Wildfire Delineation

Wildfires, also referred to as forest fires or bushfires, are uncontrolled blazes that takes place in wilderness or rural areas. They indirectly signal the state of ecological balance and can be extremely destructive, posing threats to wildlife, natural resources, residential areas, and human lives. Early detection and monitoring of wildfires facilitate an effective response. Delineating wildfires assists in risk assessment, evacuation planning, damage assessment, ecological impact analysis, and the allocation of resources. Additionally, this model can be used to segment the lava flow of volcanoes.

Using the model
Follow the guide to use the model. Before using this model, ensure that the supported deep learning libraries are installed. For more details, check Deep Learning Libraries Installer for ArcGIS.

Fine-tuning the model
This model cannot be fine-tuned using ArcGIS tools.

Input
Sentinel-2 L2A 12-bands multispectral imagery using Bottom of Atmosphere (BOA) reflectance product in the form of a raster, mosaic or image service. 

Output
Classified raster with two classes: Active fire and other.

Applicable geographies
The model is expected to work well globally.

Model architecture
It uses Normalized Burn Ratio (NBR) to delineate fire.

Limitations
This model exhibits limited effectiveness in delineation of fires within urban environments. The presence of bright regions or pixels, or the detection of a single pixel, may lead to instances of false detection. Thus it is recommended for use in mainly wild forest areas, with a threshold setting of 0.3 to 0.5, and with cluster detections.

Sample results
Here are a few results from the model.