Background

Over the past decades, an increasing body of science has highlighted the deeply complex interactions of all aspects of life and the environment. It encompasses climate change, biodiversity loss, pollution, land and water usage, food security and other interlinked threats – in short, the full spectrum of human-induced disturbances on Earth that affect both human and ecological well-being. All contributors, such as human, animal, and environment, are inseparable at planetary scale, together forming interconnected, nonlinear dynamics across global societies, ecosystems, and economies. Planetary health science has advanced the thinking by clearly identifying 9 planetary boundaries, describing the limits within which human societies can safely operate. Already, some of the boundaries have been transgressed, while others are in critical stage. While many pathways linking boundary transgressions to Human well-being are established (e.g. via climate, water, food, and disease), further research is needed to identify additional pathways, quantify causal mechanisms, thresholds, and distributional impacts across regions and populations – in order to guide decision making and interventions to halt and reverse the trends.

 

This research is deeply complex, as interactions are non-linear and ultimately can only be analyzed when all aspects are looked at within one common space. The rapid development of AI allows us to address this challenge from a new perspective, allowing virtually unlimited data points to interact with each other. In 2024, Peking University has initiated the Planetary Health Axis System (PHAS) as a breakthrough large-model AI platform to help enrich planetary health science.

 

The PHAS Advantage

PHAS is a unique large-model AI system designed from the ground up to integrate and simulate the entire planetary health ecosystem. It contains over 48,000 country–year indicators spanning human, species, nature and societal metrics. It explicitly encodes the nine planetary boundaries and a four-axis health model capturing the joint state of human, species, nature and societal health. Through this massive, multidimensional data framework, PHAS can model global-scale dynamics with unprecedented fidelity.

 

By concentrating the 48,000 metrics, this system is able to show the current status of the four health axes of each country, regionally and globally, and to estimate its distance toward planetary boundaries, the safe operating limits within the Earth's bio-geophysical systems, as well as providing warning for risk indicators. What’s more, by leveraging deep causal modeling techniques, the system not only reproduces established scientific relationships, but can also reveal novel causal linkages and feedback that have been hidden in the prior analyses. Causal AI ensures the model focuses on cause-and-effect structures, yielding more robust, interpretable and reliable insights.

Integrated Modeling and Simulation

· At its core, PHAS provides insightful practices from integrated assessment and systems modeling. It combines geophysical, ecological and social domains via a dynamic general equilibrium economic module, yielding a holistic simulation of policy and development scenarios. Using large-scale learning algorithms and risk-scenario simulators, PHAS can project outcomes under alternative forward-looking scenarios. Its dynamic equilibrium framework unifies separate elements in one coherent model. Data and Scope. PHAS draws on an enormous international data repository: thousands of indicators per country-year (demography, epidemiology, economy, land use, climate variables, etc.) plus updated metrics on nine planetary boundaries. This rich dataset underpins simulations from the country to the globe, and connects sectors from different topics.

· Causal Model Engine. A key innovation is PHAS’s causal inference engine. Grounded in AI research, it encodes known causal graphs and tests hypothesized links. This engine not only replicates well-established science (validating the model) but also mines the data for previously unrecognized causal paths. In machine learning terms, this moves beyond mere correlation towards “Causal AI”, widely acknowledged as crucial for trustworthy models.

· Validation and Reliability. To enhance the learning system, PHAS employs multidimensional expert-informed validation layers. In practice, this means staged human-in-the-loop review: experts in the corresponding field regularly double check the model’s outcomes, flag data gaps, and calibrate the model against known benchmarks. Indeed, recent work stresses that complex climate/health AI must be collaborative with human experts to ensure less-biased, credible outcomes. PHAS embodies this principle through continuous cross-checks. This rigorous process – akin to internal peer review – increases confidence that PHAS simulations are both grounded in evidence and open to correction.

· Interactive Dashboard. PHAS includes a scalable, web-based dashboard interface. Users can query the system, test policy interventions, and visualize impacts across all four health axes and nine boundaries. The dashboard is designed for scientists, policymakers, and institutions: it makes the complex model transparent and accessible, with adjustable parameters and scenario branches.

Why PHAS Now?

Planetary health science is at the cusp of a critical development. We have abundant global data and unprecedented computing power, but our analytical tools are lagging. Conventional frameworks – disciplinary models or isolated studies – fail to capture the web of interconnections between societies, economies, and the biosphere. PHAS addresses this gap by introducing a large model approach for planetary health data. It is literally at the axis of planetary health research: integrating multiple axes of health, boundaries, and economic dimensions. By doing so, PHAS can help frame the big questions: How will interconnected shocks, such as pandemics, climate extremes and conflicts, propagate across regions? What policy mixes yield better outcomes for health and sustainability? Where are hidden leverage points in the global system?

 

Importantly, PHAS is designed to enrich the global scientific community. The system’s insights will be regularly released for peer review, and its scenarios will invite refinement by experts worldwide. We envisage PHAS as a strategic platform for collaboration: bringing together institutions with complementary expertise (e.g. climate scientists, economists, epidemiologists, data scientists) around a common modeling backbone.

 

In sum, PHAS represents a breakthrough large-model AI platform built for the scale and intricacy of today’s challenges. It is timely and aligned with global priorities: climate change, pandemics, biodiversity loss, and sustainable development all intersect in the planetary health paradigm. By launching PHAS at the World Health Summit, we signal a commitment to next-generation science and multilateral cooperation. We invite leading scientists and institutions to engage with PHAS – whether by contributing data, refining models, or co-designing research questions – as we move together to safeguard planetary health.

 

While PHAS is specifically developed to address the challenges around Planetary Health, arguably the largest threat to human civilization of our times, understanding the complex epidemiological questions underpinning other mega trends in global health, such as chronic diseases and healthy aging, faces similar challenges.