This review article by Giordan et al. (2018) provides a comprehensive synthesis of how Remotely Piloted Aircraft Systems (RPASs)—commonly known as drones—have transitioned from scientific novelties to essential tools for natural hazard management. By analyzing 171 papers, the authors map out the current state of the art in using these platforms to monitor landslides, floods, earthquakes, volcanoes, and wildfires.
Bridging the Methodological Gap
Natural disasters are increasing in frequency and impact, yet traditional remote sensing often falls short during the critical "emergency phase." Satellites can be hindered by cloud cover or infrequent revisit times, while manned aircraft are expensive and slow to deploy.
The core challenge identified is a "methodological gap." While researchers have proven that drones can capture spectacular photos and create detailed 3D models, the field has historically lacked standardized, rapid-response protocols that civil protection agencies can use reliably. RPASs fill a unique niche, providing ultra-high-resolution, on-demand spatial data that bridges the gap between ground-level observations and satellite imagery.
The RPAS Workflow
As a review article, the primary contribution is a systematic analysis of existing literature. However, the authors define a generalized RPAS Workflow that serves as a benchmark for the field, moving from initial planning to final information extraction:
- Mission Planning: Defining Ground Sample Distance (GSD) and flight goals.
- Hardware Selection: Choosing between fixed-wing (for large area coverage) and multi-rotor (for complex 3D structures and oblique views).
- Photogrammetric Workflow: Using Structure-from-Motion (SfM) algorithms to transform 2D images into 3D Point Clouds, Digital Surface Models (DSMs), and Orthophotos.
- Information Extraction: Moving beyond visualization to automated feature extraction using machine learning.
Key Insights and Results
- The "Oblique" Advantage: In earthquake scenarios, traditional top-down imagery often misses structural damage. RPASs excel at capturing oblique views of building facades, which are critical for identifying collapsed first floors or "soft-story" failures.
- The Time-Quality Trade-off: During an active emergency, a "rough" image mosaic produced in minutes is often more valuable than a perfectly georeferenced 3D model that takes 24 hours to process. The paper emphasizes the need for "fast-mosaicking" methods.
- Resolution Gains: RPASs routinely achieve GSDs in the 1–5 cm range, providing a level of detail (e.g., individual cracks in a dam or specific boulders in a debris flow) that is impossible from orbit.
- Disruptive Technology: The authors argue RPASs are a "disruptive technology" because they allow for high-frequency "multi-temporal" monitoring—comparing 3D models of the same site over days or hours to detect millimetric movements in a landslide.
Limitations and Future Directions
The biggest bottleneck currently isn't technology, but policy. Varying national regulations regarding "Line of Sight" (VLOS) and flight altitudes restrict the ability to use drones in wide-scale disasters. Furthermore, most current workflows still require significant manual intervention.
The authors point toward Convolutional Neural Networks (CNNs) as the future for automatic damage detection and emphasize a critical need to move processing to the "edge"—performing SfM and change detection on-board the aircraft or on a ruggedized field laptop to provide immediate feedback to rescue teams. The paper concludes by calling for the scientific community to help government agencies "codify" these experimental methods into official emergency manuals.
Summary updated