Forest Disturbance Monitoring and Terrain Control Effects Based on LandTrendr Algorithm

By integrating the Google Earth Engine (GEE) platform with the LandTrendr time-series segmentation algorithm, this study developed a multi-dimensional median synthesis method to enhance time-series continuity. A three-tiered disturbance type classification system based on disturbance duration, recovery rate, and trajectory morphology was constructed, and the driving mechanisms of terrain factors were quantitatively analyzed. Key results revealed that the Mean Spatial Consistency (MSC) of extracted forest disturbance patches in the karst mountainous area of eastern Yunnan from 1993 to 2024 reached 85.7%, and the temporal accuracy of disturbance year detection was high (R² = 0.95, MAE = 1.15 a). Cumulative disturbance area reached (872.7 ± 15.2) km², peaking historically at 228.7 km² in 2023. The evolution exhibited three distinct phases: fluctuating decline (1993–2010), oscillating rise (2011–2020), and extreme disturbance (2021–2024). Spatially, disturbances were significantly concentrated in the mid-altitude zone (1600–1800 m; 58.6%) and gentle slopes (≤10°; 69.2%). Anthropogenic disturbance dominated (63.1%), followed by fire disturbance (27.0%, primarily human-ignited) and drought stress disturbance (9.9%), with terrain accessibility identified as the key regulatory factor. This study establishes a forest disturbance monitoring framework suitable for complex karst terrain, revealing a four-dimensional pattern encompassing temporal dynamics, spatial distribution, type composition, and terrain regulation mechanisms. It provides a technical foundation for precise identification of regional forest disturbance risks, dynamic monitoring, and adaptive management.