Review Reports

Round 1

Reviewer 1 Report

The manuscript by Mouna Amroussia et al. “Predicting spatially explicit Composite Burn Index (CBI) … ” presents a study conducted in two large forest fires occurred in 2021 in northern Tunisia in order to derive a suitable spatialization of fire severity.

The work exploits Sentinel 2A images acquired before and after fire events and analyse the correlations between field computed Composite Burn Index (CBI) and several spectral indices suggested in literature with different regression models.

The specific contribution of this work is mainly related to field work done to obtain Composite Burn Index (CBI) in correspondence to various fire severity in such environment in the South Mediterranean areas of North Africa.

Comments:

- Figure 1 presents false colour composites of Sentinel-2A of the Fernana (F1) and Takrouna (F2) study areas; please make more explicit the description on how burned area (fire perimeter) map was obtained.

- One of main merit of this work resides in the field data collection in about two hundred sites following Key&Benson and Fernández-García et al. “field protocol”. For this reason, besides a general presentation of two study areas, we expect to see a more detailed description in terms of fire severity level of the 8 sites whose pictures are shown in figure 2, or at least to read for each picture which level of fire severity is assigned by the authors.

- What is the statistical distribution of severity levels as obtained in the maps of predicted CBI values (Figure 12) ? Make a comment in comparison with distribution of ground computed CBI as shown in Figure 3a.

Line 35 -36  pattern of biomass consumption [4, 1, 5, 6), à …, 6]

Figure 6. To facilitate the comparison between the different spectral indices used as explanatory variables, the y-axis CBI should cover the same range in all the four. As indicated in fig. 5 and fig. 7 the observed CBI values range from  0.0 to 3.0.

Check duplications in References paragraph and related numbering in the manuscript.

Line 505-506  24. Eva, H.; Lambin, E.F. Remote Sensing of Biomass Burning in Tropical Regions: Sampling Issues and Multisensor Approach. Remote Sens. Environ. 1998, 64, 292–315

Line 576-577  59. Eva, H.; Lambin, E.F. Remote Sensing of Biomass Burning in Tropical Regions: Sampling Issues and Multisensor Approach. Remote Sens. Environ. 1998, 64, 292–315

Line 582-583  62. Epting, J.; Verbyla, D.; Sorbel, B. Evaluation of Remotely Sensed Indices for Assessing Burn Severity in Interior Alaska Using  Landsat TM and ETM+. Remote Sens. Environ .2005, 96, 328 -339.

Line 599-600  70. Epting, J.; Verbyla, D.; Sorbel, B. Evaluation of Remotely Sensed Indices for Assessing Burn Severity in Interior Alaska Using Landsat TM and ETM+. Remote Sens. Environ. 2005, 96, 328 339.

Author Response

 Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The paper is well written but the topic and approach are not particularly novel and the scope of the study is small, limited to 2 fires.  I recommend adding some additional detail to the methods and re-organizing the Results section to better facilitate comparisons across the models. The Discussion section was very nicely written.

Section 2.2.2 – how did the date gap between the post-image and the field work influence uncertainty. It is 2-3 months between the acquisition date and when the CBI plots were collected, did any re-vegetation occur in the intervening period? Please add text that describes this source of uncertainty and how it may influence the results.

 

Section 2.2.4 – please add either to the text or as a table all of the final parameters (e.g., learning rate, number of trees) used in each of the 4 models.

Figure 4 – The predicted values are limited to 5 different values, is this a product of a limited tree size? Consider adding a comment to the text that this is one of the limitations of this approach.

Figures 4, 5, 7, 10 – since one of the major goals of the paper is to compare statistical models, the current way these figures are organized makes it really hard to directly compare them. I don’t think the a) panel is necessary – the training predicted vs observed. I would recommend deleting the a) panels and compiling the 4 b) panels of the validation predicted vs observed into a single figure so that readers can directly compare the scatter plots.

Figure 6, 8, 9, 11 – same comment as above, it was very difficult to compare between the 4 models, I would recommend still having 4 figures, but have the first figure show the partial dependence of CBI on RBR for all 4 models, etc.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

The authors have adequately addressed my comments and the paper is acceptable for publication.