Efficient continuous land change monitoring with pyxccd and S-CCD 2.0

Su Ye , Yingchu Hu

Continuous change detection from satellite time series requires algorithms that are robust to structural breaks, computationally efficient, and accessible for reproducible large-area processing. We present pyxccd, an open-source, cross-platform Python package for continuous land change monitoring based on dense satellite time series. This package provides high-performance implementations of the newly proposed S-CCD 2.0 and COLD (the latest version of CCDC), through C kernels wrapped by a user-facing Python API, with pixel- and tile-based workflows for operational mapping. Relative to earlier S-CCD implementations, S-CCD 2.0 introduces an anomaly-break hierarchical decision rule for coarse-resolution products and exposes latent state estimates for interpretable trend-seasonality decomposition. Using 6,488 independently interpreted Landsat disturbance plots, S-CCD 2.0 achieved disturbance-detection performance comparable to COLD, with maximum F1 scores of 0.653 for S-CCD 2.0 and 0.664 for COLD. S-CCD 2.0 reduced computation by 1.4-1.9x for retrospective processing and 3-6x for near-real-time updating, with larger gains as spectral dimensionality increased. Application examples across Landsat and coarse-resolution vegetation products demonstrate utilities of pyxccd for reproducible disturbance monitoring. This new tool lowers the technical barrier to scalable, near-real-time monitoring of land-surface change from dense satellite archives.