TubigAI (AgwaSense): Hyperlocal Municipal Water-Rationing & Dam Deficit Navigator

Problem Statement

When El Niño strikes the Philippines, key reservoirs like Angat, Pantabangan, and Magat dams drop below critical minimum operating levels (e.g., Angat plunging below 180 meters), forcing NWRB and water concessionaires (Maynilad, Manila Water, and provincial water districts) to cut water allocations. This triggers rotating 8- to 16-hour rotational water interruptions affecting over 12 million Filipinos. Water schedules published on Facebook are notoriously inaccurate: high-elevation or end-of-pipe barangays receive trickles or zero pressure even when schedules claim water is on. Families store water haphazardly in open drums—breeding dengue mosquitoes—while hospitals, schools, and MSMEs struggle with emergency water truck deployments arriving blindly.

Proposed Solution

TubigAI is an AI-powered municipal water rationing optimization and citizen advisory platform designed for El Niño drought resilience.

1. For Water Districts & LGUs: It ingests real-time dam water elevations, reservoir outflow rates, and pipe pressure sensor feeds (or historical pressure logs) into a hydraulic graph neural network (GNN) and predictive model. It simulates water flow dynamics across elevation zones to generate equitable rationing schedules that prevent pipe cavitation, reduce non-revenue water (NRW) pipe bursts during re-pressurization, and prioritize critical infrastructure (public hospitals, evacuation centers).
2. For Citizens: Residents subscribe via free Smart SMS or Messenger. The AI delivers street-specific alerts in plain Tagalog/Bisaya: “Purok 3, may tubig mamayang 4:00 AM - 9:00 AM. Inaasahang mahina ang tulo bandang 8:00 AM. Mag-ipon nang may takip bago mag-7:00 AM.”
3. For Emergency Dispatch: An optimization engine routes LGU and concessionaire water delivery tankers dynamically to areas suffering zero pressure beyond their scheduled cutoff window.

Target Market

• Primary Beneficiaries: 12+ million residents in Metro Manila, Rizal, Cavite, Bulacan, and drought-hit provincial cities (e.g., Cebu, Zamboanga, Iloilo).
• Primary Buyers (B2G / B2B): Local Water Utilities Administration (LWUA), over 500 provincial water districts, Metro Manila water concessionaires (Maynilad, Manila Water), and City Disaster Risk Reduction and Management Offices (CDRRMOs).

How AI Is Used

• Hydraulic Flow & Pressure Prediction: Physics-informed machine learning (PINNs) and Graph Neural Networks model water pipe grids under varying pressure and elevation, predicting true street-level flow times rather than theoretical valve schedules.
• Dynamic Rationing Optimization: Constraint-satisfaction reinforcement learning balances dam depletion rates against community water needs, ensuring hospitals and high-density zones receive minimum survival quotas.
• Natural Language Localization: Generative AI converts complex hydraulic pressure schedules into personalized, conversational Taglish/Bisaya advisories delivered over SMS.
• Why it fails without AI: Standard static schedules assume uniform pipe pressure; in reality, gravity and topography cause uneven flow that only nonlinear predictive modeling can forecast accurately across thousands of pipe segments.

PLDT / Smart / ePLDT / Rokid Angle

• Smart Communications (SMS & Agri/Utility IoT): Smart’s nationwide SMS broadcast and two-way messaging API power the citizen notification engine. Smart IoT SIMs transmit pressure sensor telemetry from district water pump stations.
• ePLDT VITRO Data Centers & Cloud: ePLDT hosts the data-intensive hydraulic simulation engine and stores municipal infrastructure digital twin models with enterprise-grade security.
• PLDT Enterprise B2G Sales: Sold as a Smart Water Resilience Solution to water concessionaires and tier-1 LGU water utilities.

Demo Scenario (3 minutes for judges)

1. The Drought Crisis (0:00 - 0:45): The presenter shows Angat Dam’s live water elevation gauge dipping into the red zone (178m). The provincial water district dashboard triggers an emergency 40% allocation cut.
2. The AI Optimization (0:45 - 1:45): The screen shows TubigAI recalculating the water distribution plan for 15 barangays. Instead of blind blanket shutoffs, the AI shifts valve schedules to protect the provincial hospital zone and sequences water flow by elevation, saving 1.2 million liters from pipe burst pressure spikes.
3. The Citizen Experience (1:45 - 2:30): A simulated phone receives a Smart SMS alert: “Alert mula sa TubigAI: Dahil sa pagbaba ng dam, ang tubig sa Brgy. San Isidro ay bukas lamang mula 5:00 AM hanggang 10:00 AM. Mag-ipon sa mga oras na ito.” The resident checks an interactive low-bandwidth web map showing live pressure status on their exact street.
4. Tanker Triage (2:30 - 3:00): A live map displays crowdsourced reports of zero pressure; TubigAI automatically re-routes an LGU water tanker to a daycare center in a dead zone, completing the loop.

30-Hour Build Plan

Suggested stack: Python (FastAPI, NetworkX, NumPy, SciPy), Next.js, Tailwind CSS, Leaflet/Mapbox GL, Twilio/Semaphore SMS API (or mock SMS interface), Supabase (PostgreSQL).
Public data / APIs: PAGASA Dam Water Levels Portal, NWRB raw bulletins, OpenStreetMap road and elevation layers, Maynilad/Manila Water historical interruption advisories.

Risks & Privacy Notes

• Privacy (Clause 24 Compliance): Citizen subscribers only provide a mobile number and general barangay/purok location. No full names, house numbers, or identity documents are collected. Data is encrypted and automatically purged after drought emergency de-escalation.
• Operational Risk: Model suggestions must be reviewed by licensed water district engineers before valve actuators are physically opened or closed; the system acts as a decision-support co-pilot.

AI Use Disclosure (draft)

TubigAI utilizes Python-based constraint optimization models and Graph Neural Networks to forecast hydraulic pressure drops. Generative AI (LLMs) is used strictly for summarizing schedule tables into local-language SMS alerts. All core distribution models are validated against physical hydraulic flow constraints, and all public alerts undergo water district operator verification.

Rubric Scoring