Using LandTrendr to analyze forest disturbance, recovery, and attribution in Hunan province from 2001 to 2024

• Pixel-level detection of forest disturbance and recovery at the provincial scale. • Subtropical forests in Hunan show high fluctuations but low net loss (2001–2024). • Patch-scale attribution of forest disturbances using random forest classification. • Distinct spatial and morphological patterns among disturbance types. • Fire is the dominant disturbance driver, followed by logging and land conversion. Forest dynamics monitoring is an important basis for assessing the stability of regional ecosystems and formulating sustainable management strategies, and forest evolution patterns under the interaction of frequent human activities and natural disturbances in subtropical regions have not been fully revealed. This study uses Google Earth Engine and Landsat data (2001–2024) with the LandTrendr algorithm to detect spatiotemporal forest disturbances and recovery in Hunan Province, applying a random forest classifier to identify disturbance types. The results showed that from 2001 to 2024, the forest disturbance area in the study region was 4,723.46 km 2 , accounting for 3.48 % of the total forest area; the recovery area was 4,380.33 km 2 , accounting for 3.23 %; and the net loss was 343.13 km 2 , representing only 0.25 %. Among the disturbance types, weak disturbances accounted for 53.35 % of the total disturbed area, while low-level recovery accounted for 70.4 % of the total recovery area. Fire was the primary disturbance factor, accounting for 40.7 %, followed by land use conversion, which accounted for 27.3 %. This study provides a scientific basis for adaptive management, supporting targeted restoration and policy-making to enhance forest resilience and promote sustainable forest use in subtropical regions.