the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Modeling global wildfire activity in the intermediate complexity University of Victoria Earth System Climate Model (UVic ESCM 2.9): the importance of the simulated climatology
Abstract. Fire is an integral part of the Earth system, interacting in complex ways with humans, vegetation and climate. Global fire activity is an important element of the carbon cycle, and understanding its role in the context of climate change is crucial. In order to represent the transient fire-climate-vegetation interactions and to integrate these in the long term climate projections of climate models, coupling these three components is necessary. Global fire models have been coupled to climate-vegetation models with complex atmosphere modules but these models are computationally intensive. In this research, we use the University of Victoria Earth System Climate Model (UVic ESCM), an ESCM of intermediate complexity to which we couple a process based global fire model, in order to develop a computationally efficient means of studying long term fire-climate-vegetation interactions. The fire model used simulates burned area based primarily on relative humidity, soil moisture and biomass density. The UViC ESCM’s simulated relative humidity is improved by parameterizing it according to the simulated precipitation, and observational variability is added to the simulated climatology to improve the variability of simulated burned area. The best parameterization achieves a moderate spatial agreement of simulated burned area with observational data. Tropical rainforests in South America and Africa, however, display very high burned fractions, due to the poorly simulated relative humidity input; indeed, when we used observed relative humidity to simulate fire activity, the pattern of burned area in the tropics improved substantially. This research demonstrates the importance of variability and regional patterns of climatology for global wildfire activity and the corresponding limitations of ESCMs that simplify atmospheric circulation. This suggests that using pattern scaling of climate variables as an input to fire models could provide such ESCMs of intermediate complexity with the ability to integrate global fire activity.
- Preprint
(14356 KB) - Metadata XML
- BibTeX
- EndNote
Interactive discussion
Status: closed
-
AC1: 'Comment on egusphere-2022-961', Étienne Guertin, 01 Nov 2022
Hello editors,
Please note that in the preprint the DOI link to the code and data is incorrect; here is the correct link: https://doi.org/10.20383/103.0647
Étienne G.
Citation: https://doi.org/10.5194/egusphere-2022-961-AC1 -
RC1: 'Comment on egusphere-2022-961', Anonymous Referee #1, 16 Dec 2022
Review of “Modeling global wildfire activity in the intermediate complexity University of Victory Earth System Climate Model (UVic ESCM 2.9): the importance of the simulated climatology” by Étienne Guertin and H. Damon Matthews
“Modeling global wildfire activity in the intermediate complexity University of Victory Earth System Climate Model (UVic ESCM 2.9): the importance of the simulated climatology” by Étienne Guertin and H. Damon Matthews presents a newly coupled model between the Earth system model of intermediate complexity UVic and a process-based fire model. By this, the authors provide a computationally efficient model to study long-term fire-vegetation-climate interactions and feedback mechanisms. The paper demonstrates that the performance and resolution of the intermediate complexity Earth system model are limiting factors in capturing the variability and regional pattern of fire behavior. By prescribing humidity, the fire model performance improved significantly.
The introduction of a computationally efficient and fire-enabled Earth system model of intermediate complexity is useful for the community to study long-term fire-vegetation-climate interactions and the paper is well written. I am, however, not sure if the setup and performance of the model are sufficient to achieve these scientific aims. Also, the presentation of the paper, including the method section and discussion could be improved upon.
Please find in the following my general and minor suggestions and comments.
General comments:
- A large part of the method section displays the description of the fire model, which seems to have been taken almost completely from Li et al. 2012. I think it would be sufficient to show just the most important parameterizations but not repeat an already published model description. What I miss is a more conceptual explanation for these parameterizations. Almost all parameterizations are just stated, but what is the logic behind them, and where do they come from (for example empirical relationships)?
- It would be good to clearly state all differences between the model of this paper and the implementation in Li et al. Is the fire model in Li et al. for example specifically tuned or adjusted to CLM (for example to the number of PFTs)? Was the fire model already included in TRIFFID before or was it the work of this paper? What changes did you have to implement in the fire model to adjust it for the new vegetation model, for example in the case of different PFTs? Did you do offline experiments (without UVic, to judge the fire model performance with bias-corrected climate input?
- I further miss in the method part a more detailed description of what the authors did themselves. For example, the parameter optimization seems to be quite important for the performance of the model but has been described with few details. Furthermore, how was the coupling technically done, and what steps had to be undertaken?
- In the results (or possibly supplement), I miss a short evaluation of the UVic model to judge if the performance is good enough to capture fire behavior. How are roughly the temperature and precipitation biases, for example? Furthermore, the model could also be compared to other fire models and not just satellite data, which would be a more realistic comparison to judge if the model performance is sufficient. This could be done, for example, to fire data from CMIP Earth system models or uncoupled vegetation-fire models from FireMIP.
- My main concern is that the model might not be good enough to really study fire behavior in the Earth system. The performance should be discussed well in the Discussion section. The main aim of Earth system models is to project future climate changes. Can the model still look at the future changes in fire regimes when e.g. humidity or fire duration is prescribed to historic data? Without prescribed humidity, the model performance looks not good enough to be useful as e.g. in Africa most of the tropical forest is burnt down. For an updated paper version, the authors should include these and other points in the Discussion in greater detail.
Minor comments:
- L7-8: What is the name of the fire model?
- L11-13: The sentence is too long and confusing, please rephrase.
- L28-30: Revise sentence structure.
- L34/35: Please shortly mention how these opposing effects of warming and cooling are originating from fire activity.
- L78-82: What are the main differences between the simplified MOSES to the standard one? Why did the authors not use the standard MOSES?
- Figure 2:
- Consider more clearly separating the chart by processes, e.g. all fire processes in one column, etc. to improve overall clarity.
- Why are GHG emission scenarios the scientific goal? Here it looks more like a forcing to the model. Also, no arrows are in the direction of the GHG emission scenarios, therefore I would revise the label as a scientific goal.
- Why do fire impacts have a citation (Landry et al. 2015)? Fire impacts seem relatively generic and a result of the modeling. To my understanding, it would make more sense to provide a citation for the model inputs.
- L129/131: Please revise the unit names. At least a space between the words is necessary.
- 133/134: Why does fire occurrence depend on snow? Of course, temperature and moisture are very important for fire occurrence, therefore the probability of a spreading fire with snow is low but should not depend on it. In the boreal forests, fires are possible with snow (e.g. so-called zombie-fires).
- Table 1: Why do the parameter differ between the different RH model runs? How were they elected? Especially in the parameters from own work, what was the methodology to come up with these values and why do they differ from “Other literature”? The table caption should be more detailed. The different RH model runs should also be better introduced in the earlier method chapter.
- L213: Where does this 5% come from exactly? Is this number based on fire observations?
- 222-224: This is not clear: humidity is spatially homogeneous and precipitation is generated by a humidity threshold of 0.85? But then you use this precipitation (which should then also be spatially homogeneous to derive a new humidity field? This new humidity field is then also dependent on the original one. Please explain, how and why you get better humidity with this approach. Ideally, the new humidity should be independent of the old one.
- Figure 3 caption: So the humidity is generally almost constant? Spatially and temporal? Then I would not really name this “simulated”. And I would also doubt if such a model can be at all useful for fire and vegetation modeling, especially since precipitation seems to depend on this humidity.
- L253-259: This part should come earlier to explain where the different parameter values in Table 1 come from. But it is not clear, how this optimization has been performed. Was e.g. for different parameter combinations an error metric of the different model outputs calculated and the combination with the lowest error chosen? What software has been used for this? Do the new values still make sense from the perspective of fire modeling or are they just tuned to get the best results?
- L261: Please specify here which other climate forcings were used.
- L272/273: Figures should be numbered according to their first appearance in the text.
- L282-283: It is quite strange that the vegetation completely burns down with a fire fraction of around 0.1. Could you explain this mismatch?
- Fig 6 and 7: For clarity, it would be useful to use the model names rhsim and rhobs also in the figure or figure caption.
- Figure 8: Maybe you could also provide a table with the results for the different experiments. These experiments were not described much. Maybe you could elaborate a bit more, on why you used these parameter values and setup and maybe even provide maps for each experiment in the appendix.
- Line 342-344: You did not really answer the question about the feasibility of a fully coupled simulation of burned area in these lines. Demonstrating to substitute temporal variability partially does not directly validate the initial hypothesis. And is the model with these substitutions (e.g. humidity) still fully coupled?
- L354-366: You write earlier that fire is coupled in some ESMs (while not in EMICs). Maybe it would be better to compare your results to these ESMs than to uncoupled fire models, which take prescribed climatic input.
- L391: You write that it was the main objective to include fire-caused emissions in the model. But you did not show a figure or evaluation of fire emissions in the results, which would be important for climate feedback.
Citation: https://doi.org/10.5194/egusphere-2022-961-RC1 -
RC2: 'Comment on egusphere-2022-961', Anonymous Referee #2, 20 Jan 2023
The Li et al. (2012) fire model has already been commonly used in Earth system models (ESMs) and land surface models. This study is the first work that incorporates it to an ESM of intermediate complexity. It provides a valuable model platform to quantify the role of fire in the Earth system and analyze the related mechanisms in a computationally efficient way. This study also highlights the importance of simulated RH climatology in fire simulations, which could be a guideline for improving fire modeling in ESMs. This paper is interesting and well-structured. It would be a useful publication in GMD after some places are further clarified and expanded.
Specific comments
- In introduction and abstract, you could point out that earlier version of UVic ESCM uses prescribed burned area (Landry et al. 2017, ERL). Your work realizes the interactive fire modeling in UVic ESCM.
- There are three suggestions for future development of UVic ESCM coupling with Li et al. (2012) fire model. I hope that you at least discuss it in Sec. 4. First, you could tune the PFT-dependent potential maximum fire spread rate, which will improve the simulations of (at least) global total. Second, it’s fine to use precipitation as proxy to re-parameterize RH if RH is poorly simulated. Another alternative way is to use other hydrology variables and build new function of fuel flammability. Third, combustion completeness factors could be optimized if fire emissions are much overestimated or underestimated.
- P13, Sec. 2.5, what is principle of your parameter optimization? better global total, better spatial pattern, or both? Please clarification.
- Sec. 4.1, when compared with Wu et al. (2021), you could point out the IMOGEN-LPJ-SEVER used in Wu et al. (2021) did not consider aerosol influence, while UVic ESCM considers the aerosol direct, semi-direct, and indirect effects (Landry et al. 2017, ERL). It’s important to understand the impact of fire aerosols on global climate. See Li et al. (2022) for details.
Corrections
- P4, L93, change “vegetation structure” to “vegetation distribution”.
- P6, L131, unit of population density, you could consider inhabitants km-2, people km-2, or persons km-2. The “nopeople km-2” is rarely used.
- 10, L212, what does the “islands of vegetation” mean?
- P11, L229, remove “field”.
- P11, L232, change “observation-based” to “observed”.
- P19, Table 3, change “global burned area” to “GBA” in the table or remove “GBA: global burned area” in the caption.
- P19, L370, change “daily timestep” to “sub-daily timestep” or “hourly timestep”. Fire occurrence and spread used hourly timestep in Li et al. (2012).
- Figs. 1 and 3, change “observation-based” to “observations”.
Reference:
Li F., D. M. Lawrence, Y.-Q. Jiang, X. Liu, Z.-D. Lin, 2022: Fire aerosols slow down the global water cycle, Journal of Climate, 35, 3619-3633.
Citation: https://doi.org/10.5194/egusphere-2022-961-RC2
Interactive discussion
Status: closed
-
AC1: 'Comment on egusphere-2022-961', Étienne Guertin, 01 Nov 2022
Hello editors,
Please note that in the preprint the DOI link to the code and data is incorrect; here is the correct link: https://doi.org/10.20383/103.0647
Étienne G.
Citation: https://doi.org/10.5194/egusphere-2022-961-AC1 -
RC1: 'Comment on egusphere-2022-961', Anonymous Referee #1, 16 Dec 2022
Review of “Modeling global wildfire activity in the intermediate complexity University of Victory Earth System Climate Model (UVic ESCM 2.9): the importance of the simulated climatology” by Étienne Guertin and H. Damon Matthews
“Modeling global wildfire activity in the intermediate complexity University of Victory Earth System Climate Model (UVic ESCM 2.9): the importance of the simulated climatology” by Étienne Guertin and H. Damon Matthews presents a newly coupled model between the Earth system model of intermediate complexity UVic and a process-based fire model. By this, the authors provide a computationally efficient model to study long-term fire-vegetation-climate interactions and feedback mechanisms. The paper demonstrates that the performance and resolution of the intermediate complexity Earth system model are limiting factors in capturing the variability and regional pattern of fire behavior. By prescribing humidity, the fire model performance improved significantly.
The introduction of a computationally efficient and fire-enabled Earth system model of intermediate complexity is useful for the community to study long-term fire-vegetation-climate interactions and the paper is well written. I am, however, not sure if the setup and performance of the model are sufficient to achieve these scientific aims. Also, the presentation of the paper, including the method section and discussion could be improved upon.
Please find in the following my general and minor suggestions and comments.
General comments:
- A large part of the method section displays the description of the fire model, which seems to have been taken almost completely from Li et al. 2012. I think it would be sufficient to show just the most important parameterizations but not repeat an already published model description. What I miss is a more conceptual explanation for these parameterizations. Almost all parameterizations are just stated, but what is the logic behind them, and where do they come from (for example empirical relationships)?
- It would be good to clearly state all differences between the model of this paper and the implementation in Li et al. Is the fire model in Li et al. for example specifically tuned or adjusted to CLM (for example to the number of PFTs)? Was the fire model already included in TRIFFID before or was it the work of this paper? What changes did you have to implement in the fire model to adjust it for the new vegetation model, for example in the case of different PFTs? Did you do offline experiments (without UVic, to judge the fire model performance with bias-corrected climate input?
- I further miss in the method part a more detailed description of what the authors did themselves. For example, the parameter optimization seems to be quite important for the performance of the model but has been described with few details. Furthermore, how was the coupling technically done, and what steps had to be undertaken?
- In the results (or possibly supplement), I miss a short evaluation of the UVic model to judge if the performance is good enough to capture fire behavior. How are roughly the temperature and precipitation biases, for example? Furthermore, the model could also be compared to other fire models and not just satellite data, which would be a more realistic comparison to judge if the model performance is sufficient. This could be done, for example, to fire data from CMIP Earth system models or uncoupled vegetation-fire models from FireMIP.
- My main concern is that the model might not be good enough to really study fire behavior in the Earth system. The performance should be discussed well in the Discussion section. The main aim of Earth system models is to project future climate changes. Can the model still look at the future changes in fire regimes when e.g. humidity or fire duration is prescribed to historic data? Without prescribed humidity, the model performance looks not good enough to be useful as e.g. in Africa most of the tropical forest is burnt down. For an updated paper version, the authors should include these and other points in the Discussion in greater detail.
Minor comments:
- L7-8: What is the name of the fire model?
- L11-13: The sentence is too long and confusing, please rephrase.
- L28-30: Revise sentence structure.
- L34/35: Please shortly mention how these opposing effects of warming and cooling are originating from fire activity.
- L78-82: What are the main differences between the simplified MOSES to the standard one? Why did the authors not use the standard MOSES?
- Figure 2:
- Consider more clearly separating the chart by processes, e.g. all fire processes in one column, etc. to improve overall clarity.
- Why are GHG emission scenarios the scientific goal? Here it looks more like a forcing to the model. Also, no arrows are in the direction of the GHG emission scenarios, therefore I would revise the label as a scientific goal.
- Why do fire impacts have a citation (Landry et al. 2015)? Fire impacts seem relatively generic and a result of the modeling. To my understanding, it would make more sense to provide a citation for the model inputs.
- L129/131: Please revise the unit names. At least a space between the words is necessary.
- 133/134: Why does fire occurrence depend on snow? Of course, temperature and moisture are very important for fire occurrence, therefore the probability of a spreading fire with snow is low but should not depend on it. In the boreal forests, fires are possible with snow (e.g. so-called zombie-fires).
- Table 1: Why do the parameter differ between the different RH model runs? How were they elected? Especially in the parameters from own work, what was the methodology to come up with these values and why do they differ from “Other literature”? The table caption should be more detailed. The different RH model runs should also be better introduced in the earlier method chapter.
- L213: Where does this 5% come from exactly? Is this number based on fire observations?
- 222-224: This is not clear: humidity is spatially homogeneous and precipitation is generated by a humidity threshold of 0.85? But then you use this precipitation (which should then also be spatially homogeneous to derive a new humidity field? This new humidity field is then also dependent on the original one. Please explain, how and why you get better humidity with this approach. Ideally, the new humidity should be independent of the old one.
- Figure 3 caption: So the humidity is generally almost constant? Spatially and temporal? Then I would not really name this “simulated”. And I would also doubt if such a model can be at all useful for fire and vegetation modeling, especially since precipitation seems to depend on this humidity.
- L253-259: This part should come earlier to explain where the different parameter values in Table 1 come from. But it is not clear, how this optimization has been performed. Was e.g. for different parameter combinations an error metric of the different model outputs calculated and the combination with the lowest error chosen? What software has been used for this? Do the new values still make sense from the perspective of fire modeling or are they just tuned to get the best results?
- L261: Please specify here which other climate forcings were used.
- L272/273: Figures should be numbered according to their first appearance in the text.
- L282-283: It is quite strange that the vegetation completely burns down with a fire fraction of around 0.1. Could you explain this mismatch?
- Fig 6 and 7: For clarity, it would be useful to use the model names rhsim and rhobs also in the figure or figure caption.
- Figure 8: Maybe you could also provide a table with the results for the different experiments. These experiments were not described much. Maybe you could elaborate a bit more, on why you used these parameter values and setup and maybe even provide maps for each experiment in the appendix.
- Line 342-344: You did not really answer the question about the feasibility of a fully coupled simulation of burned area in these lines. Demonstrating to substitute temporal variability partially does not directly validate the initial hypothesis. And is the model with these substitutions (e.g. humidity) still fully coupled?
- L354-366: You write earlier that fire is coupled in some ESMs (while not in EMICs). Maybe it would be better to compare your results to these ESMs than to uncoupled fire models, which take prescribed climatic input.
- L391: You write that it was the main objective to include fire-caused emissions in the model. But you did not show a figure or evaluation of fire emissions in the results, which would be important for climate feedback.
Citation: https://doi.org/10.5194/egusphere-2022-961-RC1 -
RC2: 'Comment on egusphere-2022-961', Anonymous Referee #2, 20 Jan 2023
The Li et al. (2012) fire model has already been commonly used in Earth system models (ESMs) and land surface models. This study is the first work that incorporates it to an ESM of intermediate complexity. It provides a valuable model platform to quantify the role of fire in the Earth system and analyze the related mechanisms in a computationally efficient way. This study also highlights the importance of simulated RH climatology in fire simulations, which could be a guideline for improving fire modeling in ESMs. This paper is interesting and well-structured. It would be a useful publication in GMD after some places are further clarified and expanded.
Specific comments
- In introduction and abstract, you could point out that earlier version of UVic ESCM uses prescribed burned area (Landry et al. 2017, ERL). Your work realizes the interactive fire modeling in UVic ESCM.
- There are three suggestions for future development of UVic ESCM coupling with Li et al. (2012) fire model. I hope that you at least discuss it in Sec. 4. First, you could tune the PFT-dependent potential maximum fire spread rate, which will improve the simulations of (at least) global total. Second, it’s fine to use precipitation as proxy to re-parameterize RH if RH is poorly simulated. Another alternative way is to use other hydrology variables and build new function of fuel flammability. Third, combustion completeness factors could be optimized if fire emissions are much overestimated or underestimated.
- P13, Sec. 2.5, what is principle of your parameter optimization? better global total, better spatial pattern, or both? Please clarification.
- Sec. 4.1, when compared with Wu et al. (2021), you could point out the IMOGEN-LPJ-SEVER used in Wu et al. (2021) did not consider aerosol influence, while UVic ESCM considers the aerosol direct, semi-direct, and indirect effects (Landry et al. 2017, ERL). It’s important to understand the impact of fire aerosols on global climate. See Li et al. (2022) for details.
Corrections
- P4, L93, change “vegetation structure” to “vegetation distribution”.
- P6, L131, unit of population density, you could consider inhabitants km-2, people km-2, or persons km-2. The “nopeople km-2” is rarely used.
- 10, L212, what does the “islands of vegetation” mean?
- P11, L229, remove “field”.
- P11, L232, change “observation-based” to “observed”.
- P19, Table 3, change “global burned area” to “GBA” in the table or remove “GBA: global burned area” in the caption.
- P19, L370, change “daily timestep” to “sub-daily timestep” or “hourly timestep”. Fire occurrence and spread used hourly timestep in Li et al. (2012).
- Figs. 1 and 3, change “observation-based” to “observations”.
Reference:
Li F., D. M. Lawrence, Y.-Q. Jiang, X. Liu, Z.-D. Lin, 2022: Fire aerosols slow down the global water cycle, Journal of Climate, 35, 3619-3633.
Citation: https://doi.org/10.5194/egusphere-2022-961-RC2
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
511 | 177 | 38 | 726 | 33 | 26 |
- HTML: 511
- PDF: 177
- XML: 38
- Total: 726
- BibTeX: 33
- EndNote: 26
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1