Measurement Framework for Clean Cities India: Data Architecture for Tracking Pollution Reduction
Clean Cities India · 2026-07-13
The premise
Clean Cities India has already mapped the pollution landscape across ten cities, catalogued fifteen proven interventions, and identified five cities where severity meets civic readiness for first campaigns. This publication answers the next question: how will we know if any of it is working?
Measurement is not a reporting exercise. It is the feedback loop that tells a municipal commissioner whether a ward-level waste segregation pilot deserves scale-up, tells a state pollution board whether an industrial cluster's effluent standards are holding, and tells a resident whether the air her child breathes this winter is better than last year's. The framework below is designed to be honest about uncertainty, resistant to gaming, and useful to the people who must act on the data.
What the evidence supports
Air: the NCAP/PRANA architecture is the backbone; city action plans are the implementation layer
India's National Clean Air Programme (NCAP) targets 131 non-attainment and million-plus cities with a 40 % reduction in PM₁₀ by 2026 relative to the 2017 baseline. The PRANA portal (prana.cpcb.gov.in) is the designated tracking platform — it ingests city action plans, physical and financial progress, and air quality data from the National Air Quality Monitoring Programme (NAMP) and the growing continuous monitoring network.
What NAMP actually measures today
| Parameter | Averaging period | Frequency (manual stations) | NAAQS 2009 annual limit (residential) |
|---|---|---|---|
| PM₂.₅ | 24-hour | 2 days/week (104/year) | 40 µg/m³ |
| PM₁₀ | 24-hour | 2 days/week (104/year) | 60 µg/m³ |
| SO₂ | 24-hour / 1-hour | 2 days/week | 50 µg/m³ / 200 µg/m³ |
| NO₂ | 24-hour / 1-hour | 2 days/week | 40 µg/m³ / 80 µg/m³ |
| CO | 8-hour / 1-hour | 2 days/week | 2 mg/m³ / 4 mg/m³ |
| O₃ | 8-hour / 1-hour | 2 days/week | 100 µg/m³ / 180 µg/m³ |
| NH₃ | annual / 24-hour | 2 days/week | 100 µg/m³ / 400 µg/m³ |
| Pb, Benzene, BaP, As, Ni | annual | periodic | various |
Source: CPCB National Ambient Air Quality Standards (2009); NAMP operational guidelines.
What the continuous network adds — As of mid-2025, ~500 continuous ambient air quality monitoring stations (CAAQMS) report hourly PM₂.₅, PM₁₀, NO₂, SO₂, CO, O₃, NH₃, and meteorological data to the CPCB central server. This is the high-resolution layer; manual NAMP stations remain the statutory compliance backbone.
The gap this framework addresses — NCAP tracks city-level PM₁₀ trends. It does not systematically track:
- Source-sector contributions (transport, industry, dust, biomass, waste burning) at ward level
- Intervention-specific exposure reduction (e.g., did the low-emission zone in Zone 3 lower PM₂.₅ at the nearest school monitor?)
- Co-benefits: ozone formation when NOₓ drops without VOC control; PM₂.₅/PM₁₀ ratio shifts
Water: NWMP provides the river/lake backbone; urban water bodies and groundwater need a parallel track
The CPCB National Water Quality Monitoring Programme (NWMP) operates >4,000 stations across 720+ rivers, 95+ lakes, and groundwater, sampled monthly (groundwater: pre- and post-monsoon). The core parameter set:
| Core parameter | Significance | Class A limit (drinking, no treatment) |
|---|---|---|
| Dissolved Oxygen (DO) | Aquatic life, self-purification | ≥ 6 mg/L |
| Biochemical Oxygen Demand (BOD) | Organic pollution load | ≤ 2 mg/L |
| pH | Chemical/biological processes | 6.5–8.5 |
| Total coliform | Faecal contamination | 50 MPN/100 mL |
| Faecal coliform | Recent faecal pollution | — |
| Conductivity | Dissolved solids proxy | — |
| Nitrate (NO₃⁻) | Agricultural/urban runoff | 20 mg/L (BIS 10500) |
| Phosphate (PO₄³⁻) | Eutrophication driver | — |
Source: CPCB NWMP; Designated Best-Use classification (Classes A–E).
The Water Quality Index (WQI) aggregates these into a 0–100 score (higher = better). CPCB publishes station-level WQI; it does not currently roll up to a city-level urban water health score that includes drains, lakes, and groundwater — the domain where municipal action lives.
Waste: Swachh Survekshan is the only national periodic scorecard; it was not built for monthly tracking
Swachh Survekshan 2024–25 evaluates ~4,500 cities across ten thematic sections (Visible Cleanliness, Segregation, Collection & Transportation, Solid Waste Processing, Used Water Management, etc.) with a total score of 12,500 marks. Key indicators relevant to ongoing measurement:
| Indicator | What it captures | Frequency in Survekshan |
|---|---|---|
| Door-to-door collection coverage | % wards/households served | Annual |
| Source segregation rate | % waste segregated at source (wet/dry/domestic hazardous) | Annual |
| Waste processing capacity vs. generation | TPD processed / TPD generated | Annual |
| Landfill diversion rate | % waste not going to landfill/dumpsite | Annual |
| Dumpsite remediation | Area reclaimed, legacy waste processed | Annual |
| Used water treatment | % sewage treated, reuse volume | Annual |
| Citizen feedback / grievance redressal | App-based complaints, resolution time | Annual |
Source: Swachh Survekshan 2024–25 toolkit (MoHUA).
The gap — Survekshan is a ranking exercise, not a management dashboard. A commissioner needs to know this month whether segregation in Ward 17 dipped below 60 %, not where the city ranked nationally last year.
What the evidence does not support
| Common assumption | What the data says |
|---|---|
| "More monitoring stations = better management" | Station density in Indian cities correlates weakly with pollution reduction. Delhi has ~40 CAAQMS; Patna has 3. Both exceed standards. What matters is whether data triggers specific action (source apportionment → targeted intervention → verification). |
| "AQI alone tells the public what they need" | AQI is a health-communication index, not a management metric. Two cities with identical AQI 150 can have totally different source profiles and intervention needs. |
| "WQI at the river intake = city water health" | Intake WQI misses drain outfalls, urban lakes, groundwater contamination, and intermittent supply quality. |
| "Survekshan rank improvement = waste problem solved" | Cities can improve rank by documenting processes (IEC, bylaws) without changing material flows. Processing capacity on paper ≠ actual diversion. |
| "One dashboard can serve all audiences" | A commissioner needs ward-level daily segregation data; a resident needs "is my tap water safe today?"; a researcher needs raw hourly PM₂.₅. One interface cannot serve all three without distortion. |
The least-harm measurement architecture
The framework has three layers. Each layer has a distinct decision audience, refresh cadence, and accountability mechanism.
Layer 1 — City Management Dashboard (monthly, operational)
Audience: Municipal commissioner, city pollution control board, ward engineers, campaign coordinators.
Air metrics (per city, per ward where monitors exist)
| Metric | Source | Frequency | Red flag threshold |
|---|---|---|---|
| PM₂.₅ monthly mean (µg/m³) | CAAQMS + NAMP fusion | Monthly | > 60 (1.5× annual std) or ↑ >10 % MoM |
| PM₁₀ monthly mean | CAAQMS + NAMP | Monthly | > 100 or ↑ >10 % MoM |
| NO₂ monthly mean | CAAQMS | Monthly | > 40 or ↑ >15 % MoM |
| Monitor uptime | CPCB/PRANA | Daily | Monthly |
| Source-apportionment refresh | IIT/NEERI/TERI studies | Annual (target) | > 24 months since last study |
Water metrics (per city)
| Metric | Source | Frequency | Red flag threshold |
|---|---|---|---|
| WQI at each NWMP station in city limits | NWMP monthly | Monthly | < 50 (poor) or ↓ >10 pts MoM |
| BOD at major drain outfalls (mg/L) | SPCB monthly / IoT sensors | Monthly | > 30 (Class E breach) |
| Faecal coliform at supply intake (MPN/100 mL) | Jal Board / PHED lab | Monthly | > 50 (Class A breach) |
| Groundwater quality index (selected wards) | CGWB / SPCB | Quarterly | Declining trend 2+ quarters |
| Sewage treatment capacity utilization | STP SCADA / ULB | Monthly | < 80 % utilization |
Waste metrics (per city, per ward)
| Metric | Source | Frequency | Red flag threshold |
|---|---|---|---|
| Door-to-door collection coverage | ULB MIS / vehicle GPS | Monthly | < 95 % wards |
| Source segregation rate (wet/dry) | ULB MIS / RWAs sample audits | Monthly | < 60 % or ↓ >5 pts MoM |
| Wet waste processing rate | Composting/biogas plant logs | Monthly | < 70 % of collected wet waste |
| Dry waste recycling rate | MRF throughput / recyclers | Monthly | < 30 % of collected dry waste |
| Legacy waste remediated (cumulative %) | Dumpsite project reports | Quarterly | < 10 % annual progress |
| Grievance resolution time (SWM) | Swachhata App / 1969 | Weekly | > 48 hrs median |
Accountability mechanism: Each metric has a named data owner (officer + contact) and a public escalation path — if a red flag persists for two consecutive months, the dashboard auto-generates a notice to the municipal commissioner and the state pollution board chair, copied to the Clean Cities India coalition contact in that city.
Layer 2 — Campaign Verification Tracker (per intervention, 3/6/12-month gates)
Audience: Campaign leads (Clean Air Ward, Zero Burning Neighbourhoods), funders, coalition partners.
Each campaign in the Coalition Map publication has defined success metrics. The tracker records baseline → Month 3 → Month 6 → Month 12 for:
| Campaign | Primary metric | Verification method |
|---|---|---|
| Clean Air Ward (50 wards) | PM₂.₅ reduction at ward monitor vs. control wards | CAAQMS/low-cost sensor network; difference-in-differences |
| Zero Burning Neighbourhoods (200 nbhd) | % reduction in open burning events (satellite + citizen reports) | VIIRS fire counts + Swachhata App reports; ground truthing sample |
| Drain Interception Pilot (5 cities) | BOD reduction at outfall vs. baseline | Monthly SPCB grab samples + continuous DO/BOD sensors |
| Ward Composting Hubs (100 wards) | Wet waste diverted to compost (tonnes/month) | Weighbridge logs at hub; third-party audit quarterly |
| Construction Dust Control Zones | PM₁₀ at boundary monitors vs. non-zone | CAAQMS + mobile monitoring |
Least-harm safeguard: No campaign metric triggers penalties. Red flags trigger peer review (coalition partner + independent expert) and course correction, not sanction. The goal is learning, not compliance theatre.
Layer 3 — Public Accountability Dashboard (quarterly public release)
Audience: Residents, media, researchers, elected representatives, courts.
Design principles:
- One page per city. No login. Mobile-first. Hindi/English/local language toggle.
- Each metric shows: current value, trend (sparkline), benchmark (standard or peer median), data source, last updated.
- "What this means for you" — one plain-language sentence per metric (e.g., "PM₂.₅ at 55 µg/m³ means a child with asthma in this ward has ~20 % higher risk of an attack this month").
- Downloadable CSV for every chart. API endpoint for researchers.
- "Data quality flag" beside each metric: 🟢 audited / 🟡 provisional / 🔴 gap (with explanation).
City page structure:
- Air — PM₂.₅, PM₁₀, NO₂ monthly means; monitor map; source-apportionment summary (latest study year).
- Water — WQI at river/lake stations; drain outfall BOD trend; supply intake faecal coliform; groundwater trend.
- Waste — Segregation rate, processing rate, landfill diversion, grievance resolution — all ward-level where available.
- Campaigns — Clean Air Ward / Zero Burning progress bars; verification status (🟢 verified, 🟡 self-reported, 🔴 not started).
- Institutional health — Monitor uptime, STP utilization, SWM budget utilization %, vacancies in environment cell.
Gatekeeping: The public dashboard pulls only from Layer 1 (validated monthly) and Layer 2 (verified at gates). No raw or provisional data surfaces publicly without a quality flag.
Data architecture: sources, pipes, and ownership
┌─────────────────────────────────────────────────────────────────────────────┐
│ DATA SOURCES (authoritative) │
├──────────────────┬──────────────────┬──────────────────┬────────────────────┤
│ Domain │ Primary source │ Access mode │ Owner / contact │
├──────────────────┼──────────────────┼──────────────────┼────────────────────┤
│ Air (continuous) │ CPCB CAAQMS │ API (PRANA) │ CPCB Air Lab │
│ Air (manual) │ NAMP / SPCB │ Monthly CSV/FTP │ SPCB nodal officer │
│ Water (rivers) │ NWMP │ Monthly CSV │ CPCB WQM Division │
│ Water (drains) │ SPCB / ULB │ Monthly / IoT │ SPCB / City Engg. │
│ Water (supply) │ Jal Board / PHED │ Monthly lab │ MD, Jal Board │
│ Groundwater │ CGWB / SPCB │ Quarterly │ CGWB regional dir. │
│ Waste (ops) │ ULB MIS / GPS │ Daily API / CSV │ Commissioner / CE │
│ Waste (processing)│ Plant SCADA │ Daily API │ Plant operator │
│ Grievances │ Swachhata App │ API │ MoHUA / ULB │
│ Satellite (burning)│ VIIRS / Sentinel│ Daily / weekly │ NRSC / CREA │
└──────────────────┴──────────────────┴──────────────────┴────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────────────────────┐
│ INGESTION & VALIDATION LAYER │
│ • Scheduled ETL (Airflow / Prefect) — daily for APIs, monthly for CSVs │
│ • Schema validation (Great Expectations) — reject rows failing ranges │
│ • Uptime monitoring — alert if source silent > 48 hrs (air) / 14 days (water)│
│ • Deduplication — station merges, parameter harmonization (µg/m³ vs mg/m³) │
│ • Versioned warehouse (PostgreSQL + TimescaleDB) — full audit trail │
└─────────────────────────────────────────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────────────────────┐
│ COMPUTATION LAYER (derived metrics) │
│ • Monthly rollups (mean, percentile, completeness) per station/ward/city │
│ • WQI calculation (CPCB weighted arithmetic method) per station │
│ • Campaign diff-in-diff (ward vs. synthetic control) │
│ • Trend detection (Mann-Kendall, 12-month window) │
│ • Red-flag engine (threshold + MoM change + persistence) │
└─────────────────────────────────────────────────────────────────────────────┘
│
┌─────────────────┼─────────────────┐
▼ ▼ ▼
┌────────────────┐ ┌────────────────┐ ┌────────────────┐
│ Layer 1 API │ │ Layer 2 API │ │ Layer 3 API │
│ (auth, role- │ │ (campaign │ │ (public, │
│ based) │ │ owners) │ │ anonymous) │
└────────────────┘ └────────────────┘ └────────────────┘
Ownership principle: Every table, every API endpoint, every dashboard tile has a data steward (name, designation, phone, email) recorded in the metadata catalogue. If a number looks wrong, the user knows exactly who to call.
What I'm uncertain about
| Uncertainty | Why it matters | How we reduce it |
|---|---|---|
| Low-cost sensor calibration drift | Ward-level air tracking may rely on ₹15k–₹50k sensors; drift can mimic false trends. | Co-location protocol with reference monitors (monthly); open calibration logs; flag data > 90 days since co-location. |
| Drain outfall monitoring coverage | Most cities monitor < 20 % of major outfalls monthly. | Prioritize top 10 outfalls by flow × BOD load; fund IoT BOD/DO sondes at those 10. |
| Waste weighbridge data integrity | Processing rates depend on weighbridge logs; tampering risk is real. | Independent quarterly audit (weighbridge certification + random truck checks); publish audit summary. |
| Groundwater representativeness | CGWB wells are sparse in dense urban cores; may miss contamination plumes. | Supplement with ULB-drilled shallow piezometers in industrial/landfill buffer zones. |
| Source apportionment currency | Most cities have one study (2018–2022); source profiles change with fuel, fleet, industry. | Budget for biennial receptor-model updates (PMF/CMB) in all 131 NCAP cities; tie to NCAP fund release. |
| Citizen report verification | Swachhata App burning reports can be duplicate, false, or politicized. | De-dup algorithm + ward volunteer verification sample (10 %); publish verification rate. |
| Dashboard literacy | A "red flag" may be misread as "city failed" rather than "this metric needs attention." | Co-design dashboard language with RWAs, journalists, and a behavioural science partner; A/B test comprehension. |
Next steps (this framework's own gates)
- Month 1–2: Sign data-sharing MoUs with CPCB (PRANA API), 5 pilot city ULBs, and 5 SPCBs. Appoint data stewards.
- Month 2–3: Build ingestion pipelines for the 10 pilot cities (Delhi, Patna, Pune, Lucknow, Ghaziabad + 5 from Coalition Map). Validate against last 24 months of historical data.
- Month 3: Launch Layer 1 (internal) for pilot cities. Run red-flag engine shadow mode — no alerts, just log precision/recall.
- Month 4: First Campaign Verification Tracker entry (Clean Air Ward baseline + Month 3).
- Month 6: Public Layer 3 dashboard live for pilot cities. Independent usability audit.
- Month 12: Framework review — add/remove metrics, adjust thresholds, publish v1.1.
"A measurement system that cannot be read by the people it affects is not a measurement system — it is a filing cabinet. This framework is built to be read, questioned, and improved by the ward engineer, the resident, and the commissioner alike."
Submitted for review at the gate. The data architecture is only as good as the institutions that feed it; the framework names those institutions and the escalation paths when they fall silent.