the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
What controls fire size in the South American Gran Chaco? Exploring atmospheric, landscape, and anthropogenic drivers
Abstract. Wildfires are key ecological agents in the Gran Chaco, one of the world's largest dry forest ecosystems, where fire regimes are increasingly shaped by human pressure and climate variability. However, the drivers of fire size variability remain poorly understood. We analysed over 100,000 fire patches (2001–2022) from the FRYv2.0 database to assess environmental controls on fire size and morphology across the Wet, Dry, and Very Dry Chaco. High-resolution fire polygon data were combined with ERA5-Land reanalysis, vegetation and topographic metrics, and anthropogenic layers. Fire sizes were highly skewed: >80 % were <5 km², yet large events (Megafires >100 km², Gigafires >1000 km²) dominated burned area (BA). Gigafires were rare but mostly confined to the Dry Chaco, whereas the Wet Chaco had the highest BA, fire frequency, and Megafire count. Fire Weather Index (FWI)–BA correlations reached r = 0.7 in the Wet Chaco but were weaker and spatially fragmented in drier subregions, where fuel continuity and ignition context played larger roles. Lag analyses showed that in drier areas, wet-season biomass buildup (4–6 months prior) increased subsequent fire activity, while in wetter areas short-term dryness (1–3 months prior) was more predictive. During-fire meteorology, especially persistent strong winds, better explained fire morphology than pre-fire conditions. Random Forest models ranked static landscape features (elevation, land-cover evenness, slope, tree cover) highest in size prediction. Our results reveal region-specific fire–environment couplings, clarifying the interplay of meteorological, ecological, and anthropogenic factors, and providing actionable insights for fire risk forecasting and management in the Gran Chaco.
- Preprint
(4622 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-3484', Anonymous Referee #1, 14 Aug 2025
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-3484/egusphere-2025-3484-RC1-supplement.pdfCitation: https://doi.org/
10.5194/egusphere-2025-3484-RC1 -
RC2: 'Comment on egusphere-2025-3484', Anonymous Referee #2, 30 Sep 2025
What controls fire size in the South American Gran Chaco? Exploring atmospheric, landscape, and anthropogenic drivers.
General comments
The article analyzes the main drivers of fire size in the Gran Chaco forest of South America—one of the world’s largest and most threatened dry forest ecosystems. It uses a fire database derived from medium-resolution remote sensing data, covering a substantial temporal span (2001–2022) suitable for this type of analysis. The study focuses on short- and mid-term weather conditions that may influence burned area extent and fire characteristics at a regional scale, yielding novel and insightful findings. In addition, it examines various environmental factors contributing to fire size.
I believe the manuscript would benefit from a more in-depth analysis of the results, along with a reconsideration—or at least a clearer explanation—of their interpretation. Additionally, the discussion would be strengthened by a more thorough contextualization of the findings in light of the existing literature on the ecology and environmental processes in the Gran Chaco over recent decades. While some relevant studies are cited, there are other works—albeit not covering the entire Chaco ecoregion—that could provide valuable points of comparison or support for the results presented.
One of my main concerns relates to the Random Forest (RF) analysis. I am not fully convinced that elevation is truly the most important variable explaining fire size. Although the manuscript acknowledges that elevation is not a direct driver, the current explanation of its role lacks depth. Given that elevation consistently appears as a key predictor across all subregions and seasons, the authors should explore more thoroughly the ecological and land-use processes that might underlie this result. At present, the argument that elevation reflects broad ecological gradients in vegetation composition, fuel moisture regimes, and land-use history is too vague to be convincing. I strongly suggest repeating the analysis excluding elevation as an independent variable to better understand the role of other, potentially more direct, drivers.
That said, I believe the study has strong potential, particularly given the wide range of variables included and the novelty of this type of analysis for the Gran Chaco. I encourage the authors to strengthen the connections between their findings and the existing local literature relevant to the study area.
The article requires Major revisions to be accepted.
Specific comments
L42-44: I believe this statement is not sufficiently supported by the references cited. If there is evidence to back this claim, the authors should explain the connection more clearly. Specifically, are there studies demonstrating that areas invaded by exotic grasses tend to experience larger or more intense fires? Clarifying this link is essential to strengthen the argument and ensure it is grounded in existing research.
The article by D’Antonio, as well as Bravo et al. (2014), may indeed support a potential relationship between exotic grasses and increased fire intensity. However, to substantiate this argument in the context of the Gran Chaco, it is important to clarify how extensive these exotic grass-invaded areas actually are within the region. Such areas would need to be considerable in extent to justify the strength of the statement currently made in the manuscript. Moreover, there may be other underlying ecological or land-use processes—currently not discussed—that could more convincingly explain observed changes in fire regimes. Including these considerations would significantly strengthen the interpretation of your results.
The article by Naval Fernández et al. 2023 may support the idea of climate variability and larger fires: Naval Fernández, M. C., Albornoz, J. V., Bellis, L. M., Baldini, C., Arcamone, J. R., Silvetti, L. E., ... & Argañaraz, J. P. (2023). Megaincendios 2020 en Córdoba: Incidencia del fuego en áreas de valor ecológico y socioeconómico. Ecología Austral 33: 136-151.
L45: Here you refer to traditional land management or suppression strategies? I suggest rewritting these sentence to make your idea clearer.
L51-52: I suggest rewritting this sentence as Kelley et al. and Jones et al do not specifically focus in the Gran Chaco.
L68-69: Please, specify the characteristics of the environment in which the study of Saucedo and Kurtz was performed, as it is a particular environment of the Gran Chaco.
L92: climate or weather?
L97, Section 2.1. The introduction or Section 2.1 should describe better available knowledge about the fire regimes in the Gran Chaco ecoregion.
Also, it is not recommended to begin a section with a figure or table. This comment applies to the whole manuscript.
L114: clarify that these are Argentinean provinces, as readers from other countries may not be familiar with them.
L155-158: Please indicate at least the spatial resolution of the land cover product. Also, please explain why the ESA product was a better choice than annual land use-land cover maps from MapBiomas Chaco? https://chaco.mapbiomas.org/ . Global products may have significant errors in local to regional studies.
L192: Two terms are used in the same way: Fire Polygons and Fire Patches (FPs). I recommend using the same term along the manuscript.
L267: change “shaping fire activity” by “determining fire size”
L271: Here you mention fire frequency, but the Section’s title only refers to fire size.
L275, Table 1: Please, provide more detail in the Table's Caption, to be more interpretable outside of the context of the manuscript. For instance, you might mention that these are potential predictor variables of fire size in the Gran Chaco.
Table 1: Instead of “Climatic” it should be “meteorological”
Table 1: Did you obtained Land cover composition variables from the map of same year as the fire patch? I guess you did, but it would be better to explicitly mention this, as this is a dynamic variable. The same would apply for Landscape heterogeneity variables.
L289: You refer to “predictor variables” and so far, they are independent or potential predictors until the statistical analysis confirms their predictive value.
L309, Figure 3: “ignitions” in panel b, refer to the fire counts mentioned in section 2.3.3? this is confusing. Then, in lines 312 and 2013 you mention "number of fire polygons". I recommend using fire counts, as ignitions are probably much higher.
L314-316: Linear regressions were not mentioned in the Materials and Method section. The whole sentence is quite confusing, since fire extent may refer to individual fires and here it is associated with ignition frequency.
L317: “suggesting a greater role of other drivers in the later”. I don't think you should use the term "drivers" here, as you are not studying the drivers of annual burned area. The weaker association would be caused by despair fire sizes through time, which is confirmed in Figure 5.
L318: Isn't it R2?
L320: I find quite confusing the use of "ignition frequency" to refer to fire count. I recommend changing the expression.
L350: According to Figure A3, Wet Chaco had 2 Gigafires (GF) and Very Dry Chaco had 1GF. However, you include very dry chaco in the same level as the Dry Chaco.
L367: I would recommend not using the term fire frequency to refer to the number of fires, as it may be confusing with the fire metric that refers to the number of times an area burned in a certain period of time.
L393: isn't it September? (instead of October)
Figure 10: I suggest using some level of transparency, so when they overlap a more intense color would appear.
L492-497: Further interpretation of results is needed here...what does this means? the text focused solely on percentages, which are available at Fig. 10, but I would like the authors explain the meaning of these results. Also, I wonder if wind driven fires are larger, as the green color in Figure 10 highlights more than the other FWT.
Figure 11: It would be helpful for readers if you use the same colors for each FWT in the different figures.
L509-514: I agree that Figure 10 shows some interesting results of comparing pre- and during-fire clusters, but it is not clear to me to repeat this in Figure 11, as you analyze weather data ± 20 days before and after ignition. In the pre-fire cluster, the time after ignition corresponds to the time of “during fire”, and the opposite applies to the during-fire cluster. Also, your analysis of Figure 11 seems to focus on the pre-fire cluster. For instance, wind speed after ignition in the during-fire cluster is striking, but nothing is mentioned about it in the text.
L515-519: The meaning of each morphology index should be explained in the Materials and Methods section. Without the exact definition of each. It stands out to me that wind driven fires are more elongated, but have higher core-area index and lower perimeter-to-area ratio. Then, what does elongated mean? I would think that an elongated fire would have higher perimeter-to-area ratio.
L524, Section 3.6: There is no metric/index providing information about the goodness of your RF model/s, and this would be valuable to interpret and discuss your results.
L525-527: this paragraph is not necessary as it contains information from the M&M section.
Figure 13, Caption: Focus on the meaning. In this figure you want to show the contribution of each variable. Then I suggest to focus the caption with the concept and then clarify that you used SHAP.
Figure 15: what about fires at higher elevations? the plot reaches 425 m asl, but there are fires at higher elevations. Did you fix the length of the x-axis due to fewer cases? Also, what can you say about the distribution of elevation values...in your study area most land is below 500 m asl. The same question applies to slope effects: I guess you may have a small proportion of land with steep slopes.
L559-561: My interpretation of LC evenness is that there are large fires (bluer dots) all along the range of this predictor variable...unless I am reading it wrong. Instead, I agree with your interpretation of tree cover. Also, in relation to this variable, the range in the figure is 0 - 1, so you seem to have used proportion, instead of percentage.
L562-564: I see other relationships in Fig 15. For example, at certain cattle density (the variable is named "number of cattle" but it is expressed as a density) you have very low fire activity, which could be associated with lower fuel availability to burn. If I am interpreting the figure wrong, please provide more detail in the manuscript for other readers.
Also, for croplands/natural mosaics, I see that larger fires occur at proportions lower than 0.5, and then large fires are less common.
L583-589: I recommend the following papers to consider in this part of the discussion:
-Bravo, S., Kunst, C., Grau, R., & Aráoz, E. (2010). Fire–rainfall relationships in Argentine Chaco savannas. Journal of Arid Environments, 74(10), 1319-1323.
-Argañaraz, J. P., Pizarro, G. G., Zak, M., Landi, M. A., & Bellis, L. M. (2015). Human and biophysical drivers of fires in Semiarid Chaco mountains of Central Argentina. Science of the Total Environment, 520, 1-12.
L621-628: I am honestly not convinced that elevation is the main variable explaining fire size. While the manuscript acknowledges that elevation is not a direct driver, I would encourage the authors to explore more deeply the underlying factors that could be contributing to its apparent importance in the analysis. As elevation emerges as a key predictor across all subregions and seasons, the current explanation—that it reflects broad ecological gradients in vegetation composition, fuel moisture regimes, and land-use history—feels too general and insufficiently supported. A more detailed and evidence-based discussion is needed to clarify this interpretation and to justify the prominence of elevation in the results.
L629-631: Here the shrub cover is mentioned as linked to large fires, but this was not highlighted when analyzing Fig. 15.
L635-636: There are some papers addressing fuel moisture content in certain areas of the Chaco forest that it would be valuable to consider here:
-Argañaraz, J. P., Landi, M. A., Scavuzzo, C. M., & Bellis, L. M. (2018). Determining fuel moisture thresholds to assess wildfire hazard: A contribution to an operational early warning system. PloS one, 13(10), e0204889.
-Arganaraz, J. P., Landi, M. A., Bravo, S. J., Gavier-Pizarro, G. I., Scavuzzo, C. M., & Bellis, L. M. (2016). Estimation of live fuel moisture content from MODIS images for fire danger assessment in Southern Gran Chaco. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 9(12), 5339-5349.
-Bianchi, L., Defossé, G., Dentoni, M., Kunst, C., Ledesma, R., & Bravo, S. (2014). Dynamics of fuel moisture and its relation to the ecology and management of fire in the western Chaco region (Argentina) I: basic concepts.
L650-651: It would be valuable to compare your results with other studied performed in areas partially overlapping with your study area:
-Fischer, M. A., Di Bella, C. M., & Jobbágy, E. G. (2012). Fire patterns in central semiarid Argentina. Journal of Arid Environments, 78, 161-168.
L665-669: I strongly disagree with this comment. While fire is indeed widely used as a land management tool, in most cases it is not applied under controlled or regulated conditions. In fact, its use is often informal or even illegal, depending on local regulations. Moreover, there are additional sources of fire ignition in the region, including hunting practices, waste burning, and accidental ignitions. The current description of prescribed burning, as well as the way the reference to Hsu et al. (2025) is presented, gives the impression that their work characterizes fire use in the Gran Chaco ecoregion—which it does not. If you wish to retain the citation of Hsu et al., I recommend rewriting the sentence to clarify that their findings do not directly reflect the specific fire management context of the Gran Chaco.
Citation: https://doi.org/10.5194/egusphere-2025-3484-RC2
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
2,020 | 84 | 13 | 2,117 | 35 | 30 |
- HTML: 2,020
- PDF: 84
- XML: 13
- Total: 2,117
- BibTeX: 35
- EndNote: 30
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1