Boreal forests play a critical role in global carbon dynamics and climate regulation, yet their structural attributes remain poorly characterized, particularly in structurally complex ecosystems such as the northern treeline. Here, we explored the potential of Harmonized Landsat and Sentinel-2 (HLS) multispectral data to predict UAV-LiDAR-derived forest structure across sites in the western North American boreal forest. We extracted spectral features from peak and late summer HLS and used Random Forest models to predict Canopy Height and Crown Cover at 30 m resolution. Our results show strong relationships between spectral and structural metrics, with HLS NDVI and Tasseled Cap Wetness emerging as key predictors. Predictive performance is higher for dense and sparse forests than medium-density forests, and no significant differences are found between peak and late summer models. We compared our UAV-LiDAR Crown Cover estimates to the ABoVE Tree Canopy Cover product and identified overestimation of Crown Cover in the treeline ecotones. These findings highlight the value of fine-scale UAV-LiDAR structural data for algorithm building and assessments of satellite-derived products. Nevertheless, the high proportion of green understory reduces the sensitivity of HLS spectral features to canopy height and crown cover compared with applications in more productive forests. An open-access HLS–forest structure dataset is provided, containing HLS pixel-wise labeled forest structure information. By combining structural reference data with spectral HLS satellite imagery, this study contributes to filling the structural data gap at the boreal forest northern edge and to predicting forest structure in high-latitude ecosystems.