Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Development and Application of Self-Supervised Machine Learning for Smoke Plume and Active Fire Identification from the FIREX-AQ Datasets

Nicholas LaHaye, Anistasija Easley, Kyongsik Yun, Huikyo Lee, Erik Linstead, Michael J. Garay, Olga V. Kalashnikova

TL;DR

The paper addresses the challenge of identifying active fires and smoke plumes across heterogeneous remote-sensing data without extensive labeled data. It introduces SIT-FUSE, a self-supervised framework that learns representations with flexible encoders and applies deep clustering to produce per-instrument and fused smoke and fire masks, validated through SSIM against hand-labeled regions and cross-instrument comparisons. The work demonstrates strong fire and smoke detection performance across multiple instruments and fusion scenarios, including MISR/MODIS, AirMSPI/eMAS, MASTER/eMAS, and PlanetScope; it also provides open-source software and data products. The approach promises to enhance operational wildfire monitoring and support climate studies by enabling high-resolution, multi-instrument plume tracking with reduced labeling effort.

Abstract

Fire Influence on Regional to Global Environments and Air Quality (FIREX-AQ) was a field campaign aimed at better understanding the impact of wildfires and agricultural fires on air quality and climate. The FIREX-AQ campaign took place in August 2019 and involved two aircraft and multiple coordinated satellite observations. This study applied and evaluated a self-supervised machine learning (ML) method for the active fire and smoke plume identification and tracking in the satellite and sub-orbital remote sensing datasets collected during the campaign. Our unique methodology combines remote sensing observations with different spatial and spectral resolutions. The demonstrated approach successfully differentiates fire pixels and smoke plumes from background imagery, enabling the generation of a per-instrument smoke and fire mask product, as well as smoke and fire masks created from the fusion of selected data from independent instruments. This ML approach has a potential to enhance operational wildfire monitoring systems and improve decision-making in air quality management through fast smoke plume identification12 and tracking and could improve climate impact studies through fusion data from independent instruments.

Development and Application of Self-Supervised Machine Learning for Smoke Plume and Active Fire Identification from the FIREX-AQ Datasets

TL;DR

The paper addresses the challenge of identifying active fires and smoke plumes across heterogeneous remote-sensing data without extensive labeled data. It introduces SIT-FUSE, a self-supervised framework that learns representations with flexible encoders and applies deep clustering to produce per-instrument and fused smoke and fire masks, validated through SSIM against hand-labeled regions and cross-instrument comparisons. The work demonstrates strong fire and smoke detection performance across multiple instruments and fusion scenarios, including MISR/MODIS, AirMSPI/eMAS, MASTER/eMAS, and PlanetScope; it also provides open-source software and data products. The approach promises to enhance operational wildfire monitoring and support climate studies by enabling high-resolution, multi-instrument plume tracking with reduced labeling effort.

Abstract

Fire Influence on Regional to Global Environments and Air Quality (FIREX-AQ) was a field campaign aimed at better understanding the impact of wildfires and agricultural fires on air quality and climate. The FIREX-AQ campaign took place in August 2019 and involved two aircraft and multiple coordinated satellite observations. This study applied and evaluated a self-supervised machine learning (ML) method for the active fire and smoke plume identification and tracking in the satellite and sub-orbital remote sensing datasets collected during the campaign. Our unique methodology combines remote sensing observations with different spatial and spectral resolutions. The demonstrated approach successfully differentiates fire pixels and smoke plumes from background imagery, enabling the generation of a per-instrument smoke and fire mask product, as well as smoke and fire masks created from the fusion of selected data from independent instruments. This ML approach has a potential to enhance operational wildfire monitoring systems and improve decision-making in air quality management through fast smoke plume identification12 and tracking and could improve climate impact studies through fusion data from independent instruments.
Paper Structure (18 sections, 18 figures, 8 tables)

This paper contains 18 sections, 18 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Materials and Methods
  4. Methods
  5. Data Preprocessing
  6. Self-Supervised Representation Learning
  7. Deep Clustering
  8. Hand Labeling
  9. Context Assignment
  10. Contouring and Filtering
  11. Validation
  12. Materials and Tools
  13. Results
  14. Fire Detection
  15. Smoke Detection
  16. ...and 3 more sections

Figures (18)

  • Figure S1: Reference Map and Imagery
  • Figure S2: A display of the smoke plume (left) and the fire front (right) progression of the Williams Flats fire on August 06, 2019, as captured by GOES-17 and segmented via SIT-FUSE.
  • Figure S3: A display of the smoke plume (left) and the fire front (right) progression of the Williams Flats fire on August 06, 2019, as captured by eMAS and segmented via SIT-FUSE. When used in conjunction with the data in Figure \ref{['fig:goes_progression']} we can look across large spatial domains at fine temporal scales as well as fine-scale detail at higher spectral resolutions.
  • Figure S4: A display of the smoke plume (left) and the fire front (right) progression of the Williams Flats fire on August 06, 2019, as captured by eMAS, MASTER, AirMSPI, and AVIRIS-C and segmented via SIT-FUSE. This combination of instrumentation maximizes the temporal resolution at the associated high spatial and spectral (/polarization) resolutions of the airborne instrumentation. When compared to the data in Figure \ref{['fig:eMAS_progression']}, this increases the yields even more for monitoring and science capabilities when used in conjunction with geospatial and polar orbiting instrumentation.
  • Figure S5: A flow diagram for the processing of one input type (single instrument or fusion set) through SIT-FUSE
  • ...and 13 more figures