the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 12 Apr 2022
Model Comparisons for Predicting Grassland Fire Occurrence Probability in Inner Mongolia Autonomous Region, China
Abstract. Grassland fires threaten grassland ecosystem, human life and economic development greatly. However, there are limited researches focusing on grassland fire prediction, thereby it is necessary to find a better method to predict probability of grassland fire occurrence. Here we selected 16 environmental variables that may have impacts on fire occurrence, then built regression models of grassland fire probability based on historical fire points and variables in Inner Mongolia by three methods to identify the grassland fire drivers and to predict fire probability. The three methods were global logistic regression, geographically weighted logistic regression and random forest. The results showed that random forest model had the best predictive effect. The influence of 9 variables selected by geographically weighted logistic regression model on grassland fire was unbalanced spatially. The three models all showed that meteorological factors were of great importance to grassland fire occurrence. In Inner Mongolia, grassland fires occurred in different areas had various responses to the influencing drivers, and the areas with different natural and geographical condition had different fire prevention periods. Thus, the grassland fire management strategy based on local conditions should be advocated.
This preprint has been withdrawn.
-
Withdrawal notice
This preprint has been withdrawn.
-
Preprint
(786 KB)
Interactive discussion
Status: closed
-
RC1: 'Comment on nhess-2022-72', Anonymous Referee #1, 27 Jul 2022
This manuscript focused on grassland fire occurrence probability in Inner Mongolia Autonomous of China. Previous research mainly focused on fire occurrence probability in forest ecosystem in China, the fire occurrence probability research in China was very limited., so this manuscript will give some implications for grassland fire management in China.The results showed that geographically weighted logistic regression model and random forest model could well predict fire occurrence in this area. The scientific data, results and conclusions presented in a clear, concise, and well-structured, the only concern of this manuscript is the language need to polish by a native speaker. After carefully consideration, I think this manuscript could be published after minor revision.
Citation: https://doi.org/10.5194/nhess-2022-72-RC1 - AC1: 'Reply on RC1', Chang Chang, 27 Jul 2022
-
RC2: 'Comment on nhess-2022-72', Anonymous Referee #2, 29 Jul 2022
Thank you for your efforts. It is good to see the application of these models in the prediction of grassland fire occurrence, which will add new content to the body of knowledge on wildfire occurrence. This manuscript is good, but it could be better with a few changes - I list them below.
Comment 1ï¼
Figure 4 is not clear enough, the resolution needs to be further adjusted. Other figures in this paper also have similar problems.
Comment 2ï¼
In Table 4, it may be biased to only use MAE as the error evaluation criterion. In model accuracy evaluation, MAE (mean absolute error) and RMSE (root mean squared error) are both common indicators, but MAE can only reflect the average error value in a general form. RMSE is the default metric of many models and is more sensitive to outliers, it may be better to add RMSE as the index of model accuracy evaluation.
Comment 3ï¼
This study is reasonable to explain most of the mechanism of fire occurrence driving in the grassland area, but the explanation of this part is confusing: “We believed this was due to the fact that grasslands of Inner Mongolia were mainly distributed on the flat plateau lacking of steep slopes, resulting in a negative correlation between grassland fire occurrences and Slope”(Line 411-413). Can you explain it further? Or do you have more sufficient data to support this idea.
Citation: https://doi.org/10.5194/nhess-2022-72-RC2 -
AC2: 'Reply on RC2', Chang Chang, 01 Aug 2022
First, thanks for your suggestion and question. We are very pleased to reply your comments, and response them respectively.
Comment 1:
Figure 4 is not clear enough, the resolution needs to be further adjusted. Other figures in this paper also have similar problems.
Response:
We are very sorry to make trouble for you. We provide the pictures in manuscript with higher resolution in the supplement, hoping they can be clear enough.
Comment 2:
In Table 4, it may be biased to only use MAE as the error evaluation criterion. In model accuracy evaluation, MAE (mean absolute error) and RMSE (root mean squared error) are both common indicators, but MAE can only reflect the average error value in a general form. RMSE is the default metric of many models and is more sensitive to outliers, it may be better to add RMSE as the index of model accuracy evaluation.
Response:
Thanks for your valuable advice, we had calculated the RMSE values of the 3 models. We will add RMSE value in the Table 4 in the revised manuscript.
Comment 3:
This study is reasonable to explain most of the mechanism of fire occurrence driving in the grassland area, but the explanation of this part is confusing: “We believed this was due to the fact that grasslands of Inner Mongolia were mainly distributed on the flat plateau lacking of steep slopes, resulting in a negative correlation between grassland fire occurrences and Slope”(Line 411-413). Can you explain it further? Or do you have more sufficient data to support this idea.
Response:
We are very sorry to make confusion for you. In our opinion, the grasslands of Inner Mongolia were mainly distributed on the flat plateau, thus the more grassland fires occurred on the flat plateau where the terrain was flat and lacks steep slopes. Therefore, we can deduce that grassland fires prefer to occur on the flat ground where there is limited slope. Then, we get the result that a negative correlation is between grassland fire occurrences and Slope.
-
AC2: 'Reply on RC2', Chang Chang, 01 Aug 2022
-
RC3: 'Comment on nhess-2022-72', Anonymous Referee #3, 22 Aug 2022
Review
Model Comparisons for Predicting Grassland Fire Occurrence Probability in Inner Mongolia Autonomous Region, China
The author team provided an interesting manuscript and analysis on grassland fires in Inner Mongolia and developed an approach to compare three statistical model for predicting the probability of grassland fire occurrence. They used historical datasets of grassland fires and additional 16 different variables related to grassland fires for their model.
Considering the research transparency of the study I recommend major revision of the introduction and the methods section. Some of the information I missed in the introduction and method section were addressed in the results, however, I assume that the authors are aware of the information but did not made it explicit for the reader in the manuscript. I provided detailed comments and questions in the attached pdf and I am convinced that these remarks make it clearer. The current description of some parts of the methods did not allow an assessment.
Furthermore, I missed in the discussion a critical view on the chosen methods, data and resolutions, as well as their possible influence on the gained results. Such a reflection would allow also readers from other areas to learn from your study. Thus, I highly recommend to stress clearly in the manuscript what is the new and innovative part of your study.
Additional minor comments:
English editing – check abstract and a few indicated sentences in the manuscript.
- AC3: 'Reply on RC3', Chang Chang, 08 Sep 2022
Interactive discussion
Status: closed
-
RC1: 'Comment on nhess-2022-72', Anonymous Referee #1, 27 Jul 2022
This manuscript focused on grassland fire occurrence probability in Inner Mongolia Autonomous of China. Previous research mainly focused on fire occurrence probability in forest ecosystem in China, the fire occurrence probability research in China was very limited., so this manuscript will give some implications for grassland fire management in China.The results showed that geographically weighted logistic regression model and random forest model could well predict fire occurrence in this area. The scientific data, results and conclusions presented in a clear, concise, and well-structured, the only concern of this manuscript is the language need to polish by a native speaker. After carefully consideration, I think this manuscript could be published after minor revision.
Citation: https://doi.org/10.5194/nhess-2022-72-RC1 - AC1: 'Reply on RC1', Chang Chang, 27 Jul 2022
-
RC2: 'Comment on nhess-2022-72', Anonymous Referee #2, 29 Jul 2022
Thank you for your efforts. It is good to see the application of these models in the prediction of grassland fire occurrence, which will add new content to the body of knowledge on wildfire occurrence. This manuscript is good, but it could be better with a few changes - I list them below.
Comment 1ï¼
Figure 4 is not clear enough, the resolution needs to be further adjusted. Other figures in this paper also have similar problems.
Comment 2ï¼
In Table 4, it may be biased to only use MAE as the error evaluation criterion. In model accuracy evaluation, MAE (mean absolute error) and RMSE (root mean squared error) are both common indicators, but MAE can only reflect the average error value in a general form. RMSE is the default metric of many models and is more sensitive to outliers, it may be better to add RMSE as the index of model accuracy evaluation.
Comment 3ï¼
This study is reasonable to explain most of the mechanism of fire occurrence driving in the grassland area, but the explanation of this part is confusing: “We believed this was due to the fact that grasslands of Inner Mongolia were mainly distributed on the flat plateau lacking of steep slopes, resulting in a negative correlation between grassland fire occurrences and Slope”(Line 411-413). Can you explain it further? Or do you have more sufficient data to support this idea.
Citation: https://doi.org/10.5194/nhess-2022-72-RC2 -
AC2: 'Reply on RC2', Chang Chang, 01 Aug 2022
First, thanks for your suggestion and question. We are very pleased to reply your comments, and response them respectively.
Comment 1:
Figure 4 is not clear enough, the resolution needs to be further adjusted. Other figures in this paper also have similar problems.
Response:
We are very sorry to make trouble for you. We provide the pictures in manuscript with higher resolution in the supplement, hoping they can be clear enough.
Comment 2:
In Table 4, it may be biased to only use MAE as the error evaluation criterion. In model accuracy evaluation, MAE (mean absolute error) and RMSE (root mean squared error) are both common indicators, but MAE can only reflect the average error value in a general form. RMSE is the default metric of many models and is more sensitive to outliers, it may be better to add RMSE as the index of model accuracy evaluation.
Response:
Thanks for your valuable advice, we had calculated the RMSE values of the 3 models. We will add RMSE value in the Table 4 in the revised manuscript.
Comment 3:
This study is reasonable to explain most of the mechanism of fire occurrence driving in the grassland area, but the explanation of this part is confusing: “We believed this was due to the fact that grasslands of Inner Mongolia were mainly distributed on the flat plateau lacking of steep slopes, resulting in a negative correlation between grassland fire occurrences and Slope”(Line 411-413). Can you explain it further? Or do you have more sufficient data to support this idea.
Response:
We are very sorry to make confusion for you. In our opinion, the grasslands of Inner Mongolia were mainly distributed on the flat plateau, thus the more grassland fires occurred on the flat plateau where the terrain was flat and lacks steep slopes. Therefore, we can deduce that grassland fires prefer to occur on the flat ground where there is limited slope. Then, we get the result that a negative correlation is between grassland fire occurrences and Slope.
-
AC2: 'Reply on RC2', Chang Chang, 01 Aug 2022
-
RC3: 'Comment on nhess-2022-72', Anonymous Referee #3, 22 Aug 2022
Review
Model Comparisons for Predicting Grassland Fire Occurrence Probability in Inner Mongolia Autonomous Region, China
The author team provided an interesting manuscript and analysis on grassland fires in Inner Mongolia and developed an approach to compare three statistical model for predicting the probability of grassland fire occurrence. They used historical datasets of grassland fires and additional 16 different variables related to grassland fires for their model.
Considering the research transparency of the study I recommend major revision of the introduction and the methods section. Some of the information I missed in the introduction and method section were addressed in the results, however, I assume that the authors are aware of the information but did not made it explicit for the reader in the manuscript. I provided detailed comments and questions in the attached pdf and I am convinced that these remarks make it clearer. The current description of some parts of the methods did not allow an assessment.
Furthermore, I missed in the discussion a critical view on the chosen methods, data and resolutions, as well as their possible influence on the gained results. Such a reflection would allow also readers from other areas to learn from your study. Thus, I highly recommend to stress clearly in the manuscript what is the new and innovative part of your study.
Additional minor comments:
English editing – check abstract and a few indicated sentences in the manuscript.
- AC3: 'Reply on RC3', Chang Chang, 08 Sep 2022
Data sets
China monthly Vegetation index (NDVI) spatial distribution dataset Xinliang Xu https://doi.org/10.12078/2018060602
Model code and software
Windows Application for Geographically Weighted Regression Modelling Tomoki Nakaya https://download.csdn.net/download/xiaodongfly/7027693
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 684 | 290 | 38 | 1,012 | 36 | 30 |
- HTML: 684
- PDF: 290
- XML: 38
- Total: 1,012
- BibTeX: 36
- EndNote: 30
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1