Forty-Year Fire History Reconstruction from Landsat Data in Mediterranean Ecosystems of Algeria following International Standards

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Authors,

Relatively a very well written manuscript covering the BA for the Algeria region. English shall be corrected for proper sentences and third tense. Following suggestions are provided for further improvement of the manuscript:

1. Title may be modified for inclusion of Algeria, as the study primarily focus on Algeria. Currently its appears a much larger area has been studied, so giving misconception.

2. prepare separate section for Datasets/ material.

3. Methodology section should be separate from "Materials and Methods" to give proper emphasis and focus on section preparation.

4. Flowchart shall be improved. Expand essential terms, for more clarity in flowchart such as: VA Dates Tool, RP Tool, RPs. Through proper utilization of space in flowchart some Acronyms may be expanded for clarity.

5. Conclusion shall be improved and made strong on basis of the current presented study.

6. Introduction may be improved by inclusion of some of the studies of global interest in the world such as on Himalaya. Suggest following references:

a. https://doi.org/10.1007/s11069-023-05835-z

b. https://doi.org/10.5194/isprs-archives-XLII-5-469-2018

Some redundant references can be reduces, as a large number of references are used, preferably looking at the study ~ 50-60 is shall be sufficient..

Comments on the Quality of English Language

Third tense has not been used at many places.

Overall its a good writing. So, minor/moderate corrections are suggested

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

40-year fire history reconstruction from Landsat data in a Mediterranean area of North Africa following international standards

By Mostefa E. Kouachi, Amin Khairoun, Aymen Moghli, Souad Rahmani, Florent Mouillot, M. Jaime Baeza and Hassane Moutahir

 

The manuscript “40-year fire history reconstruction from Landsat data in a Mediterranean area of North Africa following international standards” focuses on a development of a Landsat-based regional burned area product (NEALGEBA) for North Eastern part of Algeria. The authors used Google Earth Engine tools to create annual burned area maps for 1984–2023 and to validate the product using Sentinel-2 data as well as MODIS and VIIRS active fire products. An extensive comparison of the generated product with several other burned area products was also performed showing its high overall performance. The manuscript is well written, relevant, and is within scope of Remote Sensing. I have only several minor comments regarding the manuscript.

Lines 288-289: I was not able to understand this sentence concerning the selection of image dates for the mosaic generation. Even within the same Landsat scene, it is possible that several fires could exist for which the lowest NBR dates will be different. Perhaps additional explanation should be added here.

Figures 4, 7 and section 3.1 report burned areas in hectares; however Figure 10b and sections 3.2.1 and 3.3.1 report it in km2. I think it would be reasonable to keep the same units across the whole manuscript.

Figures 5 and B8: Please, consider increasing font size make the figure easier to read.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

This paper presents a region-specific burned aera dataset which was derived using state-of-the-art remote sensing data sources (Landsat multispectral imagery) and data processing tools (BA Mapping Tools published by Roteta et al., 2021).

While the methods are not highly innovative (the authors applied existing methods to existing data), this study produced a new dataset of clear interest to the scientific community as well as fire and land managers. The methods are described in detail, which makes the analysis reproducible and extendable to other regions. Moreover, the manuscript is well written and structured.

However, I have two significant concerns which I think should be addressed before this manuscript can be published:

First, I have not yet understood the added value of these new results in comparison to the GABAM dataset referenced by the authors. The GABAM burned area product is global, was derived from Landsat imagery, has a 30-m spatial resolution and covers the timeframe 1985-2019. In the introduction, the authors refer to the “inherent coarse spatial resolution” (line 98) of existing datasets as a justification to produce theirs. However, their resulting dataset still has a 30-m spatial resolution, which does not provide a significant improvement over other currently available products. The comparison to other existing products and the discussion of how this work provides an advancement in the state of science should be improved.

Second, I found important limitations in the validation strategy described in Sections 2.4 and 2.5. If I understood it correctly, what the authors call validation is a comparison of the BA estimates derived from the proposed methodology against the results of applying the same BA classification methodology to Sentinel-2 data instead of Landsat’s. This means that Random Forest (RF) classification outputs are being compared against RF classification outputs –obtained with a similar RF implementation-- as a means to validate the former. It is not surprising that this framework provides high similarity metrics between NEALGEBA and the S2RD reference dataset.

This limitation in the validation methodology permeates into the comparison with other datasets. The strong dependency between NEALGEBA and the S2RD reference dataset makes it also unsurprising that NEALGEBA received higher validation scores than other –truly independent—datasets. It is surprising that NEALGEBA results were not compared directly with those other existing datasets to evaluate their agreement.

This links back to my first concern –clarity on the added value of NEALGEBA. If the primary improvement of NEALGEBA is the geographical or temporal extension of BA estimates, then outputs from the proposed methodology should be compared to other existing datasets where and when data from both is available. If the main addition of NEALGEBA is a claimed improvement in accuracy, comparison of the method against itself is not a valid validation. If the claimed improvement is an increase in resolution, then a direct comparison of estimated burned areas to local estimates is the most important validation metric.

In this regard, it was also surprising to see the computation of cross-correlation with ground-based BA estimates instead of a direct computation of BA differences between remote sensing and local estimates.

In summary, I believe that the goals and contribution of this work should be more clearly defined and that the validation of the achieved results should be revised according to those goals. In any case, the current validation scheme is insufficient. I would at least expect a direct comparison of the proposed dataset against either a gold standard or the best possible proxy to the ground truth, together with a clearer comparison against other existing datasets and ground-based BA estimates.

Below are some other minor comments:

L284-290: The methodology followed to produce pre- and post-fire mosaics (especially pre-fire mosaics) is unclear to me. Please improve description.

L291: What are unburned polygons and how were they computed?

L291-292: “burned and unburned polygons were defined incrementally over both temporal composites until the desired BA accuracy was achieved.” This methodology is very unclear. Please improve the description of how you delineate those polygons and what the “desired delineation accuracy” is.

Figure 2 lacks detail on the generation of the BA product (phase 1).

L300 mentions a 50% threshold. Which variable is that threshold applied to?

Fig. 12 should include BA estimates from other independent datasets against which NEALGEBA is being compared (GABAM, FireCCI51, EFFIS, etc).

Author Response

Please see the attachment.

Author Response File: Author Response.docx