Reply on RC2', Jiangtao Li, 02 Dec 2024
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-2000', Anonymous Referee #1, 02 Nov 2024
This work by Li et al., has successfully constructed a pollen emission model based on phenology and random forest algorithm based on the 16 years of pollen observation data observed in Beijing. Moreover, the impact of temperature and relative humidity on the pollen emission has been reported. Furtherly, this pollen emission model was integrated into the regional climate model to validate the reliability of this model. The work presents a useful pollen prediction model for future prediction.
General points:
- why the data was stopped at 2021? Could this model be used to predict those for year 2022, 2023?
- there are too many figures and tables in this manuscript. I would suggest the authors to move some of them into the supporting information.
- previous literatures on pollen prediction model should be introduced in the introduction. How about the advancement of the current model when compared to those in the literatures?
Minor points:
Line 31: RegCM. The first time it appeared, the full name should be appeared first. Similar for sDOY, eDOY in lines 134-135.
Lines 58-62: references should be provided.
Line 95, Line 103: what is the difference between numerical model and pollen emission model? It would be better to clearly explain it here.
Line 123: how was this formula obtained? Was this a novel formula or from literatures?
Line 147: what is the meaning of Rs1, Rs2, Rssig?
Line 182: Has the RF algorithm been used for pollen emission simulation in the previous literatures?
Section 3.1.1: was temperature the only variable in this autumn phenology model?
Lines 548-551: please provide more details or discussions regarding the linkage between pollen amount and climate change.
Lines 617-625: it seems that basic information regarding the RegCM should be in the introduction section. It might be better for the authors to concise the current introduction and move such information into the introduction also.
Section 3.3.1: some of them should be in the introduction section, and some may be better be in the method section.
Citation: https://doi.org/10.5194/egusphere-2024-2000-RC1 -
AC1: 'Reply on RC1', Jiangtao Li, 13 Nov 2024
Thank you for your detailed and insightful comments on our manuscript. We have addressed all of your concerns and questions in the attached PDF. In the document, your review comments are displayed in black text, and our responses and corresponding revisions are highlighted in red. Once again, we sincerely appreciate your valuable feedback.
-
RC2: 'Comment on egusphere-2024-2000', Anonymous Referee #3, 20 Nov 2024
Predicting allergenic pollen concentrations is important for urban public health management in the context of climate change. This study attempted to construct a pollen emission model using phenology and random forests, which was further integrated into RegCM to predict pollen concentrations driven by meteorological forecasts. Although the topic is important, several concerns prevent me from recommending this paper for publication in ACP. First, the novelty of the method is insufficient. The authors claim that the use of phenology and RF is a major novelty of this study, but previous studies such as Lo et al. (2018) in North America, Shokouhi et al. (2024) in Europe, and Huete et al. (2019) in southeastern Australia have all considered phenology and used RF models for pollen forecasting. Therefore, the novelty of this study is questionable. Second, plant functional type is an important information to determine the source of pollen emission, especially the partitioning of grasses into cool season C3 and warm season C4 components. For an urban study such as this one, the spatial resolution of the PFT data should be sufficient to resolve the spatial heteorogeneity of the PFT (C3/C4 partitioning). Instead of searching for a fine-scale local PFT database and validating it against in-situ measurements, the authors used a coarse-resolution (~5 km) database from NCAR's CLM, originally developed for global simulations. Uncertainties associated with the coarse-resolution PFT data are not rigorously considered or critically discussed in the error propagation. Finally, the quality of the figures presented is generally poor, often lacking critical information for the reader to appreciate the figure (e.g. Fig. 11, the reason for using two identical y-axes but with different scales is not mentioned in the figure caption, very difficult to understand). As such, I think the quality and novelty of this study does not meet the standard of ACP.
Citation: https://doi.org/10.5194/egusphere-2024-2000-RC2 -
AC2: 'Reply on RC2', Jiangtao Li, 02 Dec 2024
We sincerely appreciate the time and effort you have dedicated to reviewing our manuscript and providing valuable feedback. We are deeply sorry that our initial submission did not meet your expectations in certain areas. Your insightful comments are of great importance to us, and we have carefully considered each point raised. Below, we provide a detailed response to your concerns:
（1）On the novelty of our methodology
We value your comments regarding the novelty of our study and have reviewed the references you mentioned (Lo et al., 2018; Shokouhi et al., 2024; Huete et al., 2019). While acknowledging the contributions of these studies, we believe our work has distinct differences:
Lo et al. (2018) developed pollen calendars for North America using data from 31 National Allergy Bureau stations, focusing on analyzing the start, end, and duration of pollen seasons and annual pollen integrals for various taxa. In contrast, our study centers on developing a phenological model specifically tailored to grass pollen in the autumn season. Using temperature and sunshine hours (SSH) as inputs, our model can accurately predict the start and end dates of pollen seasons not only for the current year but also for future scenarios when meteorological forecasts are available. This approach moves beyond data analysis of existing records to establish a predictive framework for grass pollen phenology.
Shokouhi et al. (2024) and Huete et al. (2019) employed machine learning algorithms to directly simulate pollen concentrations, yielding notable results. In our study, however, the random forest (RF) model was applied solely to simulate annual pollen production due to its effectiveness in handling the numerous factors influencing this variable. Unlike these studies, RF represents only one component of our broader modeling framework, rather than the primary focus of our methodology.
In conclusion, we believe our study presents clear novelty and fundamental differences compared to previous research. By combining phenological model tailored for grass pollen with predictive capabilities and integrating it into a broader modeling framework, our work goes beyond the methods used in earlier studies. This approach not only advances our understanding of pollen dynamics but also provides tools for future forecasting under changing climate conditions, demonstrating its distinctiveness and innovation.
（2）On the spatial resolution of plant functional type (PFT) data
We greatly appreciate your insightful comments regarding the importance of PFT data and its resolution requirements. We fully agree that accurate characterization of pollen emission sources requires careful consideration of PFT spatial heterogeneity, particularly the partitioning of grasses into C3 and C4 components. Although the 5-km resolution PFT dataset provided by NCAR CLM may not be highly detailed, we believe it is sufficient for the scope of this study, given that Beijing, as a major metropolitan area in China, spans approximately 16,411 km². Moreover, for the purpose of this study, which focuses on developing and testing a pollen emission potential model, we believe that this resolution is sufficient to address the core research objectives.
Additionally, we understand that concerns about insufficient spatial heterogeneity in PFT data, combined with issues regarding figure quality, may stem from the spatial resolution of our final pollen emission model output. In our study, the model was applied at a 0.1° (~10 km) resolution across the Beijing area. We recognize that this resolution might appear coarse, but this decision was made to balance the trade-off between computational efficiency and model complexity. Simulating pollen emissions at higher resolutions would significantly increase computational demands, especially when using interpolated meteorological data and random forest algorithms for regional modeling.
That said, we agree with your suggestion to consider finer-scale PFT data and higher-resolution pollen emission modeling in future studies. Such efforts will allow us to better capture the spatial variability of pollen sources and improve the accuracy of our predictions.
（3）On the quality of figures
We sincerely apologize for the shortcomings in the quality and clarity of the figures in our original submission. We understand that figures are a crucial element for effectively communicating results. In response to your feedback, we have thoroughly reviewed and revised all figure captions to enhance their clarity and informativeness. Specifically, for the figure you mentioned (Figure 11, now revised as Figure 7), we have redesigned the layout and provided a detailed caption explaining the rationale for using two y-axes with different scales. These changes aim to eliminate any potential confusion for readers. We have also conducted a comprehensive review of all other figures to ensure they meet high standards of clarity and presentation.
Once again, we deeply appreciate your constructive feedback, which has been instrumental in helping us improve our work. We hope that the revised manuscript will better address your concerns and meet the expectations of ACP. If you have any additional comments or suggestions, we would be most grateful to receive them.
Thank you for your understanding and consideration.
Citation: https://doi.org/10.5194/egusphere-2024-2000-AC2
-
AC2: 'Reply on RC2', Jiangtao Li, 02 Dec 2024
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 228 | 52 | 255 | 535 | 23 | 16 | 18 |
- HTML: 228
- PDF: 52
- XML: 255
- Total: 535
- Supplement: 23
- BibTeX: 16
- EndNote: 18
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Jiangtao Li
Xingqin An
Zhaobin Sun
Caihua Ye
Qing Hou
Yuxin Zhao
Zhe Liu
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(2750 KB) - Metadata XML
-
Supplement
(580 KB) - BibTeX
- EndNote
- Final revised paper