The large-scale restoration of tropical and subtropical rainforests is crucial for mitigating climate change and biodiversity loss. Disturbances such as fire and wind potentially undermine efforts to restore degraded landscapes, interacting with the existing vegetation and background topography to produce complex patterns of damage. It is therefore crucial for us to understand these interactions and study the factors that contribute to disturbance resilience. Rapid developments in the field of remote sensing have provided new tools to study forest-disturbance dynamics across unprecedented spatiotemporal scales. In this thesis, a range of high-resolution remote sensing products, including aerial imagery, satellite multispectral imagery, and airborne LiDAR scans, were used to evaluate how fires and tropical cyclones have affected vegetation in wet subtropical Hong Kong. Chapter 1 provides an overview of how forest disturbances interact with restoration ecology. It then describes the vegetation history of Hong Kong, highlighting how the region represents an interesting case study as a long-running restoration project over highly degraded landscapes in the wet tropics. Chapter 2 reconstructs the fire history of Hong Kong using a 34-year Landsat imagery time series. Burn area detection in the wet tropics and subtropics is challenging due to high cloud cover and rapid revegetation of burn areas. A pipeline was developed to process hundreds of satellite multispectral images and accurately map out thousands of burnt areas. The pipeline additionally dated every detected burn area polygon and estimated burn severity for pixels in the burn area. The final product is the first of its kind in wet tropical Asia. Chapter 3 proceeds to use this burn area and severity time series to study fire-vegetation feedbacks in Hong Kong. When early successional vegetation is more fire susceptible than late-successional closed-canopy forests, positive fire-vegetation feedbacks are created. These feedbacks can then form “fire traps” that undermine restoration of degraded landscapes. Here, fire occurrence and post-fire recovery in different vegetation types were investigated. The results provided compelling evidence for the presence of strong fire traps in Hong Kong. Chapter 4 further expands on these results by investigating how landscapes can escape these fire traps. The results demonstrate that fire suppression increased forest cover in Hong Kong, and these changes could be accurately modelled by parameters estimated from remote sensing. To meet restoration objectives, land managers would benefit from explicit, quantifiable targets based on model predictions to ensure the restoration success is not undermined by changes in fire regimes under climate change. Chapter 5 describes a pipeline to model long-term mean and typhoon maximum wind speeds across the rugged topography of Hong Kong, as a precursor for Chapter 6. Specifically, wind models based on computation fluid dynamics (CFD) modelling were validated by wind data collected from a dense network of weather stations and our own anemometers. Chapter 6 analyses the resulting wind maps and a repeated LiDAR dataset (2010, 2017, 2020) to study forest resilience against strong tropical cyclones. The LiDAR dataset captured the forest damage incurred during Typhoon Mangkhut in 2018, which was the strongest tropical cyclone to affect Hong Kong in over 40 years. Plantations, tall forests, and normally wind-sheltered forests were found to be more susceptible to tropical cyclones. Effects of tropical cyclones and wind exposure cascaded through time to create strong wind-related limits on local forest height. Overall, this thesis provides a detailed account of the patterns of resilience against fire and tropical cyclones in the wet tropics. Such knowledge on resilience could help land managers better plan and restore degraded landscapes in the wet tropics under changing climate and disturbance regimes.