.png){kind=icon}
.png){kind=icon}
Return
Burn probability and wildfire risk modeling of the protected areas in Golestan Province, NE Iran
.png){kind=icon}
OA
AI
S
R
M
M
DOI：10.1186/s40677-026-00379-3.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Wildfires are one of the most significant natural hazards affecting ecosystems worldwide. They not only cause severe ecological damage but also pose substantial threats to human life and property. Therefore, modeling the spatial distribution of fire likelihood and potential fire behavior through Burn Probability (BP) and wildfire risk models is essential for effective planning, mitigation, and climate adaptation strategies. In recent decades, the Hyrcanian temperate forest region in northern Iran, particularly its protected areas, has been significantly affected by wildfires. This research aims to produce high-resolution and accurate BP maps, along with fire risk assessments for six protected areas in Golestan Province, northeastern Iran. The Minimum Travel Time (MTT) fire growth algorithm was implemented in FlamMap to estimate spatial BPs, utilizing customized fuel models based on the study area and considering climatic, physiographic, vegetation, and anthropogenic variables. Support Vector Regression (SVR), employing the ε-SVM approach, was used to predict fire risk rates. The accuracy of the classification maps was evaluated using historical fire ignition data. The fire risk prediction maps were categorized into four risk classes. The study analyzed the impact of various factors on BP rates across the different sites. Results indicated significant variations in BP under different conditions and across different protected areas, ranging from near zero in high-elevation zones at all sites to a maximum of 0.096 in the Golestan National Park (GNP). The highest BPs were associated with lower elevations (< 500 m) and fine fuel models (FM1, FM3, and FM4). Notably, regions such as GNP, Loveh, and both Upper-Zav and Lower-Zav showed a higher concentration of fire occurrences in high and extreme-risk zones. The models achieved high overall accuracies, ranging from 88–95%, with areas classified as high and very-high risk accounting for 54–72% of the observed fire ignitions across the study regions. The findings suggest that support vector regression (SVR) methods, when combined with region-specific calibration and spatial analysis, offer a powerful and transferable approach for predicting wildfire risk.
Keywords:
Burn probability
Hyrcanian forest
Protected areas
Support vector regression (SVR)
Wildfire risk map
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
4
Papers:
264
Citations:
1.1K
Organization
Cited Papers