Cyclic Trends of Wildfires over Sub-Saharan Africa

Round 1

Reviewer 1 Report

This article is very interesting and very well structured, showing important findings that complement the research on the topic under analysis. Therefore, the results obtained are relevant for the scientific community.

Besides the important theoretical framework on which this work is cast, based on extensive references (although excessively based on regional African references) but with several updated references, the results are well supported by the data analysis.

The text is clear and, properly drafted. However, I suggest the manuscript to be revised by  a native English (seems to have some minor problems). Also some minor corrections (see the attached document). The abstract seems to be too long. Consider reduce it.

The figures and tables seem to be appropriate (attention with the formatting) .

I consider the contents of this paper to be of good quality. It is very clear in terms of the methodology employed, which seems to be appropriate, and reveals a clear and logical structure.

Comments for author File: Comments.pdf

Author Response

Thank you very much for your review of our manuscript. Please find attached our point by point response to the points your raised.

Yours sincerely,

Reason Machete

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors present an interesting study of the occurrence of wildfires in sub-Saharan Africa, showing a decreasing trend in occurrence over recent decades. It would have enhanced the paper if some discussion of the possible reasons for such a finding were discussed. The authors touch on climate cycles (fewer lightening strikes?) and indicate that the drop in wildfire occurrences is primarily in the savannah. Does this discount crop land fires which may be set by farmers? This latter did not show any trend which could be expected if farming is maintained at the same level regardless of rainfall.

The authors should review the manuscript for conformance to journal requirements. On line 78, the reference to Roy at al should be assigned a numeric reference number, while on line 77 "Dintwe et al" should be inserted prior to "[37]". Similarly, on line192, "Andela et al" should be inserted prior to "[38]"; on line 210, "Jiang et al" should be inserted prior to "[39]"; on line 216, "Farkonedehmaal and Ghaffarzadegan" should be inserted before "[40]"; and, on line 223, "van Wilgen et al" should be inserted before "[43]".

There are a few typos: line 148 (approaches), line 175 (autocorrelation), line 191 (continental), and line 237 (oscillations).

Section 2.2 and the introductory portion of section 3 should be deleted: the relevance to the paper is not clear and this text distracts from the message of the manuscript. In figure 3, delete the last two zeros of the latitude and longitude as these are not relevant.

The paper presents an interesting observation of the occurrence of wildfires in southern Africa and is worthy of publication.

Author Response

Thank you very much for your review of our manuscript. We have tried our best to respond positively to your comments and hope you now find our manuscript suitable  for publication. Please find attached our point by point response to your comments.

Yours sincerely,

Reason Machete

Author Response File: Author Response.pdf

Reviewer 3 Report

The manuscript titled “CYCLIC TRENDS OF WILDFIRES OVER SUB-SAHARAN AFRICA” reports the results of an analysis of wildfire trends in sub-Saharan Africa. The authors concluded that the there was a declining trend in wildfires.

The paper lacks complete methods. There are not citations in the methodology. The authors do not provide a context for their chosen methodology and use for wildfire trend detection. The authors are inconsistent across the document about what they studied. I would suggest that the authors include direct statements about their questions, what their independent and depend art variables are, etc. It is unclear on what their analysis unit is. The authors do put their usage of “trend” into wildfire context nor do they clearly define it. 

The paper has the potential to improve the field, but currently is not complete or clear.

Some specific comments follow:

1) Lines 8-9:

Reviewer: How are declining trends of wildfires driven by “savannas, shrublands, and grasslands”? What does this mean?

2) Lines 45-59:

Reviewer: This paragraph could be reworded/edited/restructured to improve clarity.

3) Lines 61-63: “In this study, we assessed fire frequencies and cycles over a period of two decades. We hypothesize that fire frequency and return interval would change as climate change continues to take toll.”

Reviewer: The authors statement of what they assessed is not consistent with the rest of the manuscript. This hypothesis feels contrived and not particularly informative.

4) Lines 65-66: “The study was conducted in savannas and grasslands of sub-Saharan Africa (SSA) (Figure 1).”

Reviewer: In the abstract, results, and discussion, the authors speak about “savanna, grasslands, shrublands, croplands, and forests.” Was the study not conducted for these other vegetation structure types? 

5) Regression and Spectral Analysis section

Reviewer: The authors do not place these methods into context for use with fire trend analysis. Furthermore, there is no citations in this section.

6) Simulation example section

Reviewer: Again, there are no citations in this section and the authors have not placed their methods or this example into any wildfire framework. 

7) Lines 109:

Reviewer: No methods described for the results that include proportion.

8) Lines 112-113: “These were all significant at a level of 5 %. In fact, the largest P-Value was for South of the Equator, and it was 2.3 %”

Reviewer: Do the authors mean an alpha level of 0.05? 

9) Line 145:

Reviewer: These other vegetation types are not included in the final paragraph of the introduction.

10) Lines 153-154: “The Forests yielded a trend of 0.0018 at a P-Value of 0.023 whilst nonlinear regression of the detrended series and power spectrum gave periods of 5.26 and 5.99 years respectively.”

Reviewer: Here and elsewhere can you relate the values for these trends more directly to the fire frequency (periods?)? This will help the reader. What does a value of 0.0018 for the trend mean?

11) Lines 159-160: “Graphs of corresponding autocorrelation function and power spectrum gave periods of 2, 8, and 2.3 years”

Reviewer: I do not think that the methods adequately described how the “corresponding autocorrelation function and power spectrum gave periods of 2, 8, and 2.3 years.” Please expand.

12) Line 190: “[…] we considered the proportion of burned area […]”

Reviewer: Proportion is not introduced until the results. There is no mention of proportion in the abstract, introduction, or methods.

13) Lines 200-201: “[…] we found a significant declining trend of the proportion of burned area […]”

Reviewer: Trend of proportion of burned area? Please explain this.

Author Response

Thank you very much for your review of our manuscript. It has helped us a great deal to improve it. We have tried our best to respond positively to your comments and hope you shall find our manuscript now suitable for publication. Please find attached our point by point response to your comments.

Yours sincerely,

Reason Machete

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

The manuscript titled “CYCLIC TRENDS OF WILDFIRES OVER SUB-SAHARAN AFRICA” reports the results of an analysis of wildfire trends in sub-Saharan Africa. The authors conclude that the there is a declining trend in wildfires.

In general the new content in this revised version seems to address the letter but not the spirit of the requested revisions I previously submitted. They seem incomplete and do not read as a “final version”.

The paper lacks complete methods. What methods the authors did add in this revision are not clear and frankly read like a draft. The authors still do not adequately provide a context for their chosen methodology and use for wildfire trend detection. Moreover, the authors’ now state their method is novel. Yet, they do not provide the level of detail needed to understand why it is novel. Help the reader out. If it is novel, great!, but put this into understandable context. 

I am still unsure to what the analysis unit is. The new methodological content about the proportion of burnt area is not clear and seems rushed. The authors do not put their usage of “trend” or “cycle” into wildfire context nor do they clearly define either term. 

The paper has the potential to improve the field, but currently is not complete or clear.

1) Lines 8-9:

I still have an issue with this sentence. There is no quantity in “[…] seemingly largely driven by the savannas, shrublands and grasslands.” What is driving the trend? E.g., What is about savannas?

2) Lines 45-59: 

This paragraph still could be restructured to improve clarity.

3) Lines 61-62: “In this study, we assessed African fire proportions and their cycles over a period of two decades.”

Proportions of what?

4) Line 63:

Is “the corresponding return interval (period)” the same as “cycle” used in the previous sentence? If so, please be consistent.

5) Lines 67-68: “The study was conducted in savannas and grasslands of sub-Saharan Africa (SSA) (Figure 1).” 

Previously I said “In the abstract, results, and discussion, the authors speak about “savanna, grasslands, shrublands, croplands, and forests.” Was the study not conducted for these other vegetation structure types?” The authors’ responded: “We have made the correction; our study covers savannas, grasslands, shrublands, croplands, and forests”.

The sentence at Lines at 67-68 has no edits to it in the revised version. Please explain how the correction was made?

6) Lines 86-87: “trend and oscillations” how does these terms link with the content in the last paragraph on the introduction?

7) Lines 84-89: The authors’ present:

Total Area = A

ap = pixel area

Ab = burned area

They then state that “the total number of pixels in the area under consideration is n = A/ap”.

I have at least two questions. 1) Why does the value of A differ from ap?; 2) If A does not equal ap, How does the quotient of A and ap become the number of pixels? Both A and ap are areas, right? 

8) Lines 84-89:

The authors’ added this section (2.1 Proportion of Burnt Area) in response to my question 7 of my first review. This section does not seem complete. For example, see my above comment.  The new content does not clearly describe this seemingly important part of the work. 

9) Lines 91-99:

The new content provided here read more like draft content that final content in terms of clarity. The authors should spend more time describing some of the fundamental components of this analysis and how it connects with fire. The authors state that this is a novel application of the analytical method. Given that declaration, more information describing the method is warranted.

10) Methods in general. It still is not clear what the difference between ‘cycle’ and ‘linear trend’ is.

11) Line after 102: “As with all natural signals, there is observational noise in fire data”

This is an incomplete sentence.

12) Lines after 102: “This has been documented in [42] and [43] who discuss methods for assessing noise in wildfire data”

This is not sufficient detail for the reader to understand how these citations align with what they propose. Also, citation 42 is for a whole dissertation which does not quite direct the reader to an easily findable connection between the citation and the work presented.

13) Lines after 102: “It is therefore, important to investigate how the foregoing tools can perform in the presence of noise.”

The two sentences before this one do not provide information needed to meaningfully state “it is therefore, important”.

14) Lines 123: “The number of burns is a simple count of the pixels in which burns were detected.”

Are the pixels independent? How is each pixel counted as an independent fire? 

15) Lines 124-125: “In the subsequent analyses, we also study the proportion of burnt area, computing both the gradient of the trend line and the period of the underlying oscillations”

Where are the methods for performing these analyses with proportions of burnt area?

16) Figure 2 caption. The authors state " Graphs to highlight the efficacy of power spectrum and auto-correlation to detect underlying frequencies.”

How do they highlight the efficacy? The authors do not provide details here or in the methods on what this efficacy is. In the paragraph after Line 102 they state “A simulation example will be used to illustrate the impact of white noise on the efficacy of the power spectrum and the auto-correlation function in detecting underlying signal frequencies.” The authors do not define “efficacy” or how it is evaluated.

17) The figure legends might be more helpful if they referenced the fires in a clearer manner.

18) Figure 3.

Did the authors treat time (years) as a continuous variable? If so, why are their separate lines per year. If not, why is the scale for years a continuous color gradient?

19) Line 180

Is this referring to “proportion of burnt area”?

20) Lines 197-98: “These trends correspond to decreases of 17,500 km2, 1,400 km2 and 6,300 km2 per annum for Savannas, Grasslands and Croplands respectively”

Decreases of what? The areal extent of these vegetation types?

21) Figure 4 caption

What do the lines mean? Why is one dotted? What are the plus signs? Why do the panels use different colors?

Author Response

Please se the attachment in which we provide a point by response to the reviewer's comments.

Author Response File: Author Response.pdf