The paper entitled: “Environmental Factors Affecting Wildfire Burned Area In South-Eastern France, 1970-2019” investigates the spatiotemporal evolution in burned area over a 50-year period (1970-2019) and its interactions between topography (slope inclination and aspect) and vegetation type in south-eastern France by exploiting Geographic Information System databases.

I think this paper has a great potential for the large dataset of mapped burned scars; but authors don´t take advantage of this long temporal fire geodatabase. I think this paper is very simplistic mainly because of authors only show a descriptive analysis of the burned area frequency in different topographic features and vegetation types. This paper doesn´t make any scientific contribution in its current form. Following, I propose several statistical approaches to improve this paper:

1. i) Frequency analysis should be carried out using more appropriate statistics as the Resource Selection Indices (please see Moreira et al. 2009; Moreno et al. 2011; among others):

Moreira F, Vaz P, Catry F, Silva JS (2009) Regional variations in wildfire susceptibility of land-cover types in Portugal: implications for landscape management to minimize fire hazard. International Journal of Wildland Fire 18, 563–574. doi:10.1071/WF07098

Moreno, J. M., Viedma, O., Zavala, G., & Luna, B. (2011). Landscape variables influencing forest fires in central Spain. International Journal of Wildland Fire, 20(5), 678-689.

1. ii) Trends could be assessed by a Man-Kendall Trend analysis, for example. Changes in the burned area time series could be assessed by breaking points or Pettit analysis. Here there are some examples:

Urbieta, I. R., Franquesa, M., Viedma, O., & Moreno, J. M. (2019). Fire activity and burned forest lands decreased during the last three decades in Spain. Annals of Forest Science, 76(3), 1-13.

Moreno, M. V., Conedera, M., Chuvieco, E., & Pezzatti, G. B. (2014). Fire regime changes and major driving forces in Spain from 1968 to 2010. Environmental Science & Policy, 37, 11-22.

iii) Spatial patterns Analysis in a GIS: spatial clustering (k- Ripley), Getis-Ord index, spatial autocorrelation…

1. iv) Size patterns: Gini index
2. v) Geographically Weighted Regression to assess the role of the different topographic and vegetation factors over time. These models could be carried out at decadal scale to see the different role of those independent variables. See for example:

Nunes, A. N., Lourenço, L., & Meira, A. C. (2016). Exploring spatial patterns and drivers of forest fires in Portugal (1980–2014). Science of the Total Environment, 573, 1190-1202.

There are some typographic errors over the paper.

SPECIFIC COMMENTS

In introduction other several references could be included to support many sentences. Overall, the objective of this paper is to assess temporal changes in BA spatial patterns but there is not any reference about this topic.

Please see: Viedma et al. 2018 although it is related to fire number (fire frequency).

Viedma, O., Urbieta, I. R., & Moreno, J. M. (2018). Wildfires and the role of their drivers are changing over time in a large rural area of west-central Spain. Scientific reports, 8(1), 1-13.

For example, to support this sentence:

“Among the environmental characteristics, several studies provide evidence of spatial patterns relating topography to forest fire probability (Dickson et al., 2006; Padilla and Vega-García, 2011)”

Please, read and include this paper:  Viedma, O., Urbieta, I. R., & Moreno, J. M. (2018). Wildfires and the role of their drivers are changing over time in a large rural area of west-central Spain. Scientific reports, 8(1), 1-13.

To support this sentence:

“Csontos and Cseresnyés (2015) observed an exponential velocity increase in upslope fire spread with the increase in slope inclination whereas downslope fire spread velocity was unaffected by slope angle and was similar to rates detected on flat terrain”

Please, read and include these papers:  

Viedma, O., Quesada, J., Torres, I., De Santis, A., & Moreno, J. M. (2015). Fire severity in a large fire in a Pinus pinaster forest is highly predictable from burning conditions, stand structure, and topography. Ecosystems, 18(2), 237-250.

Viedma, O., Chico, F., Fernández, J. J., Madrigal, C., Safford, H. D., & Moreno, J. M. (2020). Disentangling the role of prefire vegetation vs. burning conditions on fire severity in a large forest fire in SE Spain. Remote Sensing of Environment, 247, 111891.

To support this sentence:

“…and the probability of large fires in landscapes with dense shrublands is greater than in forested ecosystems in the Mediterranean basin (Moreira et al., 2011; Ruffault and Mouillot, 2017).”

Please, read and include this paper:  

Urbieta, I. R., Franquesa, M., Viedma, O., & Moreno, J. M. (2019). Fire activity and burned forest lands decreased during the last three decades in Spain. Annals of Forest Science, 76(3), 1-13.

Here, authors showed that treeless areas tend to burn more than treed areas in Spain during the last decades.

To support or enlarge this sentence with the trends during the last decades:

“…broadleaved forests are usually less prone to burning than coniferous species which present a greater fire hazard (Moreira et al., 2009; Oliveira et al., 2014).”

Please, see Urbieta et al. 2019: In Spain oak forests are burning more than conifers in the last decades.

There is a great confusion with the cell size of the grid to extract the frequency data. For example, in Lines 124: “A 500x500 m grid was created and overlaid on the study area in order to measure the percentage of the area that was burned inside every 25 m cell for each year”.  Later, author say that each cell is 25 ha but a 25 x 25 m cell is 625 m². Sorry, but I don´t understand anything. Please, clarify this.

On the other hand, it is said that “…BA by vegetation type used the CLC layer closest to the BA data”

 Please, be careful because the dates of the LULC maps were not before of several forest fires and we expect that you have checked that the LULC represented by the maps always indicated prefire vegetation.

In fig. 5 and later over the paper, you use the term “fuel type”. Please, change it by forest type or vegetation type or even Land cover type; because you are not working with fuel types, but only with land covers.

Please, improve the quality of fig. 6 and others done in Excel. Remove internal lines of the plot and be careful with the borders of the figure. Letters in black better than grey.

This paragraph must be in discussion section, not in results:

“Overall, the patterns described here are coherent with known interactions between fire ignition, vegetation continuity, and wind speed: fire ignition occurs most frequently in proximity to human activities (Badia et al., 2011; Chas-Amil et al., 2013; Jiménez-Ruano et al., 2017; Lampin-Maillet et al., 2011) and BA depends on fuel continuity and wind speed (Dueane et al., 255 2015; Fernandes et al., 2016). “

This sentence reflects one of the limitations of this paper:

Line 265: “However, on the northern shore of the Mediterranean, there are generally more S-facing (sum of SW, S, SE) than N-facing (sum of NW, N, NE) slopes, and BA distribution may therefore be a simple reflection of area rather than susceptibility to burn. “…”In order to compensate for this, BA is plotted as a percentage of the burned forested slopes”

As you say, this is a limitation and other type of statistical analysis should be carried out as the resource Selection Index:

See this paper: Moreno, J. M., Viedma, O., Zavala, G., & Luna, B. (2011). Landscape variables influencing forest fires in central Spain. International Journal of Wildland Fire, 20(5), 678-689.

Resource selection index (RSI). The RSI is commonly used in studies of habitat selectivity by animals (Manly et al. 1993). We used Savage’s (1931) forage ratio: Wi=Ui/Ai. The index is calculated as follows for LULC types (similar calculations were made for the other variable): Ui designated the area of LULC class i burned by each fire divided by the total area of that fire, and Ai represented the area covered by LULC class i in the entire study site divided by the total area of the study site. A class with a burned area proportionate to its availability was thus represented by the value Wi=1, a class with a burned area exceeding that expected by chance (i.e. ‘selected’ by a fire) had a value Wi > 1, and a class with a less-than-expected burned area (i.e. ‘avoided’ by a fire) had a value Wi < 1.

There is confusion in the figure numeration:

It is not figure 9 but figure 8. The same with figure 10 in line 261 (it is figure 8)

To support this sentence and make comments in discussion section:

Below 30°, there are no clear differences between slope inclination categories. Above 30°, the percentage BA drops abruptly. Temporal fluctuations of the distributions show a general shift from high inclination slopes (40° or greater) to lower inclination slopes (≤20°).”

Results are like those found in Viedma et al. 2018: a shift of fire frequency to flatter areas.

See: Viedma, O., Urbieta, I. R., & Moreno, J. M. (2018). Wildfires and the role of their drivers are changing over time in a large rural area of west-central Spain. Scientific reports, 8(1), 1-13.

Please reconsider to change the figure caption of figures 12 and 13. I propose these:

Fig. 12. Percentage of burned vegetation according to the area of vegetation types by decade

Fig. 13. Percentage of burned vegetation according to the total burned area by decade

n this manuscript, the authors investigate whether (and if yes, how) topography and vegetation types affect the probability of burning in three administrative regions located in south-eastern France over the period from 1970 to 2019. Their main objective is to determine whether the proportion of fires that burn a certain land cover type or topographical configuration (disproportionally to their availability) has changed over time as a consequence of the implementation of a new fire policy that took place in France in the early 1990s and showed high efficiency at reducing burnt areas and fire numbers. For this purpose, they analyze a long-term georeferenced fire-shape dataset (which in my knowledge has never been analyzed but by Ganteaume and Barbero 2018, doi.org/10.5194/nhess-19-1055-2019) and compared this dataset with LULC data over time. The main results show the increasing importance of south-facing slope over time and higher burned area in Sclerophyllous vegetation than in other vegetation types.

While the subject of this manuscript is of interest to the fire community and the data gathered sounds relevant to answer the general hypotheses outlined here above, I have a number of broad concerns and comments that should be carefully addressed before I feel comfortable recommending this paper for publication. My opinion is that the manuscript suffers from serious flaws, including (i) inappropriate analyses of the data and over-interpretation of the results, (ii) some doubts regarding the homogeneity of the wildfire database that might lead to some bias in the interpretation of the results, (iii) the need to better use the literature to introduce and discuss the hypothesis and results of this work. To put it clearly, I think that the issues raised in this paper and data gathered to answer them are promising but it is most likely that addressing these concerns would require an important amount of work, including new analyses and a great amount of rewriting. I detail below these three main concerns as well as other comments I noted while reading this manuscript. As it is my belief that this manuscript would require important rewriting, I did not feel it would be relevant to provide, at this stage, a list of very detailed comments.

(i) Perhaps, my main concern regards the analyses whose results are shown in Fig, 8,9,10, 11, 12, and 13) and that do not seem appropriate for the objectives raised in this paper.  Indeed, despite an overall agreement of your results with the literature, there is no guarantee that the conclusions drawn from your current analyses are not affected by other drivers of fire spread, such as fuel connectivity, weather conditions, and many other possible interactions including those that might occur between your studied factors. The presentation of results is very convoluted (see for instance the section from L261 to L279 that is very difficult to follow) and no framework currently rigorously demonstrates that LULC influence is significant from a statistical point of view. In light of these remarks, it seems essential to rethink the analytical framework to ensure that these three points are correctly addressed by developing a more appropriate analytical framework that takes into account (or at least minimizes) these interactions and presents the results in a more relevant manner. One interesting possibility would be to determine whether the observed values of fire selectivity are significantly different from those that would be observed if fires were to occur at random locations in the landscape. I would advise you to look at the methodologies shown in Moreno et al. (2011) and Barros et al. (2014) that seem relevant because they assess fire selectivity through a null-based model that is independent of spatial relationships or any bias of fire data. Note that other approaches might also be relevant. Besides I found that the approach developed in Fig 7 and the conclusions drawn from these analyses were highly descriptive and speculative (see L224-L255). No analyses on the spatial pattern are provided, simply a description of the maps. These analyses should therefore also be improved to strengthen the results and their interpretation.  

Barros, A. M., & Pereira, J. M. (2014). Wildfire selectivity for land cover type: does size matter?. PloS one, 9(1), e84760.

Moreno, J. M., Viedma, O., Zavala, G., & Luna, B. (2011). Landscape variables influencing forest fires in central Spain. International Journal of Wildland Fire, 20(5), 678-689.

(ii) I was surprised by the results in Fig 6 showing no trend in fire numbers, as it differs from the decreasing trend generally reported for this area see for instance Fig. 2 in Curt and Fréjaville, 2018). A possible explanation for such a discrepancy is that your database could be biased by a non-constant reporting of fires over time, which in turn might affect your results and conclusions. What I suspect is that a higher proportion of fires are now reported compared to what it was in the early 19070’s especially for the smallest fires. To reduce the potential biases induced by this inconsistency, one solution would be to compare your dataset against the french promethee fire database for fires > 1ha (that is currently considered to be a relevant and robust fire size threshold to study fire ignitions in France, Pimont et al. 2021) in order to determine the fire size threshold above which you consider your database is not affected by an evolution of detection /reporting over time. Furthermore, I think that this manuscript would really benefit from including, somehow, a size factor in the analytical framework. Indeed, it would be relevant to test whether under severe fire weather, fires are expected to become larger and less dependent on land cover, which is generally reported in southern Europe.

(iii) I found that references to previous studies were not enough detailed, sometimes vague, or even missing and my opinion is that this manuscript could be greatly improved on this matter. As mentioned in one of my previous comments, several important papers, whose vast majority address fire regimes over the Iberian Peninsula, investigate the impact of LULC on fires but were not cited in this manuscript, including for instance Carmo et al. (2011), Bajocco et al. (2008), Moreno et al. (2011), Nunes et al. (2005), Koutsias et al. (2012) among many others.  Furthermore, like any other area, the French Mediterranean region has its own peculiar context and history regarding fire regimes, whose description and analyses could also be improved. For instance, in France, the papers from Fréjaville and Curt (2016), Ruffault and Mouillot (2015), Ruffault et al. (2016), and Evin et al. (2018) studied the consequences of the introduction of this new fire policy on different metrics. Some of their results and discussions might provide relevant results for your discussion and the building of your hypotheses.  There are a few papers also that have explored the drivers of spatial fire weather and or fire hazard in Southern France that might help, including the works from Ruffault et al. (2017) and Pimont et al. (2021). Note that this is by no means a list of papers that need to be cited but rather some references that the authors you might find useful to improve the quality of your manuscript.

- Evin, G., Curt, T., & Eckert, N. (2018). Has fire policy decreased the return period of the largest wildfire events in France ? A Bayesian assessment based on extreme value theory. Natural Hazards and Earth System Sciences, 18(10), 2641-2651. https://doi.org/10.5194/nhess-18-2641-2018

- Fréjaville, T., & Curt, T. (2017). Seasonal changes in the human alteration of fire regimes beyond the climate forcing. Environmental Research Letters, 12(3), 035006. https://doi.org/10.1088/1748-9326/aa5d23

- Ruffault, J., & Mouillot, F. (2015). How a new fire‐suppression policy can abruptly reshape the fire‐weather relationship. Ecosphere, 6(10), 1-19. http://dx.doi.org/10.1890/ES15-00182.1

- Ruffault, J., Moron, V., Trigo, R. M., & Curt, T. (2016). Objective identification of multiple large fire climatologies : an application to a Mediterranean ecosystem. Environmental Research Letters, 11(7), 075006. doi:10.1088/1748-9326/11/7/075006

- Ruffault, J., Moron, V., Trigo, R. M., & Curt, T. (2017). Daily synoptic conditions associated with large fire occurrence in Mediterranean France: evidence for a wind‐driven fire regime. International Journal of Climatology, 37(1), 524-533. https://doi.org/10.1002/joc.4680

- Pimont, F., Fargeon, H., Opitz, T., Ruffault, J., Barbero, R., Martin‐StPaul, N., ... & Dupuy, J. L. (2021). Prediction of regional wildfire activity in the probabilistic Bayesian framework of Firelihood. Ecological applications, e02316. https://doi.org/10.1002/eap.2316

- Carmo, M., Moreira, F., Casimiro, P., & Vaz, P. (2011). Land use and topography influences on wildfire occurrence in northern Portugal. Landscape and Urban Planning, 100(1-2), 169-176. https://doi.org/10.1016/j.landurbplan.2010.11.017

- Bajocco, S., & Ricotta, C. (2008). Evidence of selective burning in Sardinia (Italy): which land-cover classes do wildfires prefer?. Landscape Ecology, 23(2), 241-248.

- Moreno, J. M., Viedma, O., Zavala, G., & Luna, B. (2011). Landscape variables influencing forest fires in central Spain. International Journal of Wildland Fire, 20(5), 678-689. https://doi.org/10.1071/WF10005

- Nunes M, Vasconcelos M, Pereira J, Dasgupta N, Alldredge R, et al. (2005) Land cover type and fire in Portugal: Do fires burn land cover selectively? Landscape Ecology 20: 661–673. https://doi.org/10.1007/s10980-005-0070-8

- Pereira, M. G., Aranha, J., & Amraoui, M. (2014). Land cover fire proneness in Europe. Forest Systems, 23(3), 598-610. http://dx.doi.org/10.5424/fs/2014233-06115

- Koutsias, N., Arianoutsou, M., Kallimanis, A. S., Mallinis, G., Halley, J. M., & Dimopoulos, P. (2012). Where did the fires burn in Peloponnisos, Greece the summer of 2007? Evidence for a synergy of fuel and weather. Agricultural and Forest Meteorology, 156, 41-53. https://doi.org/10.1016/j.agrformet.2011.12.006