From lakes and glades to viability algorithms: automatic classification of system states according to the topology of sustainable management

Tim Kittel, Finn Müller-Hansen, Rebekka Koch, Jobst Heitzig, Guillaume Deffuant, Jean‐Denis Mathias, Jürgen KurthsThe European Physical Journal Special Topics 2021 · DOI · arXiv

Introduction

In sustainability science, concepts like "planetary boundaries" and "safe operating spaces" have become central to policy discussions. However, identifying boundaries is only the first step. To manage complex systems like the global climate-economy, we must understand how internal dynamics—such as economic growth and carbon cycles—interact with management choices. This paper bridges the gap between qualitative sustainability frameworks and quantitative control theory by automating the classification of system states using the Topology of Sustainable Management (TSM).

The Topology of Sustainable Management (TSM)

The TSM framework partitions a system's state space into intuitive regions based on their long-term sustainability and the necessity of intervention. These regions are often described using aquatic metaphors to represent different levels of risk and manageability:

  • Shelters: Safe zones where the system remains in a desirable state indefinitely without any management intervention.
  • Glades and Lakes: Regions where safety is reachable, but the system may need to pass through "dark" (undesirable) states first.
  • Backwaters: Areas where the system can only remain safe through constant, active management.
  • Trenches: Inevitable failure zones where no management can prevent the system from entering an undesirable state.
Conceptual illustration of the Topology of Sustainable Management (TSM) regions using aquatic metaphors.

While conceptually powerful, TSM was initially limited to low-dimensional models that required manual analysis. This research introduces a computational toolkit to automate this classification for more complex, higher-dimensional models.

Methods: Operationalizing TSM with Viability Theory

The authors translate TSM components into the language of Viability Theory, a subfield of control theory. The classification relies on two core concepts: Viability Kernels (states where you can stay safe forever) and Capture Basins (states from which you can reach a target).

To compute these sets, the authors utilize and extend the Saint-Pierre Algorithm. This algorithm discretizes the state space and iteratively "prunes" points that fail to meet viability or capture criteria.

The decision tree used for the automatic classification of system states into the main cascade and manageable partitions.

Key technical innovations include:

  • Handling "Eddies": A new extension to identify regions where the system can cycle in and out of desirable states.
  • Numerical Stability: To handle real-world economic models, the authors introduce coordinate transformations (mapping unbounded variables like wealth to a bounded cube) and time homogenization to account for varying dynamic speeds.

Case Study: The AYS Model

The methods were tested on the AYS Model, a 3D dynamical model representing Atmospheric carbon stock ($A$), economic Yield ($Y$), and renewable energy knowledge Stock ($S$). Parameters were estimated using data from the World Bank and IPCC.

The analysis revealed the "Lake Dilemma": the current global state likely resides in a "time-limited lake." This forces a choice between maintaining current "sunny" conditions (low growth) or temporarily entering a "dark" region of high carbon emissions to rapidly build the knowledge stock necessary to reach a permanent "shelter."

Key Results and Insights

The automated approach successfully visualized complex theoretical shapes like "backwaters" in 3D space. Crucially, the research demonstrates that management does not just move the system's state; it changes the topology of safety. For instance, aggressive energy transition policies can expand "Glade" regions, making sustainable attractors reachable from a wider range of starting points.

Furthermore, the paper introduces "TSM bifurcations." Similar to phase transitions in physics, a system can undergo a topological shift where a safe "backwater" suddenly collapses into a "trench" if management parameters, such as growth rates, cross a critical threshold.

Limitations and Future Work

Despite its success, the Saint-Pierre algorithm faces the "curse of dimensionality" due to its reliance on grid discretization. Scaling to very high-dimensional models may require machine learning-based classification. Additionally, while the current framework is deterministic, future iterations aim to incorporate stochasticity to better reflect the noise and random shocks of real-world systems.

Summary updated

BibTeX
@article{kittel2021from,
  author = {Tim Kittel and Finn Müller-Hansen and Rebekka Koch and Jobst Heitzig and Guillaume Deffuant and Jean‐Denis Mathias and Jürgen Kurths},
  title = {{From lakes and glades to viability algorithms: automatic classification of system states according to the topology of sustainable management}},
  journal = {The European Physical Journal Special Topics},
  year = {2021},
  doi = {10.1140/epjs/s11734-021-00262-2},
}
Figure