Speaker
Description
The argan tree (Argania spinosa L. Skeels), endemic to Morocco and recognized by UNESCO, is a key component of ecosystems in the Province of Essaouira. However, this forest is increasingly affected by climate change and human pressures, leading to progressive degradation. Monitoring tree density at a large scale is therefore essential for sustainable management.
An integrated approach combining high-resolution aerial imagery, manual annotation, and a deep learning model (YOLOv8 Nano) was applied. A total of 1,000 images (0.0655 km² each) were collected across the province, of which 56% contained argan trees and were annotated. The trained model was then applied to 102 forest units. Field validation was conducted in 6 forests, using 10 plots per site.
The model achieved a mAP@50 of 0.72, with precision 0.74 and recall 0.68. Estimated tree densities were generally consistent with field data across different density classes (low, medium, high), indicating the potential of the approach for operational argan tree density mapping.
This study highlights the effectiveness of combining remote sensing and artificial intelligence for large-scale monitoring of argan tree density. The approach provides a reliable tool to support conservation and sustainable management of argan ecosystems.