Key points are not available for this paper at this time.
Monitoring large-scale afforestation projects in arid and semi-arid environments requires accurate, high-resolution, and repeatable methods to assess tree survival and growth. In this study, we integrated unmanned aerial vehicle (UAV) multispectral imaging with an advanced object detection framework to evaluate vegetation establishment in the Shuayb Al-Budai afforestation site, part of the Imam Turki bin Abdullah Royal Natural Reserve, Kingdom of Saudi Arabia (KSA). Multispectral datasets were acquired using a MicaSense Altum-PT sensor and processed through a masked Region-based Convolutional Neural Network (RCNN) with two backbone architectures: ResNet-101 and VGG19-BN. The Mask R-CNN–ResNet-101 model achieved superior performance, with an overall accuracy of 90% and stable validation loss, outperforming VGG19-BN in per-class accuracy, generalization, and resistance to overfitting. The classification results indicated that 23% of pits contained no trees, 51% contained small trees, and 26% contained large trees, revealing heterogeneous establishment patterns linked to microtopographic and soil conditions. The methodology proved robust against site-specific challenges, including variable illumination, dust deposition, small canopy sizes, and spectral confusion with bare soil. The outputs provide actionable insights for targeted replanting, irrigation optimization, and performance tracking. By successfully scaling the detection framework to assess over 43,000 planting pits across a 5 km² area, this study provides a robust baseline for ecological monitoring. Furthermore, it establishes a spatial foundation that could be integrated with future ground-based Internet of Things (IoT) sensor networks to advance toward real-time, data-driven restoration management.
Morgan et al. (Fri,) studied this question.