The Physics-Informed World Model for Water Solvency.
Water decisions are made with fragmented data, disconnected models, and no shared way to understand where the water goes. Powered by Physics-Informed Machine Learning (PIML), H₂IQ gives us the intelligence to understand and manage water at the scale of entire watersheds, so we can save water, restore lakes and rivers, protect aquifers, and help secure a livable future.
Pilot cohorts forming · Research in progress · No public release yet
Managing Water in the Dark
Some data is missing. Much of what exists is fragmented across sensors, models, agencies, and records. Without a unified view of the watershed, we can't reliably understand where water is, where it goes, or what we can do about it.
The Data Lives in Silos
Sensors, satellites, water rights, field records, and operational systems each capture part of the picture, but rarely share a common language.
Some of the Picture Is Missing
Critical parts of the water system are poorly measured, unmonitored, or estimated. We often don't know what we don't know.
The Models Don't Agree
Hydrologic, operational, agricultural, and climate models describe different pieces of the system, often using different assumptions, scales, and definitions.
We Don't Speak the Same Water
Agencies, utilities, researchers, farmers, and technology systems describe the same physical system in different ways.
Water Has No Balance Sheet
We can measure individual flows, diversions, reservoirs, groundwater, ET, and return flows, but we lack a continuously reconciled picture of how those pieces connect.
We Have to Act Without Knowing
When information is incomplete or contradictory, conservation, allocation, infrastructure, and environmental decisions become harder to defend and harder to optimize.
We Act. We Measure. But We Still Don't Know What Worked.
Water systems are changing faster than our ability to understand them. Decisions made today can take years to reveal their consequences. We need to know what is happening now, while there is still time to act.
We Find Out Too Late
The state of a watershed can take months or years to become clear. Data arrives from different sources, important measurements are missing, and understanding the system requires enormous analytical effort.
Build a living model of the watershed.
H₂IQ continuously integrates observations, models, and physical relationships to show what is happening now, identify what we don't know, and reveal where better measurements are needed.
We Can't Tell What Worked
A conservation project may reduce measured water use, but that doesn't necessarily mean water reached the river, lake, or aquifer we intended to protect. The Jevons paradox shows how even well-intentioned efficiency measures can produce unexpected system-level outcomes.
Trace interventions through the system.
H₂IQ connects actions to their physical consequences, measuring changes in consumptive use, transit, return flows, and downstream delivery so we can determine whether an intervention actually produced the intended outcome.
We Can't See the Consequences Before We Act
Water rights, senior and junior users, impairment, timing, physical losses, and changing behavior make interventions inherently interconnected. Changing one part of the system can create consequences somewhere else.
Model the system before we change it.
H₂IQ lets water managers evaluate interventions against a physics-informed model of the watershed, revealing tradeoffs, unintended consequences, and likely outcomes before committing scarce water or resources.
From Disparate Data to Defensible Action
Four progressive stages under active development — spanning conservation impact measurement and wet water shepherding verification.
Multi-Source Data Ingestion
Bringing together the observations, measurements, and records needed to build a complete picture of the watershed.
- Satellite-derived ET
- Streamflow/Gage
- Soil/Weather
- Water Rights
- Reservoir/Infrastructure
The PIML Advantage: Beyond Standard AI
Pure machine learning fails in hydrology because statistical models violate conservation-of-mass physics. Pure physical modeling fails because unmetered basins lack ground-truth data. H2IQ combines both.
Physics-Guided Neural Networks (PGNNs)
Embeds fundamental conservation-of-mass equations directly into the neural network architecture, so predictions respect physical law rather than overfitting to sparse observations.
Multi-Source Data Ingestion
Fuses satellite-derived ET, streamflow gages, soil sensors, weather grids, water-rights ledgers, and infrastructure records into a unified basin state.
Data Void & Anomaly Resolution
Infers microclimate thermodynamics across unmetered surrounding sub-basins to fill critical data voids automatically and flag anomalies before they distort decisions.
Who we're building the first pilots with
We are onboarding a limited number of design partners across three tracks during the private beta.
State Regulators & DNR
Audit-proof conservation verification and automated compliance reporting built on evidence a court can follow.
Apply for this pilot track →Water Conservancy Districts
Predictive transit loss math and zero-impairment proof, so delivery decisions stand up before junior users do.
Apply for this pilot track →Hyperscale Data Centers
Pre-construction micro-hydrology due diligence and local 'Water Positive' replenishment matching.
Apply for this pilot track →Bridge the Gap Between Paper Water and Physical Reality.
We're selecting basins for early pilot builds. Tell us where you sit and we'll share the development roadmap.