To main content

ALDSI project - Automatic Landslide Detection with Satellite Imagery.

Abstract

As part of the ALDSI (Automatic Landslide Detection using Satellite Imagery) project, we are developing a web application using AI-assisted programming to streamline landslide mapping and review. The application brings established methods for landslide detection and back-dating into shared, repeatable workflows, with consistent metadata and processing records for imagery and polygon outputs. Reliable information on landslide occurrence and extent is essential for understanding geohazards and supporting disaster response. Satellite observations can help fill reporting gaps in remote areas, although cloud cover, terrain and ambiguous surface changes complicate interpretation. The application supports complementary use of optical and synthetic aperture radar (SAR) imagery and provides a framework for documenting the evidence behind mapping decisions. We have developed workflows for three applications: systematic mapping following major triggering events in Norway; back-dating landslide polygons in a region of Vietnam; and comparing landslide signatures in NISAR and Sentinel-1 imagery in a global study. The project aims to produce training data to improve automatic landslide detection models for testing in an operational setting by NVE. For a dataset of 14,000 landslides in Vietnam, preliminary tests have automatically assigned dates to 44% using Sentinel-1 and Sentinel-2 time series. Manual inspection suggests potential to increase this proportion. Further work will assess additional Norwegian triggering events, improve back-dating coverage and examine differences between NISAR and Sentinel-1 signatures. The presentation discusses how AI-assisted development can make research workflows more accessible and reproducible, while improving the documentation and traceability of the resulting data.

Category

Conference lecture

Language

English

Author(s)

Affiliation

  • SINTEF Community / Infrastructure

Presented at

Geofaredagen 2026

Place

Trondheim

Date

22.10.2026 - 23.10.2026

Organizer

NGU

Date

22.10.2026

Year

2026

View this publication at Norwegian Research Information Repository