Understanding sensitivity and accuracy of time-series change detection algorithm across multiple land disturbance scenarios

Dense time series data have been widely applied in mapping land-cover disturbances. Numerous continuous land-cover change detection algorithms have been developed, among which Continuous Change Detection and Classification (CCDC) is one of the most prominent. However, its parameter sensitivity and applicability across diverse scenarios remain insufficiently explored. In this study, we built a multi-scenario land-cover change validation dataset by combining the LANDFIRE annual disturbance dataset and visual interpretation, and further comprehensively evaluated the uncertainty and robustness of the CCDC algorithm across four disturbance scenarios (Harvest, Fire, Insect/Disease, and Weather). Specifically, the algorithm's scenario-specific applicability and parameter sensitivity were evaluated using multiple performance metrics together with Sobol sensitivity analysis. Then, parameter optimization was conducted from both single-metric and multi-metric perspectives based on TOPSIS (Technique for Order of Preference by Similarity to Ideal Solution). The results indicate that: 1) CCDC achieves higher overall accuracy in Fire (0.404) and Harvest (0.409) scenarios compared to Insect/Disease (0.340) and Weather (0.350) scenarios; 2) CCDC is sensitive to the key parameters among all scenarios, among which the consecutive observation threshold holds the most significant influence on model performance (SN ≥ 0.358), followed by the chi-square threshold and regularization coefficient; 3) CCDC favors optimal observation thresholds varied distinctively by scenario, specifically 8 and 9 for Fire and Harvest, while 11 and 14 for Weather and Insect/Disease, respectively. In contrast, the TOPSIS-based multi-metric optimization provides a more balanced evaluation, yielding stable and reliable parameter configurations for practical application. Ultimately, this study bridges the knowledge gap concerning CCDC's parameter sensitivities and applicability, offering practical guidelines for parameter optimization across diverse disturbance scenarios.