The future of tropical forests hinges on the balance between disturbance rates, which are expected to increase with climate change, and tree growth. Whereas tree growth is a slow process, disturbance events occur sporadically and tend to be short-lived. This difference challenges forest monitoring to achieve sufficient resolution to capture tree growth, while covering the necessary scale to characterize disturbance rates. Airborne LiDAR time series can address this challenge by measuring landscape scale changes in canopy height at 1 m resolution. In this study, we present a robust framework for analysing disturbance and recovery processes in LiDAR time series data. We apply this framework to 8000 ha of old-growth tropical forests over a 4–5-year time frame, comparing growth and disturbance rates between Borneo, the eastern Amazon and the Guiana shield. Our findings reveal that disturbance was balanced by growth in eastern Amazonia and the Guiana shield, resulting in a relatively stable mean canopy height. In contrast, tall Bornean forests experienced a decrease in canopy height due to numerous small-scale (<0.1 ha) disturbance events outweighing the gains due to growth. Within sites, we found that disturbance rates were weakly related to topography, but significantly increased with maximum canopy height. This could be because taller trees were particularly vulnerable to disturbance agents such as drought, wind and lightning. Consequently, we anticipate that tall forests, which contain substantial carbon stocks, will be disproportionately affected by the increasing severity of extreme weather events driven by climate change.