PCA-OS: A Planetary Climate Adaptation Operating System

Official repository and project page for the KDD 2026 Blue Sky Ideas Track paper:

PCA-OS: A Planetary Climate Adaptation Operating System Chaoyue He¹, Xin Zhou¹, Di Wang¹, Hong Xu¹, Wei Liu², Chunyan Miao¹* ¹ Nanyang Technological University, Singapore · ² Alibaba Group · * Corresponding author Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2 (KDD 2026), August 9–13, 2026, Jeju Island, Republic of Korea.

TL;DR

Climate ML is spectacularly good at producing read-only futures — hazard maps that say what might happen, but not what to do, where, when, for whom, and under which futures. PCA-OS is a field-level provocation: adaptation should be framed as a continual learning and decision loop —

measurement → causal estimation → robust, equity-constrained optimization

— in which interventions (not hazard maps) are the primary scientific object: first-class, versioned, causally evaluated, and auditable.

The Three Core Objects

PCA-OS is a decision-support operating abstraction built on an intervention-aware global causal knowledge graph. It standardizes schemas, versioned updates, query primitives, and audit interfaces across three shared infrastructural artifacts:

	Object	What it does
🧾	Adaptation Intervention Ledger	Records where and when interventions occur — type, intensity, footprint — with provenance and uncertainty, fused from EO change signals, SAR/event response, administrative text, and operational/participatory traces.
🗺️	Causal Effect Atlas	Stores scenario-indexed, spillover-aware causal estimates as explicit objects: estimands, identification assumptions, diagnostics, sensitivity bounds, and transportability warnings.
📊	Robust Portfolio Decision Layer	Turns estimates into intervention portfolios under explicit budget, equity-floor, and bounded-harm constraints, scored by Robust Decision Regret across climate scenarios.

Every recommendation must be traceable back to specific ledger versions, atlas entries, constraints, and human overrides — making the OS claim falsifiable.

Four Design Commitments

1. Decision-first — outputs are intervention, effect, and portfolio objects, not hazard maps alone.
2. Interventions as first-class objects — versioned, geolocated, uncertain, and causally queryable.
3. Uncertainty-forward — measurement, identification, and climate-scenario uncertainty propagate into the decision layer.
4. Normative constraints as infrastructure — equity, no-harm, and contestability are encoded inside the optimization and the interface, not post-hoc commentary.

AdaptBench

We outline AdaptBench, an evaluation suite where systems are tested on intervention mapping, causal estimation, and portfolio choice — not prediction alone. Systems can fail for untraceable, maladaptive, or harmful decisions despite accurate hazard prediction. Metrics include ledger calibration, interval coverage, spillover error, robust decision regret, inequity gaps, and bounded-harm violation rates.

Running Exemplars

• Cool roofs for extreme heat — the minimum viable, city-scale deployment: detect parcel-level retrofits from EO + permit text, estimate avoided heat with neighborhood spillovers, optimize subsidy rollout under budget, equity-floor, and no-harm rules.
• Flood defenses with spillovers — levees and drainage retrofits that can redirect water and externalize harm.
• Urban greening & nature-based solutions — cooling benefits vs. green-gentrification risks and distributive outcomes.
• Compound environmental burdens — heat, ozone, traffic, and logistics-burden interactions that shift risk unevenly.

Repository Layout

├── paper/
│   ├── PCA-OS_KDD26_BlueSky.pdf   # Camera-ready paper (CC BY 4.0)
│   ├── pca-os.bib                 # BibTeX entry
│   └── figures/                   # Vector overview figure
├── docs/                          # Project page (GitHub Pages)
└── CITATION.cff                   # Citation metadata (GitHub "Cite this repository")

Citation

If you find this work useful, please cite:

@inproceedings{he2026pcaos,
  author    = {He, Chaoyue and Zhou, Xin and Wang, Di and Xu, Hong and Liu, Wei and Miao, Chunyan},
  title     = {{PCA-OS}: A Planetary Climate Adaptation Operating System},
  booktitle = {Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery
               and Data Mining V.2 (KDD 2026)},
  year      = {2026},
  address   = {Jeju Island, Republic of Korea},
  publisher = {ACM},
  doi       = {10.1145/3770855.3818654},
  isbn      = {979-8-4007-2259-2}
}

Acknowledgments

This research is supported by the RIE2025 Industry Alignment Fund–Industry Collaboration Projects (IAF-ICP) (Award I2301E0026), administered by A*STAR, and by Alibaba Group and NTU Singapore through the Alibaba–NTU Global e-Sustainability CorpLab (ANGEL).

License

This work is licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0), matching the paper's publication license.