Accurately monitoring forest fire activities is critical to understanding carbon dynamics and climate change. Three-dimensional (3D) canopy structure changes caused by fire make it possible to adopt Light Detection and Ranging (LiDAR) in burned forest classification. This study focuses on the effects of spatial resolution when using LiDAR data to differentiate burned and unburned forests. The National Aeronautics and Space Administration’s (NASA) Ice, Cloud, and land Elevation Satellite-2 (ICESat-2) mission provides LiDAR datasets such as the geolocated photon data (ATL03) and the land vegetation height product (ATL08), which were used in this study. The ATL03 data were filtered by two algorithms: the ATL08 algorithm (ILV) and the adaptive ground and canopy height retrieval algorithm (AGCH), producing classified canopy points and ground points. Six typical spatial resolutions: 10, 30, 60, 100, 200, and 250 m were employed to divide the classified photon points into separate segments along the track. Twenty-six canopy related metrics were derived from each segment. Sentinel-2 images were used to provide reference land cover maps. The Random Forest classification method was employed to classify burned and unburned segments in the temperate forest in California and the boreal forest in Alberta, respectively. Both weak beams and strong beams of ICESat-2 data were included in comparisons. Experiment results show that spatial resolution can significantly influence the canopy structures we detected. Classification accuracies increase along with coarser spatial resolutions and saturate at 100 m segment length, with overall accuracies being 79.43 and 92.13% in the temperate forest and the boreal forest, respectively. Classification accuracies based on strong beams are higher than those of using weak beams due to a larger point density in strong beams. The two filtering algorithms present comparable accuracies in burned forest classification. This study demonstrates that spatial resolution is a critical factor to consider when using spaceborne LiDAR for canopy structure characterization and classification, opening an avenue for improved measurement of forest structures and evaluation of terrestrial vegetation responses to climate change.