An autoencoder-based model for forest disturbance detection using Landsat time series data

Monitoring and classifying disturbed forests can provide information support for not only sustainable forest management but also global carbon sequestration assessments. In this study, we propose an autoencoder-based model disturbance detection, which considers disturbances as anomalous events in temporal trajectories. The autoencoder network is established trained to reconstruct intact Then, the detection threshold derived by Tukey’s method based on reconstruction error of trajectory. assessment result shows that using NBR time series performs better than NDVI-based model, with overall accuracy 90.3%. omission commission errors are 7% 12%, respectively. Additionally, NBR-based implemented two test areas, accuracies 87.2% 86.1%, indicating robustness scalability model. Moreover, comparing three common methods, proposed better, especially accuracy. This study provides a novel robust approach acceptable enabling be identified regions limited reference data.