Anderson, J & Pearse, B 2025, 'Monitoring success: remote sensing and artificial intelligence for tracking success 
of individual seedlings at a revegetation trial in northern Canada', in S Knutsson, AB Fourie & M Tibbett (eds), Mine Closure 2025: Proceedings of the 18th International Conference on Mine Closure, Australian Centre for Geomechanics, Perth, pp. 1-14, https://doi.org/10.36487/ACG_repo/2515_67
(https://papers.acg.uwa.edu.au/p/2515_67_Anderson/)
Abstract: Progressive reclamation and reclamation trials are vital to mine closure success, providing opportunities to refine and validate reclamation prescriptions. Comprehensive monitoring is essential for evaluating the relative success of treatments over time. The research objective was to identify newly planted tree seedlings, classify them by species, and develop a robust indicator of seedling health using high-spatial-resolution multispectral imagery, object-based image analysis, and machine learning. In July 2023, a 2 hectare revegetation trial established on a mined waste rock dump was planted with aspen (Populus tremuloides), lodgepole pine (Pinus contorta), and white spruce (Picea glauca). Multispectral imagery was acquired via remotely piloted aircraft system (RPAS) in August 2023, and 30 permanent sampling plots (PSPs) were established across the trial, where each seedling’s position and species were recorded (n = 1,545). The multispectral imagery was filtered using a vegetation index to isolate individual seedlings across the trial, and a Random Forest model was used to classify seedlings by species. Subsequent RPAS imagery and vigour rankings for seedlings within PSPs were collected in August 2024 and analysed using a similar approach to classify survival (dead or alive) and vigour (0–5) or generate seedling size percentiles as indicators of health. Species and survival were classified with an average F1-score of 0.94 and 0.98, respectively. Vigour classifications for aspen, lodgepole pine, and white spruce yielded average F1-scores of 0.82, 0.86, and 0.83, respectively. Increases in size percentiles coincided with increases in field-assessed vigour, demonstrating a clear relationship between the two. This study resulted in the precise geotagging of 4,942 aspen, 2,524 white spruce, and 2,819 lodgepole pine across the reclamation trial, demonstrating the ability to monitor the progression of individual seedling performance immediately after planting. This approach improves the efficiency of monitoring planting programs, enabling deeper insights into the long-term effects of different treatments on seedlings.

Keywords: revegetation trial, drone, UAV, RPAS, remote sensing, machine learning