Abstract
To overcome limitations in long-term forest disturbance monitoring in mountainous areas
arising from persistent cloud contamination and observation sparsity, this study proposed a forest
disturbance identification method based on multi-source satellite data fusion, combining Landsat and
Sentinel spectral time-series data (2014-2023) in the Qinling Mountains and employing an optimized
LandTrendr parameter scheme to derive disturbance year, magnitude, and duration, achieving detailed
disturbance detection and mapping and validating spatiotemporal consistency using GFC, FAGE
products, and visually interpreted reference samples. The results demonstrate that the proposed multisource
data fusion and optimized LandTrendr approach attained an overall accuracy of 96.51% with
a Kappa coefficient of 0.908, exhibiting strong agreement with validation data and superior detection
performance compared with the single-sensor LandTrendr model. During the period 2014-2023, the total
forest disturbance area in the Qinling Mountains reached about 4.01×103 km2, and disturbances were
predominantly distributed in foothill zones characterized by elevations near 800 m and slope gradients
of 15°-35°. The proposed forest disturbance detection framework and high-precision disturbance
products offer valuable support for long-term forest monitoring and sustainable management in regions
with complex topography.