Estimating wildfire ignition probabilities with geographic weighted logistic regression.

Area of Science:

• Environmental Science
• Geospatial Analysis
• Risk Management

Background:

• Wildfire ignition probabilities are crucial for effective wildfire management, risk assessment, and resource prepositioning.
• Northern Portugal, specifically the Baião municipality, experiences frequent wildfires during its fire season.
• Accurate ignition probability data can significantly aid firefighting authorities in identifying vulnerable areas and combating fire occurrences.

Purpose of the Study:

• To estimate fire ignition probabilities in the Baião municipality, Portugal.
• To develop quantitative models for wildfire risk management and resource allocation.
• To assist local authorities in identifying fire-prone zones for enhanced wildfire response.

Main Methods:

• Geographically Weighted Regression (GWLR) with an exponential kernel was employed to estimate ignition probabilities.
• Logit and probit link functions were utilized alongside independent variables: population density, distance to roads, altitude, forest proportion (land use), and Normalized Difference Moisture Index (NDMI) from LANDSAT 8.
• A binary dependent variable (wildfire ignition occurrence 2011-2020) was used, with data split into training (70%) and test sets via stratified sampling.

Main Results:

• The study generated useful application models for estimating wildfire ignition probabilities.
• Model performance was rigorously evaluated using metrics such as accuracy, ROC curve AUC, precision, recall, specificity, balanced accuracy, and F1 score.
• The developed models offer valuable insights for wildfire management in Portugal, complementing existing reference models.

Conclusions:

• The developed GWLR models provide a robust framework for predicting wildfire ignition probabilities.
• These models can be integrated into quantitative risk management strategies for fuel and resource management.
• The findings support enhanced decision-making for firefighting authorities in fire-prone regions like Baião.