01
Connect the evidence
Bring monitoring data, spatial context, and model output into one traceable scientific view.
Project RAIN is funded by the U.S. Department of Energy through the Genesis Mission to accelerate AI-enabled scientific discovery.
AI for water science
Follow every signal through the system.
RAIN advances science-informed AI for complex water systems—connecting observations, physical models, and uncertainty into research workflows people can inspect and trust.
The project
Scientific AI, designed to move with the water.
Water research rarely fits inside one dataset or one model. RAIN is built around the relationships between them: observations and simulations, local measurements and system-scale behavior, fast AI inference and physical understanding.
A U.S. Department of Energy-led national initiative
RAIN and the Genesis Mission
The Genesis Mission is led by the U.S. Department of Energy and its 17 National Laboratories. Project RAIN is funded by DOE through the Genesis Mission, joining a national effort to connect scientific data, advanced computing, experimental facilities, and AI.
02
Disentangle Relationships
Use information theory and unsupervised ML to tease apart nonlinear, complex interdependencies while respecting the full descriptive power of the data.
03
Respect the physics
Use physical constraints to guide learning, expose implausible results, and focus computation where it matters.
04
Make uncertainty useful
Turn uncertainty from a footnote into an explicit part of research questions, comparisons, and decisions.
Project partners
Built across institutions and disciplines.
RAIN is being developed with national-laboratory and university collaborators whose expertise spans water science, subsurface-system modeling, scientific computing, and artificial intelligence.
Scientific foundation
From drought signals to coupled water systems.
RAIN builds on an established technical lineage of interpretable AI for water science. In a 2022 study co-authored by EnviTrace CTO Velimir Vesselinov, researchers combined climate projections, a process-based hydrologic model, and NMFk to reveal spatial and seasonal drought patterns across 134 Colorado River Basin subwatersheds.
The study showed how AI can extract interpretable spatial and temporal patterns that researchers can relate to physical processes, while keeping differences between climate scenarios visible. RAIN carries those principles forward—from regional surface-water analysis to integrated research on coupled surface-water and subsurface systems, using observations, physical models, advanced computing, and explicit uncertainty.
Related peer-reviewed work
Characterizing Drought Behavior in the Colorado River Basin Using Unsupervised Machine Learning
Talsma, Bennett & Vesselinov · Earth and Space Science · 2022Read the paper ↗
Figure 2 · From data to patterns
NMFk organizes 134 subwatershed time series into a matrix, extracts shared signals, and maps watersheds with similar behavior—the paper's clearest example of moving from raw model output to interpretable structure.
Figure 6 · Where patterns change
Spatial clusters compare historical and future dry-day behavior, plus the delta between them, under dry and wet scenarios. Blank panels indicate that NMFk found no acceptable solution; colors identify separate clusters within each panel and should not be compared across panels.
Figure 7 · When patterns change
Seasonal dry-day signals correspond to the spatial clusters in Figure 6. Individual lines represent subwatersheds and dashed lines mark cluster medians, making the timing and scenario dependence of those patterns visible.
Figures reproduced without modification from Talsma, Bennett & Vesselinov (2022), Earth and Space Science, doi:10.1029/2021EA002086. © 2022 The Authors, licensed under CC BY 4.0. Captions condensed for the web.
Research workflow
From raw signal to defensible insight.
RAIN frames AI as a scientific collaborator: one that can organize evidence, learn across scales, and keep the physical story visible.
01
Record
Observe, collect and align measurements across space, time, and source.
02
Represent
Create analysis-ready descriptions of the system and its context.
03
Reason
Combine AI, simulation, and domain constraints to test explanations.
04
Refine
Use new evidence and uncertainty to guide the next research cycle.
Explore EnviCloud
Explore related demos
Try interactive public demos that bring these ideas to life.
- Public demo
Information Theory Explorer
See how patterns and useful detail emerge in one- and two-dimensional data.
Open demo (opens in a new tab) - Public demo
WEF Nexus Decision Lab
Explore cost, carbon, and freshwater tradeoffs in a connected resource system.
Open demo (opens in a new tab) - Public demo
Model Diagnostics
Calibrate a model and identify influential, uncertain, or poorly constrained inputs.
Open demo (opens in a new tab)
Built for discovery
Let's make complex water systems easier to understand.
Connect with EnviTrace to discuss the RAIN project, scientific collaboration, or related AI and water-system research.
Start a conversation