Google Earth Engine App

Sihlwald dead trees
(2014-2024)

This project uses deep learning to automatically detect dead trees in multi-temporal, high-resolution aerial imagery of Switzerland's Sihlwald forest. Covering 2014–2024, it maps temporal changes in the number and spatial distribution of dead trees. Detections can be filtered by model confidence and bounding-box area. An interactive map supports the exploration of forest mortality patterns and long-term forest health assessment.

How to use

Filters

Confidence

The model's certainty that a detection is a dead tree. Keep only detections above this level; raise it for fewer, more reliable results.

Detection size (BBox area)

Keep only detections within this size range, to exclude small or oversized detections.

Counts & chart

Year

Pick a year to see its count and detections on the map.

Dead tree count

Total detections for the selected year, under the current filters.

Bar chart

Compares all years (2014–2024) under the current filters. Updates only when filters change.

Reference

Study

Tree Crown Mortality and Defoliation Assessment in Temperate Forests Using Aerial Imagery and Deep Learning.

Aerial imagery

SWISSIMAGE 10 cm orthophotos (swisstopo).