Expanding Connected Components from Alternative Terminals: Global Optimization for Freshwater Fishes Under the UN's 30x30 Conservation Goal

Yue Mao, Zhongdi Qu, Imanol Miqueleiz, Aaron M. Ferber, Sami Wolf, Marc Grimson, Sebastian Heilpern, Felipe Siqueira Pacheco, Alexander S. Flecker, Peter B. McIntyre, Carla P. GomesIJCAI 2025 · DOI

Abstract

The United Nations' "30x30" target aims to protect 30% of the Earth's land and water by 2030. Freshwater ecosystems are often neglected in these efforts despite the rapid decline of aquatic biodiversity. This study introduces the Expansion of Connected Components from Alternative Terminals (ECCAT) framework, a scalable optimization model designed to maximize freshwater fish protection while ensuring habitat connectivity. By applying a Mixed Integer Programming (MIP) approach to global hydrological and species data, the researchers demonstrate that connectivity-aware conservation is computationally feasible and significantly more effective than current strategies.

Introduction

Freshwater fish are declining at an alarming rate, yet current protected areas (PAs) often leave species-rich rivers and lakes vulnerable. Unlike terrestrial conservation, freshwater protection must prioritize connectivity; fish require contiguous habitats to migrate and survive. Mathematically, this necessitates expanding existing PAs into connected networks rather than selecting isolated high-value spots. This creates a massive combinatorial optimization problem: maximizing biodiversity protection under a strict 30% area budget while ensuring all new areas are physically linked to existing ones.

Global distribution of freshwater fish species richness, highlighting areas of high conservation priority.

Methodology

The researchers formulated the problem as ECCAT and proved it to be strongly NP-complete. They proposed a Mixed Integer Programming (MIP) formulation using a flow-based encoding for connectivity. In this model, existing PAs act as "sources," and "protection fluid" flows into adjacent units only if they are selected for protection, ensuring physical linkage. The study utilized three massive datasets: HydroBASINS (190,675 hydrological units), the IUCN Red List (13,740 fish species), and the World Database on Protected Areas (WDPA).

Results and Discussion

The algorithm successfully solved the global problem (200,000 variables) in just 1 hour and 20 minutes. A key insight is the "Easy-Hard-Easy" pattern in computational complexity, where the hardest instances occur at an area budget ratio of approximately 15%.

Runtime analysis across different area budget ratios for planar grids, showing the peak computational difficulty at the 15% threshold.

The global optimization approach increased the aggregate species protection score by 286% and effective protection (Minimum Viable Range) by 31 times compared to current levels. Furthermore, the study found that river networks (tree structures) are roughly 1,000 times faster to solve than planar grids, due to their hierarchical nature. To address geographic imbalances, a "Per Main Basin" approach was tested, achieving a 216% increase in protection while distributing conservation efforts more equitably across countries.

Conclusion

While the ECCAT framework provides a robust ecological baseline for the 30x30 goal, it currently excludes sociopolitical factors, land costs, and human infrastructure like dams. Future research should incorporate multi-objective optimization and dynamic connectivity to account for shifting habitats due to climate change.

Summary updated

BibTeX
@inproceedings{mao2025expanding,
  author = {Yue Mao and Zhongdi Qu and Imanol Miqueleiz and Aaron M. Ferber and Sami Wolf and Marc Grimson and Sebastian Heilpern and Felipe Siqueira Pacheco and Alexander S. Flecker and Peter B. McIntyre and Carla P. Gomes},
  title = {{Expanding Connected Components from Alternative Terminals: Global Optimization for Freshwater Fishes Under the UN's 30x30 Conservation Goal}},
  booktitle = {International Joint Conference on Artificial Intelligence},
  year = {2025},
  doi = {10.24963/ijcai.2025/1090},
}
Figure