عوامل فیزیوگرافی و اقلیمی مؤثر بر الگوی مکانی آتش‌سوزی در جنگل‌های تالش با استفاده از مدل‌های آماری FR و WOE

Background and Objectives: Forest fire is one of the most significant natural–human phenomena that imposes wide-ranging environmental, economic, and social impacts. Over recent decades, the frequency and intensity of wildfires have increased due to climate change, rising temperatures, declining humidity, land-use changes, and expanding human activities. The northern forests of Iran, particularly the Talesh forests as part of the Hyrcanian ecosystem, despite their humid climatic conditions, are increasingly exposed to fire risks as a result of climatic fluctuations and anthropogenic pressures. Owing to their diverse topography, variable slopes, altitudinal gradients, and microclimatic variations, these forests represent an ideal setting to investigate the influence of physiographic and climatic factors on the spatial pattern of fire occurrence. Therefore, the present study aimed to identify and analyze the physiographic and climatic variables affecting the occurrence and spatial distribution of forest fires in the Talesh region and to develop a spatial–statistical framework to support preventive fire management. To achieve this, Frequency Ratio (FR) and Weights of Evidence (WOE) models were employed in a Geographic Information System (GIS) environment to assess the role of slope, aspect, elevation, temperature, and precipitation in shaping the spatial dynamics of wildfires.

Methodology:The study area is located in Talesh County, Gilan Province, within a mountainous landscape characterized by a humid to semi-humid climate, between latitudes 37°80′N and longitudes 48°90′E. Spatial datasets, including the Digital Elevation Model (DEM), slope and aspect layers, Normalized Difference Vegetation Index (NDVI) derived from Landsat imagery, and vector data of roads, rivers, and settlements, were collected and processed. Climatic parameters such as annual precipitation, temperature, and wind speed were obtained from local meteorological stations and the WorldClim global database. All datasets were converted to the same coordinate system (UTM) and resampled to a 30-meter spatial resolution for accurate spatial analysis. Physiographic and climatic variables were classified into meaningful categories using the Natural Breaks (Jenks) method. The FR model was applied to calculate the ratio between the percentage of fire occurrences and the corresponding class area, where values greater than 1 indicate a positive effect and values below 1 a negative effect. The WOE model was used to compute the logarithmic relationship between fire occurrence probability and background probability for each class; positive WOE values indicate a positive correlation, and negative values a negative effect. To correct small-sample bias, an adjustment term (ε = 10⁻⁶) was applied. Subsequently, FR and WOE values were extracted for slope, elevation, aspect, temperature, precipitation, and wind speed. The correlation matrix and Variance Inflation Factor (VIF) were computed to identify and eliminate multicollinearity. Finally, all weights were integrated in the GIS environment to produce fire susceptibility maps, enabling the identification of high-risk zones.

Results:The results indicated that the spatial distribution of forest fires in Talesh is governed by a complex interplay between physiographic and climatic variables. Slope was identified as the most influential factor, with the highest fire frequency occurring on moderate to steep slopes (8–20 degrees). These slopes are more prone to fires due to stronger natural ventilation, higher fuel accumulation, and restricted accessibility for suppression. Gentle slopes (<5°) and very steep slopes (>25°) had FR values less than 1, reflecting their lower susceptibility. Regarding elevation, the mid-altitudinal range (1000–2000 m a.s.l.) showed the highest fire frequency, corresponding to the dense Hyrcanian forests with moderate temperature and abundant dry fuels. Lower elevations (<500 m) exhibited less fire due to land-use change, while higher elevations (>2000 m) were less prone to fire because of increased humidity and lower temperatures. The temperature variable revealed that classes between 15°C and 21°C exhibited the strongest correlation with fire occurrence, forming the so-called “climatic fire window,” where surface fuels reach optimal combustibility. Lower (<15°C) and higher (>21°C) temperatures reduced fire probability due to excessive moisture or fuel limitation. The precipitation factor showed a dual effect: the highest fire risk occurred in the 1000–1200 mm range (FR>1 and positive WOE), where vegetation growth is abundant but becomes highly flammable during dry seasons. Conversely, areas with rainfall below 800 mm or above 1200 mm demonstrated lower fire risk due to limited vegetation or excessive humidity. Analysis of wind speed indicated that values between 8.5–17 m/s had the greatest impact on fire spread, while extremely weak or strong winds suppressed fire propagation. Spatial modeling results revealed that southern and southwestern slopes, especially at mid-elevations and with moderate-to-steep slopes, experienced the highest density of fire points, thus identified as critical fire-prone zones. The strong agreement between FR and WOE results confirmed the consistency and reliability of the combined modeling approach.

Conclusion: The spatial assessment of wildfires in the Talesh forests demonstrated that fire occurrence is not a random event but rather the outcome of intricate interactions among physiographic and climatic variables. Moderate-to-steep slopes, mid-altitudinal ranges, moderate temperatures, and intermediate rainfall constitute the ecological conditions most conducive to wildfire occurrence. The integrated use of FR and WOE models proved effective for identifying high-risk areas and quantifying the contribution of each variable to fire susceptibility. From a management perspective, the recognition of such critical zones provides a foundation for developing fire susceptibility maps and prioritizing preventive measures. It is recommended that preventive strategies including continuous remote-sensing monitoring, community training, establishment of firebreaks, and development of climate-based early warning systems be implemented in these high-risk regions. Furthermore, integrating FR and WOE models with advanced machine-learning algorithms such as Random Forest and XGBoost can enhance prediction accuracy in future studies. Overall, the present research not only provides an accurate spatial understanding of wildfire behavior in the Talesh forests but also proposes a localized model applicable to other northern Iranian forest ecosystems. The findings emphasize the importance of data-driven ecological management and informed decision-making for sustainable natural resource governance. Understanding the relationship between physiography and fire occurrence is essential not only for advancing scientific knowledge but also for ensuring the ecological and economic resilience of Iran’s forest landscapes.