Ranks Bhopal’s seven ghats on a six-factor risk score for the BMC immersion planning officer, linking every data point to its MPPCB report page.
The problem
Ahead of Durga Visarjan on 20 October 2026, Bhopal Municipal Corporation (BMC) officers must allocate limited aeration equipment, emergency water testing kits, and cleanup crews across 7 idol-immersion ghats. Today, these high-stakes environmental decisions rely solely on memory and intuition because 3 years of Madhya Pradesh Pollution Control Board (MPPCB) water quality data (1,524 test readings from 2023–2025) remain buried in dense 90–120 page government PDFs that nobody cross-references against legal standards.
The hidden reality is severe: at Bairagarh (Sant Hirdaram) kund, dissolved oxygen crashed to 0 mg/l four separate times across 2023 and 2025 (supporting zero aquatic life against the CPCB Class B bathing minimum of 5.0 mg/l). Across all Bhopal ghat readings, 42 of 116 DO readings violate bathing criteria, and 100 of 112 BOD readings exceed legal limits. Furthermore, official reports omit units, record asymmetric "BDL" (Below Detection Limit) values, and omit entire sampling cycles without explanation. Officers had no tool to rapidly identify the most vulnerable ghats before immersion day.
How it works
We built Ghat Watch, a decision-support system engineered specifically for the BMC Immersion Planning Officer:
6-Factor Audited Risk Ranking (0–100): Ranks all 7 Bhopal ghats by combining Dissolved Oxygen failure rate (35%), BOD failure rate (25%), immersion impact drop (15%), failure to recover post-immersion (10%), worst reading severity (10%), and pH violation rate (5%), with a 1.5× weight applied to Durga Puja cycles. Audit-Grade Traceability: Every data point, baseline, and failure links directly to its primary source MPPCB PDF report and exact page (#page=N), enabling instant administrative verification. Scientific Honesty & Missing Data Integrity: Adheres strictly to water-quality science: zero unsubstantiated health claims, asymmetric BDL interpretation (BDL BOD passes organic criteria; BDL DO flags severe hypoxia), and visually distinct hatched warning cells for unmonitored phases (e.g., missing 2025 Durga tables at Bairagarh & Malikhedi). Sensitivity Stress-Testing: Allows officers to verify rank stability under 3 alternative weighting scenarios (DO Priority, Impact & Recovery Priority, Severity Focus). Forwardable 1-Page Bilingual Brief (English & Hindi): Generates an A4 print-ready executive memo with verified administrative Devanagari terminology for immediate WhatsApp/email forwarding to municipal commissioners and ground teams.
How Claude was used
We leveraged Claude and Claude Code as an end-to-end systems and domain-engineering partner throughout the hackathon:
Data Auditing & Domain Logic: Interpreted 1,524 rows of MPPCB data, established asymmetric BDL handling (distinguishing low organic load from catastrophic hypoxia), and handled missing table anomalies. Formula Formulation & Verification: Formulated the transparent 0–100 composite risk scoring engine and verified its mathematical stability across sensitivity schemes. Full-Stack Implementation: Architected and wrote the modular FastAPI backend, CPCB Class B compliance evaluators, and the responsive Next.js frontend with accessible visual design and hatched missing-state indicators. Bilingual Localization: Crafted and verified culturally and administratively accurate Hindi Devanagari translations for municipal officers without robotic transliteration errors.