Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Multispectral Indices for Wildfire Management

Afonso Oliveira, João P. Matos-Carvalho, Filipe Moutinho, Nuno Fachada

TL;DR

Wildfire management demands timely, accurate mapping of fuels, water barriers, and infrastructure. The paper surveys multispectral indices and evaluation metrics, and demonstrates through two Portuguese case studies how NDVI, MNDWI, and MSR enable effective vegetation, water, and artificial-structure extraction from Landsat-8 and Sentinel-2 data. It shows that index performance is region- and feature-dependent, with NDVI-family indices excelling in vegetation, MNDWI in water, and MSR for built-up areas, while corrections and data fusion can enhance robustness. The findings provide practical guidance for wildfire monitoring and planning, and highlight directions for integrating machine learning and hyperspectral data to further improve accuracy.

Abstract

The increasing frequency and severity of wildfires requires advanced methods for effective surveillance and management. Traditional ground-based observation techniques often struggle to adapt to rapidly changing fire behavior and environmental conditions. This paper examines the application of multispectral aerial and satellite imagery in wildfire management, emphasizing the identification and analysis of key factors influencing wildfire behavior, such as combustible vegetation and water features. Through a comprehensive review of current literature and the presentation of two practical case studies, we assess various multispectral indices and evaluate their effectiveness in extracting critical environmental attributes essential for wildfire prevention and management. Our case studies highlight several indices as particularly effective for segmentation and extraction: NVDI for vegetation, MNDWI for water features, and MSR for artificial structures. These indices significantly enhance wildfire data processing, thereby supporting improved monitoring and response strategies.

Multispectral Indices for Wildfire Management

TL;DR

Wildfire management demands timely, accurate mapping of fuels, water barriers, and infrastructure. The paper surveys multispectral indices and evaluation metrics, and demonstrates through two Portuguese case studies how NDVI, MNDWI, and MSR enable effective vegetation, water, and artificial-structure extraction from Landsat-8 and Sentinel-2 data. It shows that index performance is region- and feature-dependent, with NDVI-family indices excelling in vegetation, MNDWI in water, and MSR for built-up areas, while corrections and data fusion can enhance robustness. The findings provide practical guidance for wildfire monitoring and planning, and highlight directions for integrating machine learning and hyperspectral data to further improve accuracy.

Abstract

The increasing frequency and severity of wildfires requires advanced methods for effective surveillance and management. Traditional ground-based observation techniques often struggle to adapt to rapidly changing fire behavior and environmental conditions. This paper examines the application of multispectral aerial and satellite imagery in wildfire management, emphasizing the identification and analysis of key factors influencing wildfire behavior, such as combustible vegetation and water features. Through a comprehensive review of current literature and the presentation of two practical case studies, we assess various multispectral indices and evaluate their effectiveness in extracting critical environmental attributes essential for wildfire prevention and management. Our case studies highlight several indices as particularly effective for segmentation and extraction: NVDI for vegetation, MNDWI for water features, and MSR for artificial structures. These indices significantly enhance wildfire data processing, thereby supporting improved monitoring and response strategies.
Paper Structure (23 sections, 3 equations, 15 figures, 10 tables)

This paper contains 23 sections, 3 equations, 15 figures, 10 tables.

Table of Contents

  1. Introduction
  2. Multispectral Imaging
  3. Index Comparability and Metrics
  4. Feature Extraction
  5. Vegetation
  6. Soil
  7. Water
  8. Artificial Structures
  9. Burnt Areas
  10. Practical Case Studies
  11. Study I
  12. Data Collection
  13. Regions of Interest
  14. Results and Analysis
  15. Study II
  16. ...and 8 more sections

Figures (15)

  • Figure 1: Electromagnetic spectrum with approximate location of relevant bands.
  • Figure 2: Multispectral indices grouped by formulaic configuration and function, their respective formula, and a reference to the study that introduced them. i - Simple greenness indicators; ii - Anthocyanin reflectance indices; iii - Enhanced vegetation indices; iv - Soil-adjusted vegetation indices; v - Modified bare soil indices; vi - Terrain adjusted vegetation index; vii - ndhd composites; viii - ; ix - Water/moisture extraction indices; x - Burnt area extraction indices; xi - Artificial surface index and components; xii - Road extraction indices;
  • Figure 3: Feature extraction for wildfire prediction using multispectral indices.
  • Figure 4: RGB representation of the three regions of interest in Study I: A -- greater Lisbon area; B -- southern Portugal; and, C -- central Portugal.
  • Figure 5: Study I results. Columns A, B, and C represent three distinct test regions: the greater Lisbon area, southern Portugal, and central Portugal, respectively. Rows correspond to the outcomes of the following indices: NVDI, NDWI, ARI, MARI, ASI, REI, NBR, and BAI.
  • ...and 10 more figures