Constrained Multiscale Dynamics (CMD): a reproducible framework and pipeline for detecting multiscale transitions in complex systems

CMD is a general-purpose, reproducible framework to detect multiscale transitions and event-like reorganizations in complex systems using time series data. This record contains the complete reference implementation of Constrained Multiscale Dynamics (CMD): a fully deterministic, end-to-end analysis pipeline designed for systems in which no global invariant laws, stationary regimes, or stable governing equations can be assumed. CMD targets intrinsic multiscale reorganization, including transitions, breakpoints, and event-like dynamics, by combining: scale-resolved statistical structure, temporal divergence and irreversibility diagnostics, residual-based change detection, and event-centered comparative analysis. Rather than characterizing systems through averaged structural states, CMD is explicitly designed to quantify dynamical change, scale transitions, and temporal asymmetry as primary signatures of complex behavior. CMD builds upon earlier retrospective, event-centered approaches to collective statistical organization, in particular the TAMC-FRANJAMAR framework for multistation seismic networks(https://doi.org/10.5281/zenodo.18085580), and its subsequent domain-specific extensions to: Volcanic seismicityhttps://doi.org/10.5281/zenodo.18348740 Controlled aerodynamic systemshttps://doi.org/10.5281/zenodo.18288994 While those works focused on specific physical domains and network structures, CMD generalizes the underlying methodology into a domain-agnostic multiscale framework, explicitly designed to characterize reorganization and transitions without relying on stationarity assumptions or invariant governing laws. Repository contents This repository includes: a fully deterministic, end-to-end analysis pipeline, documented Python scripts for multiscale metrics, breakpoint detection, transition analysis, and event detection, all configuration and execution logic required to reproduce the reported results, the accompanying paper describing the CMD framework and its empirical application. External reference data (the daily geomagnetic Kp index from NOAA/NGDC) are included exclusively for independent validation and comparison, and are not used for event detection within CMD. Position within the TAMC-FRANJAMAR research line CMD constitutes the domain-agnostic generalization of the TAMC-FRANJAMAR event-centered methodology. While TAMC-FRANJAMAR was originally developed for multistation seismic networks, CMD abstracts its core principles — multiscale organization, transition detection, temporal irreversibility, and event-centered comparative analysis — into a framework designed to operate across heterogeneous complex systems. The methodological core has been validated and deployed through multiple reproducible derived implementations, including: Multistation seismic networks (TAMC-FRANJAMAR core framework) Extreme heterogeneous events (DPRK underground tests, Beirut explosion, Chelyabinsk airburst, Brumadinho collapse) Volcanic reorganization dynamics (Stromboli, Kilauea, Icelandic systems) Controlled aerodynamic systems (dynamic reorganization under imposed modifications) These derived works provide empirical case studies demonstrating the transferability and structural robustness of the framework across physically distinct systems. Intended scope CMD is intended as a methodological and exploratory framework applicable to physical, biological, social, or synthetic systems exhibiting multiscale dynamics. The approach prioritizes transparency, reproducibility, and conceptual clarity over predictive claims.