Water Intelligence Platform

Every query has a cost. Every model has a thirst.

Fresh water covers less than 3% of Earth's surface — and less than 1% of that is accessible to humans. Yet as AI infrastructure has scaled into one of the fastest-growing industries on the planet, it has quietly become one of its most water-intensive. Training a single large language model can consume hundreds of thousands of liters of water for cooling. A data center running inference at scale draws as much water daily as a small city. And unlike energy — which can be generated from renewables — water consumed for cooling is largely lost to evaporation, removed from local watersheds permanently.

Meanwhile, freshwater systems are under unprecedented pressure. Droughts are intensifying. Aquifers are being drawn down faster than they recharge. Rivers that once ran year-round now run dry in summer. The communities and ecosystems that depend on these systems don't get a quarterly report from the hyperscalers drawing from the same water table.

This platform exists to make that connection visible.

By combining real hydrological measurements from U.S. Geological Survey stream gauges with transparent, deterministic modeling, the Water Intelligence Platform translates raw water data into terms that make the AI industry's water footprint tangible: not as abstract statistics, but as a live, station-by-station accounting of what our digital infrastructure costs the natural world. No black boxes. No obfuscation. Just data, math, and honesty.

Water is not an infinite resource. Neither is our right to use it without accountability.

---

A public-facing environmental transparency platform that makes real U.S. water data accessible and translates hydrological measurements into AI sustainability equivalents.

Live data from the U.S. Geological Survey. No mocked data. No paid AI APIs. Fully explainable.

---

---

Architecture

water-intelligence-platform/
├── backend/     NestJS API (TypeScript)
└── frontend/    React + MUI + Recharts (TypeScript)

Backend modules

Module	Responsibility
WaterModule	Fetch & normalize USGS OGC API data
AiImpactModule	Deterministic sustainability modeling
IntelligenceModule	Moving averages, anomaly detection, scoring
AnalyticsModule	In-memory observability (request tracking)

Primary endpoint

GET /intelligence?stationId=01646500

Other endpoints

GET /stations?state=VA
GET /analytics/summary
GET /analytics/public
GET /health

---

Local Development

Prerequisites

• Node.js 20+
• npm 9+

Backend

cd backend
cp .env.example .env
npm install
npm run start:dev
# API available at http://localhost:3001

Frontend

cd frontend
cp .env.example .env
# Edit REACT_APP_API_BASE_URL=http://localhost:3001
npm install
npm start
# UI available at http://localhost:3000

---

Production Deployment

Backend → Render

1. Push repository to GitHub.
2. Create a new Web Service on render.com.
3. Point it to the backend/ directory.
4. Set Build Command: npm install && npm run build
5. Set Start Command: npm run start:prod
6. Add environment variables from .env.example.

Frontend → GitHub Pages

1. In frontend/package.json, set "homepage" to your GitHub Pages URL:

   "homepage": "https://yourusername.github.io/water-intelligence-platform"

2. Update REACT_APP_API_BASE_URL to your Render backend URL.
3. Deploy:

   cd frontend
   npm run deploy

---

Docker

Backend only

cd backend
docker build -t water-intelligence-api .
docker run -p 3001:3001 --env-file .env water-intelligence-api

Full stack (docker-compose)

# From repo root
docker-compose up

---

Environment Variables

Backend (backend/.env)

Variable	Default	Description
PORT	3001	API port
CORS_ORIGIN	http://localhost:3000	Frontend origin
USGS_BASE_URL	https://api.waterdata.usgs.gov/ogcapi/v0	USGS API base
CACHE_TTL_SECONDS	300	In-memory cache TTL
WATER_PER_KWH	1.8	Liters of water per kWh (data center cooling)
KWH_PER_AI_INFERENCE	0.001	kWh per AI inference
KWH_PER_GPU_TRAINING_HOUR	1.2	kWh per GPU training hour

Frontend (frontend/.env)

Variable	Default	Description
REACT_APP_API_BASE_URL	http://localhost:3001	Backend API URL

---

USGS Station IDs — Quick Reference

Station	ID	Location
Potomac at Point of Rocks	01638500	MD/VA
Mississippi at St. Louis	07010000	MO
Colorado at Lees Ferry	09380000	AZ
Columbia at The Dalles	14105700	OR
Rio Grande at El Paso	08364000	TX

---

Intelligence Engine — Algorithm Reference

Moving Averages

Simple arithmetic mean over N most recent daily values. If fewer than N values exist, all available values are used.

Volatility Index

volatility = stdDev / mean (coefficient of variation) Capped at 2.0 for display. Values > 0.5 indicate high volatility.

Anomaly Detection (rule-based)

if (today / MA7) >= 2.0  → severity: "severe"
if (today / MA7) >= 1.5  → severity: "moderate"
else                      → severity: "none"

Sustainability Score (0–100)

Starts at 100; deductions applied:

• Volatility > 0.5: −20
• Severe anomaly: −30
• Moderate anomaly: −15
• Flow < 10th percentile of recent series: −15

---

Future Roadmap

• PostgreSQL persistence layer for analytics
• Time-series caching with Redis
• User authentication for personalized station watchlists
• WebSocket real-time updates
• Export to CSV / GeoJSON
• Mobile-first PWA

---

Data Attribution

Water data provided by the U.S. Geological Survey Water Data for the Nation.

• API: https://api.waterdata.usgs.gov/ogcapi/v0
• Website: https://waterdata.usgs.gov

This platform is not affiliated with or endorsed by USGS.

License

MIT