Reconstructing pre-fire vegetation condition in the wildland urban interface (WUI) using artificial neural network

Area of Science:

• Remote Sensing
• Geospatial Analysis
• Wildfire Science

Background:

• Urban wildfires pose significant risks to life, property, and infrastructure.
• Accurate fire risk assessment is critical for managing wildlands, especially smaller urban areas.
• Pre-fire vegetation characteristics are vital fuel parameters but are difficult to obtain due to high temporal variation.

Purpose of the Study:

• To develop an Artificial Neural Network (ANN) model for reconstructing pre-fire vegetation spectral characteristics.
• To improve the accuracy of fire risk assessment in urban wildland interfaces by incorporating pre-fire fuel conditions.
• To enhance the spatial resolution of vegetation data for better fire management strategies.

Main Methods:

• An Artificial Neural Network (ANN) was designed to reconstruct spectral characteristics of vegetation from Landsat imagery.
• The method was tested on urban vegetation and applied to a 2016 wildfire site in Haifa, Israel.
• Reconstructed data was used to sharpen original Landsat data, generating high-resolution Normalized Difference Vegetation Index (NDVI) maps.

Main Results:

• The reconstructed RGB image accurately mapped green vegetation locations in urban areas.
• Normalized Difference Vegetation Index (NDVI) maps effectively determined vegetation presence and dryness levels.
• The method demonstrated potential for identifying fuel characteristics and improving fire risk assessment models.

Conclusions:

• The ANN-based spectral reconstruction method enhances the understanding of pre-fire vegetation conditions.
• This approach improves the spatial accuracy of vegetation mapping and dryness estimation in urban wildlands.
• The findings support the integration of reconstructed spectral data into fire-risk assessment and behavior modeling.