Wildfire Risk Forecasting Using Weights of Evidence and Statistical Index Models
Round 1
Reviewer 1 Report
Manuscript entitled ~Forest Fire Risk Forecasting Using Weights of Evidence and Statistical Index Models~ it is interesting and evaluated a problem about fire risk potential using WoE SI models.
Here are some comments:
1.Introduction - to add a paragraph on the methods applied and in what context.
2.Materials and Methods - Figure 1 - I suggest changing the color palette for elevation with terrain palette color.
3.Dataset and Preprocessing - line 98 -the authors will talk about the classification of layers according to expert opinion. Can you explain in more detail?
4. Figure 2 - I suggest that the authors remove the fire points from the thematic maps. Those points are relevant in evaluating the final results.
5. Discussion - from line 183 to 234 the authors analyze other studies, the factors that amplify the occurrence of fires and the applied models. I suggest that this information be passed on, rather than the introduction. This section discusses the results obtained in this study.
6. Conclusions: rewrite some phrases for a better explanation of conclusion according of your results.
7. The authors correct the way of writing the equations.
Author Response
Dear Reviewer,
We thank you very much for your time and constructive comments that help us to improve our manuscript. We think we have addressed all your comments. We highlighted the changes in the manuscript as well. Please let us know if you have any further comments.
Manuscript entitled “Forest Fire Risk Forecasting Using Weights of Evidence and Statistical Index Models” it is interesting and evaluated a problem about fire risk potential using WoE SI models.
Here are some comments:
- Introduction - to add a paragraph on the methods applied and in what context.
Response. We added a paragraph toward the end of the Introduction:
“In Iran, wildfire is seen as a significant threat to forests and pastures. Some estimates suggest that an average of 400 fire events occur per year that burn over 6000 hectares of land [57]. This trend is expected to continue and may even increase in the future due to ongoing climate and land-use changes, as well as increasing human activities [58]. On the other hand, and based on many conducted surveys, the application of the WoE and SI models in forest fire susceptibility mapping is novel. Therefore, this research was conducted to evaluate the WoE and SI models for wildfire susceptibility mapping in Sanandaj county. In other words, the aim of this study was to identify high-risk and low-risk areas of wildfire in the study area using the WoE and SI models to manage and reduce the risk of fire. The maps of factors affecting the fire including altitude, slope percentage, slope direction, distance from the road and distance from the river, Land Use/Land Cover (LULC), average annual rainfall, and average annual temperature were prepared using the available statistical data (2011-2020) and Geographic Information System (GIS). Then with the help of mathematical functions (WoE and SI models), the fire potential map of the area was forecasted and finally, the two models were compared using the ROC curves.”
- Materials and Methods - Figure 1 - I suggest changing the color palette for elevation with terrain palette color.
Response. We have changed the color palette with terrain palette color.
- Dataset and Preprocessing - line 98 -the authors will talk about the classification of layers according to expert opinion. Can you explain in more detail?
Response. This part is related to the classification of slope, direction, altitude, and other factors of environmental variables and is completely dependent on different variables and is not often mentioned in the literature. Therefore, we decided to remove lines 97 and 98.
- Figure 2 - I suggest that the authors remove the fire points from the thematic maps. Those points are relevant in evaluating the final results.
Response. Thank you for your suggestion. However, each of those maps provides useful information to researchers on their own, and some may be interested to look at the individual factors triggering wildfire and may be interested to see the conditions around the fire points. Also, the thematic maps are based on fire points, and the omission of these points creates ambiguity for the audience as to what and how these maps were made. In fact, fire points are an important part of these plans.
- Discussion - from line 183 to 234 the authors analyze other studies, the factors that amplify the occurrence of fires and the applied models. I suggest that this information be passed on, rather than the introduction. This section discusses the results obtained in this study.
Response. Thank you for your suggestion. However, in this study, the effect of 8 factors on the occurrence of wildfires has been investigated. The weight and importance of the classes of each factor in the occurrence of wildfires using two methods of weights of evidence and statistical index (lines 183 to 234) have been investigated in this section. Therefore, this section is part of the model results and the weighted results obtained in this study have been compared with other studies.
- Conclusions: rewrite some phrases for a better explanation of conclusion according of your results.
Response. We have rewritten some parts of the Conclusion and added a few more sentences:
“… According to the wildfire risk forecasting map, the highest number and density of forest fires occurred in the lands and forest parks around the Sanandaj city, where the human factor has the most role in these fires. Therefore, in order to prevent and deal with potential hazards due to fires in the areas of natural resources and environment of Sanandaj city, measures should be taken to increase monitoring in areas with high and very high fire risk potential, such as increasing the number of human resources, creating firebreaks, and allocating most of the financial resources related to forest and rangeland firefighting, i.e., by launching an appropriate fire extinguishing system, using modern technologies and appropriate equipment, as well as training expert forces. Other strategies to prevent fires in the fields of natural resources and the environment include culture building on the importance and use of these national resources….”
- The authors correct the way of writing the equations.
Response. The equations have been checked for their correctness, and the clearest and most explanatory way of representing them is used to help readers understand them easier.
We hope that the changes mentioned above are satisfactory.
Respectfully yours,
Dr. Ebrahim Ghaderpour
On behalf of all the authors
Author Response File: Author Response.pdf
Reviewer 2 Report
General Comments:
It's a repetitive manuscript from the Zagros region of Iran, without any new findings and comprehensive analysis (for example comparing many models). As mentioned in the paper, you considered wildfire in all LULC classes (not only Forest). So, change the title from a forest fire to wildfire and do it in the rest of the manuscript.
Specific Comments:
** Line 15: Due to the similar accuracy of both models, a better model is a model with fewer input parameters. Revise it.
** Keywords: Add a geographical location of the study area Sanandaj and Kurdistan.
** Line 39: Just focus on the study area, I mean Zagross forest and use citation of this region, about the change of rainfall pattern, for example:
Attarod, P., Rostami, F., Dolatshahi, A., Sadeghi, S. M. M., Amiri, G. Z., & Bayramzadeh, V. (2016). Do changes in meteorological parameters and evapotranspiration affect declining oak forests of Iran?. Journal of Forest Science, 62(12), 553-561.
Attarod, P., Sadeghi, S. M. M., Pypker, T. G., & Bayramzadeh, V. (2017). Oak trees decline; a sign of climate variability impacts in the west of Iran. Caspian Journal of Environmental Sciences, 15(4), 373-384.
** Line 57: Here, need to show what are the disadvantages of other models and what are the advantages of Woe and SI models?
** Line 58: "After agricultural and urban activities, forest fires are the second most common cause of ecosystem degradation" Where? globally?
** Line 60: Forests and Rangelands Organization of Iran...There is no organization with this name, revise it.
** Line 60: ... 5,000 to 6,000 hectares of forest area are reduced annually. Where? In Iran or in the Zagros region?
** Line 66: Specify what is the necessity of research? This study attempted to fill which kind of research gap?
** At the end of the introduction, I did not see any literature reviews of Forest fire studies in the Zagros region (at least 10 papers, to my knowledge, have been published). I would like to say the introduction is not well, framed, and organized, and the authors need to check for more recent references.
Author Response
Dear Reviewer,
We thank you very much for your time and constructive comments that help us to improve our manuscript. We think we have addressed all your comments. We highlighted the changes in the manuscript as well. Please let us know if you have any further comments.
General Comments:
It's a repetitive manuscript from the Zagros region of Iran, without any new findings and comprehensive analysis (for example comparing many models). As mentioned in the paper, you considered wildfire in all LULC classes (not only Forest). So, change the title from a forest fire to wildfire and do it in the rest of the manuscript.
Response. Using the same model in different regions has different results because the influencing factors vary in different places. This study was carried out in a part of North Zagros using WOE and SI models, and the map of high-risk and low-risk areas of the fire was identified where these models had not been used before, so both models and results for this region are new. Furthermore, it is practically almost impossible to compare many models at the same time for one region because it requires a lot of data, time, and money. On the other hand, the possibility of error increases. Therefore, most articles compare between 2 or 3 models.
We have changed the title to Wildfire and did this in several places in the manuscript. We also defined the link between wildfire and forest fire.
Specific Comments:
** Line 15: Due to the similar accuracy of both models, a better model is a model with fewer input parameters. Revise it.
Response. The input parameters were the same for both models. In this study, the performance of the models was investigated using the Receiver Operating Characteristic (ROC) curve method. The performance of the models was relatively similar, but the performance of the weights of evidence model was slightly better than the performance of the statistical index model. This was probably due to the type and number of parameters considered as well as the condition of the study area.
We added the following sentence in the Abstract:
“Although the input parameters for both models were the same, the WoE model showed a slightly higher AUC value compared to the SI model and potentially can be used for predicting future fire risk in the study area”
** Keywords: Add a geographical location of the study area Sanandaj and Kurdistan.
Response. We added Kurdistan, Sanandaj county, and Zagros Forests in the keywords.
** Line 39: Just focus on the study area, I mean Zagros forest and use citation of this region, about the change of rainfall pattern, for example:
Attarod, P., Rostami, F., Dolatshahi, A., Sadeghi, S. M. M., Amiri, G. Z., & Bayramzadeh, V. (2016). Do changes in meteorological parameters and evapotranspiration affect declining oak forests of Iran?. Journal of Forest Science, 62(12), 553-561.
Attarod, P., Sadeghi, S. M. M., Pypker, T. G., & Bayramzadeh, V. (2017). Oak trees decline; a sign of climate variability impacts in the west of Iran. Caspian Journal of Environmental Sciences, 15(4), 373-384.
Response. We have significantly improved the Introduction and added your two suggested articles [34,35]:
“Herein, the study area is the part of the Zagros forests of western Iran dominated by oak trees. In recent years, the area of Zagros forests has decreased mainly due to climate change. The effect of climate change on reducing the area of Zagros forests has been confirmed in various studies [34,35]. Furthermore, climate change and the rise of mean annual temperature have increased the fire rate in forests and pastures. It seems that the fire affects the vegetation and changes the forest stand structures. The results of a study in [36] showed that a decade after fire occurrence, the share of oak trees has been decreased, while the proportion of Amygdalus and Crataegus has been increased.”
** Line 57: Here, need to show what are the disadvantages of other models and what are the advantages of Woe and SI models?
Response. We added the following sentences:
“…Each method has its own advantages and weaknesses.
Zagros forest fires have been studied by many researchers. In research conducted in the Sardasht forests, it was shown that wildfire has a higher chance to occur between June and September [53]. In another study, fire danger maps were created using SVM, GLM, FDA, and random forest. The results of that study showed that FDA (0.777) and GLM (0.772) algorithms generated the most accurate fire danger maps [54]. Jaafari et al. [55] modeled wildfire probability across the Zagros mountains of Iran using the WoE model. Findings of that study clearly demonstrated that the probability of a fire is strongly dependent on the topographic characteristics of landscapes and, perhaps more importantly, human infrastructure and associated human activities. The main advantage of the WoE and SI models is that they calculate the weighted value of the factors based on a statistical formula, and thus they avoid the subjective choice of weighting factors. In addition, input maps with missing data (incomplete coverage) can be accommodated in the models that do not significantly impact the results [40,56]. However, the main shortcoming of the WoE and SI models is that the weight values calculated for different areas are not comparable in terms of the degree of hazard [56].”
Please note that our aim in this study is not to examine the disadvantages and advantages of other models, but to evaluate the performance of the WOE and SI models to predict the risk of fire in the study region.
** Line 58: "After agricultural and urban activities, forest fires are the second most common cause of ecosystem degradation" Where? globally?
Response. “in the world”. We, however, removed this sentence as its reference was old. Instead, we added the following sentence:
“In 2015, approximately 98 million hectares of forest around the world were affected by fires [21].”
** Line 60: Forests and Rangelands Organization of Iran...There is no organization with this name, revise it.
Response. Corrected: “Natural Resources and Watershed Management Organization".
** Line 60: ... 5,000 to 6,000 hectares of forest area are reduced annually. Where? In Iran or in the Zagros region?
Response. We replaced it with
“Iran is one of the low-forest cover countries in the world. Therefore, investigation of fire consequences in the forests of Iran and recognition of opposition methods to fire in these forests is essential to present a solution to decrease these fires [32]. Reports indicate the occurrence of 1124 cases of fires and burning of 7364 hectares of forests and rangelands of Kurdistan province is only between the years 2007 and 2016 [33].”
** Line 66: Specify what is the necessity of research? This study attempted to fill which kind of research gap?
Response. We added a paragraph toward the end of the Introduction:
“In Iran, wildfire is seen as a significant threat to forests and pastures. Some estimates suggest that an average of 400 fire events occur per year that burn over 6000 hectares of land [57]. This trend is expected to continue and may even increase in the future due to ongoing climate and land-use changes, as well as increasing human activities [58]. On the other hand, and based on many conducted surveys, the application of the WoE and SI models in forest fire susceptibility mapping is novel. Therefore, this research was conducted to evaluate the WoE and SI models for wildfire susceptibility mapping in Sanandaj county. In other words, the aim of this study was to identify high-risk and low-risk areas of wildfire in the study area using the WoE and SI models to manage and reduce the risk of fire. The maps of factors affecting the fire including altitude, slope percentage, slope direction, distance from the road and distance from the river, Land Use/Land Cover (LULC), average annual rainfall, and average annual temperature were prepared using the available statistical data (2011-2020) and Geographic Information System (GIS). Then with the help of mathematical functions (WoE and SI models), the fire potential map of the area was forecasted and finally, the two models were compared using the ROC curves.”
** At the end of the introduction, I did not see any literature reviews of Forest fire studies in the Zagros region (at least 10 papers, to my knowledge, have been published). I would like to say the introduction is not well, framed, and organized, and the authors need to check for more recent references.
Response. We have added 20 recent articles and expanded the Introduction and talked more about Zagros region. All the changes are highlighted in the revised manuscript.
We hope that the changes mentioned above are satisfactory.
Respectfully yours,
Dr. Ebrahim Ghaderpour
On behalf of all the authors
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
I have seen the improvements made to the article and I agree with its publication in its current form.
Author Response
Dear Reviewer.
Thank you very much for your time and comprehensive comments.
Best regards,
Ebrahim Ghaderpour
Reviewer 2 Report
The revised version of the manuscript is generally in good shape, overall well-structured and well written.
Author Response
Dear Reviewer,
Thank you very much for your time and comprehensive comments.
Best regards,
Ebrahim Ghaderpour