the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
A global fire emission dataset using the three-corner hat method (FiTCH)
Abstract. Fire carbon emissions contribute to the accumulation of atmospheric CO2 and affect climate change. It is crucial to accurately monitor the dynamics of global fire emissions for fire management and climate change mitigation. However, there are large uncertainties in the existing satellite-based global fire emission products. This study analyzed the uncertainties of six state-of-the-art fire emission products and merged them using the three-corner hat method (TCH), producing a new global fire emission dataset, FiTCH. Our results revealed that satellite-based products such as the Global Fire Assimilation System (GFAS), the Quick Fire Emissions Dataset (QFED), and the Global Fire Emissions Database (GFED) had low uncertainties in fire emissions, while the Fire INventory from National Center for Atmospheric Research (NCAR) (FINN), the Fire Energetics and Emissions Research (FEER), and Xu et al. (2021) data had high uncertainties. The proposed FiTCH dataset presented the lowest uncertainties with a mean annual fire emission of 1978.47 Tg C in 2001–2021. Among biomes, tropical forests and tundra showed higher uncertainties than other biomes such as temperate forests and Mediterranean forests. In drought years, forests showed increased fire emissions, especially in boreal forests, while non-forest regions like grasslands displayed decreased emissions. By integrating the FiTCH data and historical fire emissions in the late 20th century, 1994 was identified as a break year, before which global fire emissions increased significantly and after which the emissions decreased. Global land temperatures and fire emissions have decoupled in the past two decades. However, climate change still causes threats to forest carbon sequestration, especially for boreal forests. This study highlights the importance of forest fire monitoring and management for effective climate mitigation and ecosystem conservation. The proposed FiTCH dataset is available from: https://doi.org/10.6084/m9.figshare.22647382.v1 (Liu and Yang, 2023).
- Preprint
(3258 KB) - Metadata XML
-
Supplement
(1665 KB) - BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on essd-2023-150', Anonymous Referee #1, 09 Jul 2023
This study compared six widely-used fire emission products and merged them using the three-cornered hat method (TCH). A new global fire emission dataset, FiTCH, was developed to quantify the fire emissions between 2001 and 2021. Fire emissions in different regions and biomes were derived and analyzed. The impact of drought on fire emissions was also evaluated. This study is timely and valuable as global climate change is expected to cause more frequent extreme events, such as extreme droughts and fires. Figuring out the uncertainties of the existing fire emission products and producing accurate fire emission data are important for global climate change analysis. Overall, this manuscript is solid and well-structured. However, the descriptions of some important points are inadequate. I suggest an accept after addressing the concerned and comments below.
Comments:
1. In Section 2.3, the impact of drought on fire emissions was quantified, which helped to illustrate the influence of drought on fire. Can the authors also analyze the effects of different drought severity? For example, -3 < PDSI < -2, -4< PDSI < -3, and PDSI < -4 usually indicate moderate drought, severe drought, and extreme drought, respectively. The fire emissions might change under different drought severity. Maybe add this extra analysis to the supplementary material.2. In Section 3.5, the correlations between fire emissions and temperature were described. However, only the RETRO data and the proposed FiTCH data were used. Can the authors also analyze the relationships between the six fire emission products and the temperature? For example, use a scatterplot for each product like Figure S3, with the x axis and the y axis for temperature and fire emissions, respectively. This analysis may also go to the supplement. Otherwise, Figure 8 will be too large.
3. The manuscript used the term “the three-corner hat”, however, it is more common to use “the three-cornered hat”. Maybe revise the term to make it consistent with the current literature.
Citation: https://doi.org/10.5194/essd-2023-150-RC1 - AC1: 'Reply on RC1', Meng Liu, 17 Jul 2023
-
RC2: 'Comment on essd-2023-150', Anonymous Referee #2, 13 Jul 2023
Liu & Lang have assessed six different biomass burning emissions datasets using the three-cornered hat method. They constructed a new dataset and assessed relations between fires and climatological parameters. The paper reads well but I have fundamental doubts about the fidelity of the approach.
As mentioned by the authors, the TCH method was developed to evaluate the uncertainties of different products when the true values were unknown. Each dataset is treated the same and considered independent. In reality that is not the case. GFAS is derived from GFED3 (which is rather similar as GFED4), and QFED builds on GFAS. Those datasets find their origin in the MODIS burned area data, which is also used by Xu et al. (2021). That means that four datasets share the same origin and FiTCH resembles those and yields low uncertainties to these four and relatively high to FEER and FINN which provide substantially higher emissions.
Now, in reality we do know to some degree what the true values are for biomass burning emissions. FEER uses top-down constraints and some recent work using Sentinel-2 burned area data indicates that MODIS severely underestimates burned area and the derived emission products thus severely underestimate emissions. See for example Ramo et al. (2021, PNAS, https://doi.org/10.1073/pnas.2011160118). Top-down studies also point towards higher emissions than calculated in the MODIS-based products, see for example Van der Velde et al. (2021) https://doi.org/10.1038/s41586-021-03712-y. Clearly these are regional studies but a wider look at the literature shows the same pattern; MODIS misses burned area.
The three-cornered hat method ignored this knowledge and considers the MODIS-based estimates as the best ones (I assume because they resemble eachother) while in reality it is much more likely that FEER and FINN are closer to the truth.
As a minor note, I would also be careful with using RETRO to go back in time. The uncertainties in the modelling and AVHRR approaches are very large and there has been more work done since that publication (BB4CMIP for example)
In summary, I am sorry to say that I feel the hard work of the authors does not help the field forward. If ‘true values’ indeed would be unknown I would see the benefit of this work although there is still the issue that the products are not independent, but in fact there are hints towards the right magnitude of fire emissions in the recent literature and they point to the opposite conclusion of this paper.
Citation: https://doi.org/10.5194/essd-2023-150-RC2 -
AC2: 'Reply on RC2', Meng Liu, 17 Jul 2023
Thanks for your valuable suggestions. We have improved our manuscript accordingly. Please see the attachment.
-
RC3: 'Reply on AC2', Anonymous Referee #2, 19 Jul 2023
Nobody knows the true values of biomass burning emissions. However, in my review I have referred to recent literature that points out that MODIS burned area misses a lot of burned area and thus has a strong bias. Possibly a factor two, much more than the difference between GFED4s and GFED3. All derived emission products, also when based on FRP or active fires but tuned to datasets derived from MODIS burned area, have the same issue. This is becoming well established and accepted by the community and it means that global burned area and thus emissions are substantially higher than at least four of the datasets indicate.
The authors introduce a new dataset that very likely contains the same bias as GFED, GFAS etc because it is based solely on a statistical technique ignoring the new insights I point towards in my review. From a methodological perspective the 3CH method might be interesting, but in my opinion it does not help the field forwards. In contrast, it presents a new dataset that suffers from the same issue as some of the older ones but claims its uncertainty is very low.
A final note, I disapprove of the comment “Additionally, can you provide the “true values” frequently mentioned in your comments? The detailed data sources including downloading links, citations, and user manuals are necessary. It is amazing that you have true global fire emission data for the past two decades. “. I fully understand the frustration of a negative review but it is up to the editor to decide whether it is a fair review or not. In my review I have pointed towards pieces of information that are as close to reliable estimates as we can come at this stage and I feel it is our duty to build on that.
Citation: https://doi.org/10.5194/essd-2023-150-RC3 - AC4: 'Reply on RC3', Meng Liu, 24 Jul 2023
-
RC3: 'Reply on AC2', Anonymous Referee #2, 19 Jul 2023
-
AC2: 'Reply on RC2', Meng Liu, 17 Jul 2023
-
AC3: 'Comment on essd-2023-150', Meng Liu, 17 Jul 2023
Hi all,
Thanks for all your help with improving the manuscript titled "A global fire emission dataset using the three-cornered hat method (FiTCH)". We have revised our manuscript based on the reviewers' comments. However, it seems that there is nowhere to submit the revised manuscript (and the supplement). So, we have to submit them as an attachment here. Sorry for the inconvenience.
Sincerely,
Meng Liu and Linqing Yang
-
RC4: 'Comment on essd-2023-150', Anonymous Referee #3, 24 Sep 2023
The manuscript entitled “A global fire emission dataset using the three-cornered hat method (FiTCH)” has been reviewed. The authors analyzed the uncertainties of size state-of-the art fire emission products and merged them using the three-cornered hat method (TCH), producing a new global fire emission dataset, FiTCH. The manuscript is well written and can be easily understood. However, this study does not satisfy the high standard of ESSD due to the following reasons.
- “The proposed FiTCH dataset presented the lowest uncertainties……” (Line 15). Actually, there are no true values for validation. The “uncertainty” in this manuscript looks like variance/covariance of the yearly time series data according my understanding on section 2.2. Datasets of FEER, FINN and Xu et al. show stronger fluctuations than the others (Figure 3a), generally corresponding to higher variances. Meanwhile, the so-called uncertainties of the three datasets are also higher than the other datasets (Section 3.1, lines 194-195). Therefore, the lowest “uncertainty” is more like the smallest fluctuation for FiTCH (Figure 3a).
- FiTCH is the weighted sum of other existing datasets (Lines 170-175). It may reduce some uncertainties by merging these existing datasets. However, it is still unclear about the specific benefits using the TCH method besides the so-called uncertainty. For example, does the drawbacks in one dataset have been solved based on another datasets?
- There are three primary goals in the manuscript (lines 76-78). However, according to my understanding on ESSD, generally the second goal is what ESSD care about. Unfortunately, I cannot find enough descriptions on the quality of the resulting dataset in both results and discussions.
- Figure 8. We can also use other existing datasets to do similar analysis and to get similar conclusions. What are the benefits using FiTCH compared to other datasets?
Citation: https://doi.org/10.5194/essd-2023-150-RC4
Status: closed
-
RC1: 'Comment on essd-2023-150', Anonymous Referee #1, 09 Jul 2023
This study compared six widely-used fire emission products and merged them using the three-cornered hat method (TCH). A new global fire emission dataset, FiTCH, was developed to quantify the fire emissions between 2001 and 2021. Fire emissions in different regions and biomes were derived and analyzed. The impact of drought on fire emissions was also evaluated. This study is timely and valuable as global climate change is expected to cause more frequent extreme events, such as extreme droughts and fires. Figuring out the uncertainties of the existing fire emission products and producing accurate fire emission data are important for global climate change analysis. Overall, this manuscript is solid and well-structured. However, the descriptions of some important points are inadequate. I suggest an accept after addressing the concerned and comments below.
Comments:
1. In Section 2.3, the impact of drought on fire emissions was quantified, which helped to illustrate the influence of drought on fire. Can the authors also analyze the effects of different drought severity? For example, -3 < PDSI < -2, -4< PDSI < -3, and PDSI < -4 usually indicate moderate drought, severe drought, and extreme drought, respectively. The fire emissions might change under different drought severity. Maybe add this extra analysis to the supplementary material.2. In Section 3.5, the correlations between fire emissions and temperature were described. However, only the RETRO data and the proposed FiTCH data were used. Can the authors also analyze the relationships between the six fire emission products and the temperature? For example, use a scatterplot for each product like Figure S3, with the x axis and the y axis for temperature and fire emissions, respectively. This analysis may also go to the supplement. Otherwise, Figure 8 will be too large.
3. The manuscript used the term “the three-corner hat”, however, it is more common to use “the three-cornered hat”. Maybe revise the term to make it consistent with the current literature.
Citation: https://doi.org/10.5194/essd-2023-150-RC1 - AC1: 'Reply on RC1', Meng Liu, 17 Jul 2023
-
RC2: 'Comment on essd-2023-150', Anonymous Referee #2, 13 Jul 2023
Liu & Lang have assessed six different biomass burning emissions datasets using the three-cornered hat method. They constructed a new dataset and assessed relations between fires and climatological parameters. The paper reads well but I have fundamental doubts about the fidelity of the approach.
As mentioned by the authors, the TCH method was developed to evaluate the uncertainties of different products when the true values were unknown. Each dataset is treated the same and considered independent. In reality that is not the case. GFAS is derived from GFED3 (which is rather similar as GFED4), and QFED builds on GFAS. Those datasets find their origin in the MODIS burned area data, which is also used by Xu et al. (2021). That means that four datasets share the same origin and FiTCH resembles those and yields low uncertainties to these four and relatively high to FEER and FINN which provide substantially higher emissions.
Now, in reality we do know to some degree what the true values are for biomass burning emissions. FEER uses top-down constraints and some recent work using Sentinel-2 burned area data indicates that MODIS severely underestimates burned area and the derived emission products thus severely underestimate emissions. See for example Ramo et al. (2021, PNAS, https://doi.org/10.1073/pnas.2011160118). Top-down studies also point towards higher emissions than calculated in the MODIS-based products, see for example Van der Velde et al. (2021) https://doi.org/10.1038/s41586-021-03712-y. Clearly these are regional studies but a wider look at the literature shows the same pattern; MODIS misses burned area.
The three-cornered hat method ignored this knowledge and considers the MODIS-based estimates as the best ones (I assume because they resemble eachother) while in reality it is much more likely that FEER and FINN are closer to the truth.
As a minor note, I would also be careful with using RETRO to go back in time. The uncertainties in the modelling and AVHRR approaches are very large and there has been more work done since that publication (BB4CMIP for example)
In summary, I am sorry to say that I feel the hard work of the authors does not help the field forward. If ‘true values’ indeed would be unknown I would see the benefit of this work although there is still the issue that the products are not independent, but in fact there are hints towards the right magnitude of fire emissions in the recent literature and they point to the opposite conclusion of this paper.
Citation: https://doi.org/10.5194/essd-2023-150-RC2 -
AC2: 'Reply on RC2', Meng Liu, 17 Jul 2023
Thanks for your valuable suggestions. We have improved our manuscript accordingly. Please see the attachment.
-
RC3: 'Reply on AC2', Anonymous Referee #2, 19 Jul 2023
Nobody knows the true values of biomass burning emissions. However, in my review I have referred to recent literature that points out that MODIS burned area misses a lot of burned area and thus has a strong bias. Possibly a factor two, much more than the difference between GFED4s and GFED3. All derived emission products, also when based on FRP or active fires but tuned to datasets derived from MODIS burned area, have the same issue. This is becoming well established and accepted by the community and it means that global burned area and thus emissions are substantially higher than at least four of the datasets indicate.
The authors introduce a new dataset that very likely contains the same bias as GFED, GFAS etc because it is based solely on a statistical technique ignoring the new insights I point towards in my review. From a methodological perspective the 3CH method might be interesting, but in my opinion it does not help the field forwards. In contrast, it presents a new dataset that suffers from the same issue as some of the older ones but claims its uncertainty is very low.
A final note, I disapprove of the comment “Additionally, can you provide the “true values” frequently mentioned in your comments? The detailed data sources including downloading links, citations, and user manuals are necessary. It is amazing that you have true global fire emission data for the past two decades. “. I fully understand the frustration of a negative review but it is up to the editor to decide whether it is a fair review or not. In my review I have pointed towards pieces of information that are as close to reliable estimates as we can come at this stage and I feel it is our duty to build on that.
Citation: https://doi.org/10.5194/essd-2023-150-RC3 - AC4: 'Reply on RC3', Meng Liu, 24 Jul 2023
-
RC3: 'Reply on AC2', Anonymous Referee #2, 19 Jul 2023
-
AC2: 'Reply on RC2', Meng Liu, 17 Jul 2023
-
AC3: 'Comment on essd-2023-150', Meng Liu, 17 Jul 2023
Hi all,
Thanks for all your help with improving the manuscript titled "A global fire emission dataset using the three-cornered hat method (FiTCH)". We have revised our manuscript based on the reviewers' comments. However, it seems that there is nowhere to submit the revised manuscript (and the supplement). So, we have to submit them as an attachment here. Sorry for the inconvenience.
Sincerely,
Meng Liu and Linqing Yang
-
RC4: 'Comment on essd-2023-150', Anonymous Referee #3, 24 Sep 2023
The manuscript entitled “A global fire emission dataset using the three-cornered hat method (FiTCH)” has been reviewed. The authors analyzed the uncertainties of size state-of-the art fire emission products and merged them using the three-cornered hat method (TCH), producing a new global fire emission dataset, FiTCH. The manuscript is well written and can be easily understood. However, this study does not satisfy the high standard of ESSD due to the following reasons.
- “The proposed FiTCH dataset presented the lowest uncertainties……” (Line 15). Actually, there are no true values for validation. The “uncertainty” in this manuscript looks like variance/covariance of the yearly time series data according my understanding on section 2.2. Datasets of FEER, FINN and Xu et al. show stronger fluctuations than the others (Figure 3a), generally corresponding to higher variances. Meanwhile, the so-called uncertainties of the three datasets are also higher than the other datasets (Section 3.1, lines 194-195). Therefore, the lowest “uncertainty” is more like the smallest fluctuation for FiTCH (Figure 3a).
- FiTCH is the weighted sum of other existing datasets (Lines 170-175). It may reduce some uncertainties by merging these existing datasets. However, it is still unclear about the specific benefits using the TCH method besides the so-called uncertainty. For example, does the drawbacks in one dataset have been solved based on another datasets?
- There are three primary goals in the manuscript (lines 76-78). However, according to my understanding on ESSD, generally the second goal is what ESSD care about. Unfortunately, I cannot find enough descriptions on the quality of the resulting dataset in both results and discussions.
- Figure 8. We can also use other existing datasets to do similar analysis and to get similar conclusions. What are the benefits using FiTCH compared to other datasets?
Citation: https://doi.org/10.5194/essd-2023-150-RC4
Data sets
A global fire emission dataset using the three-corner hat method (FiTCH) Meng Liu and Linqing Yang https://doi.org/10.6084/m9.figshare.22647382.v1
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
795 | 260 | 62 | 1,117 | 39 | 50 | 62 |
- HTML: 795
- PDF: 260
- XML: 62
- Total: 1,117
- Supplement: 39
- BibTeX: 50
- EndNote: 62
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1