Pyxccd: An Efficient Python Package for Break-aware Time Series Analysis of Earth Observation Data

Read full text Summarise with VIRAT
Authors: Su Ye, Yingchu Hu
Cited by: 0
References: 21

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Pyxccd is an open-source, cross-platform Python package (installable via PyPI) for break-aware analysis of Earth observation time series, supporting retrospective disturbance mapping and near-real-time (NRT) monitoring. It implements the two CCDC-like algorithms: COLD (the latest version) and S-CCD 2.0 (state-space formulation to enable NRT application). Additionally, S-CCD 2.0 adds an anomaly-break hierarchical decision rule that improves robustness for coarse-resolution products and can output latent states for interpretable decomposition. A hybrid C/Python architecture provides high performance with a user-friendly API, plus pixel- and tile-based workflows and utilities for large-area orchestration. On 6,488 independently interpreted Landsat disturbance plots, COLD and S-CCD 2.0 achieve comparable accuracy (maximum F1=0.664 vs 0.653). S-CCD 2.0 is 1.4–1.9× faster for retrospective processing and 3–6× faster for NRT updating, with increasing gains as band numbers grow. Overall, pyxccd lowers the barrier to reproducible, efficient, and operational continuous change detection from Earth observation time series.

Fields of Study & Subjects

Remote-Sensing Image Classification Remote Sensing in Agriculture Remote Sensing and Land Use

Publication details

Publication date
21 March 2026
Language
EN
Document type
Preprint