About Delhi Water Intelligence

How “Days of Water Left” Works

We compute three scenarios based on current reservoir storage, daily consumption, and inflow patterns:

Home / dashboard

Delhiowns no reservoir. Roughly 90% of its raw water arrives from other states under a stack of instruments - the 1994 five-state Yamuna MoU at Wazirabad, the 102-km Munak carrier from Haryana (about 70% of the lifeline), a BBMB-resolved share of Bhakra 240 km away, and Tehri's release riding the Upper Ganga Canal - with the city's own over-drafted groundwater making up the rest. So the headline is the supply chain itself, not a days-of-water runway: nine DJB treatment plants turn that inflow into about 960 MGD (~4,365 MLD) against an assessed requirement above 1,290 MGD.

The structural numbers are audit-grade rather than self-reported. The CAG's performance audit of the Delhi Jal Board (Report No. 3 of 2025, tabled in the Delhi Legislature on 23 March 2026) is the spine: non-revenue water at 51-53% of supply, a Rs 4,988 crore revenue impact, Rs 66,595 crore of outstanding loans and interest as of March 2022, and only 40% of the water produced ever billed. The source-wise split of raw water is DJB's own, published in the Delhi Economic Survey 2023-24 (Chapter 13) - the same chapter that states 52.35% of the city's water is "wasted or pilfered by tanker mafia". Where DJB's claimed availability and the CAG's audited production disagree, we show both rather than reconciling them.

What islive daily is rainfall: IMD's 0.25-degree gridded series for the NCT grid point (56 years, long-term mean 624.8 mm - within a millimetre of the figure IN-GRES reports for Delhi) extended by Open-Meteo reanalysis for the months IMD has not yet published, refreshed by the daily cron and superseded automatically as IMD catches up.

Groundwater

The choropleth is CGWB's Dynamic Ground Water Resource assessment at Delhi's own reporting unit - the 11 mapped districts, plus the non-spatial "Nazul Land" estate unit that appears in the tables but has no polygon. In the 2024-25 cycle four districts are Over-Exploited (New Delhi 123.2%, Shahdara 112.2%, North East 106.0%, South 103.4%) and the NCT as a whole draws 92.1% of what recharges - so the capital is pumping close to, and in places well past, what nature returns.

The per-district panel carries four assessment cycles (2021-22, 2022-23, 2023-24, 2024-25) labelled the way the source labels them. Two caveats are stated on the panel itself: the assessment only became annual in 2021-22, and before that Delhi was assessed by tehsil (~34 units) rather than by district, so older editions cannot honestly be stitched onto this series. The 2022-23 cycle has no mirrored dataset anywhere and was taken from IN-GRES directly.

Underneath the district choropleth sit 237 CGWB observation wells, each with a monthly depth-to-water series from 2015 to 2025. We had previously recorded that this network was unreachable because India-WRIS refuses connections from outside India. That was wrong, and worth stating plainly: only the map-tile host is geo-restricted, while the data endpoint is open. It merely behaved as though Delhi had no data, because a query that omits a district returns an empty result rather than every district.

The readings needed care before publication. The sign of a reading depends on which programme installed the well, so depth is resolved per station rather than by taking absolute values, which would have quietly turned two faulty sensors into plausible readings and erased the genuine cases where water sits above the sensor in floodplain wells. The result matches Delhi's known hydrogeology: about 2 m under Jagatpur on the Yamuna floodplain against about 68 m under Gadaipur on the ridge.

Per-ward depth interpolation remains deliberately off. A shaded surface between a floodplain well and a ridge well would invent intermediate depths no instrument ever recorded. Points are shown where they were measured, and the ward ranking uses the wells that are actually near a ward. Note also that the network stops reporting on 20 September 2025, the same month the Bhakra Beas Management Board's reservoir page froze.

Rivers

This page is Yamuna-basin scoped and labelled as such: the reach runs from the Hathnikund barrage in Haryana through the 22-km Wazirabad-to-Okhla city stretch that carries roughly 80% of the whole river's pollution load. Five channels are mapped - the Yamuna, the Western Yamuna Canal / Munak carrier that is the city's lifeline, the Hindon, and the Najafgarh and Sahibi courses, which are the same water: OpenStreetMap maps most of the Najafgarh drain as the Sahibi's engineered reach, so the two entries cross-reference each other rather than pretending to be separate rivers.

Water quality is Delhi's one genuinely high-cadence public feed, and it is monthly- DPCC samples eight Yamuna stations and about 39 drain points every month, which no other state pollution board publishes at that granularity. The station panel is derived from that monthly file so the two can never drift apart. The shape of the river's year is visible in it: water arrives at Palla meeting the bathing standard and leaves at Asgarpur with dissolved oxygen at nil.

Known gaps: the Barapullah and Shahdara drains are not mapped under those names in OpenStreetMap, and the DPCC reports are scanned PDFs, so extending the series is an OCR task rather than an API call. The 13 CETP monthly archives (2019-2024) are indexed with a transcribed sample but not yet bulk-extracted.

Flood risk

Delhi has no public flood model - no CFLOWS or iFLOWS equivalent - so unlike Chennai there are no modelled hazard zones or return-period extents to show, and we do not substitute something weaker while implying otherwise. What Delhi does have is a release-driven threshold structure with a 36-72 hour fuse: water released at Hathnikund reaches the Old Railway Bridge gauge two to three days later, where 204.50 m is warning, 205.33 m danger and 206.00 m evacuation, and the first flood warning goes out when Hathnikund discharge crosses about one lakh cusecs.

Seven events are documented from 1978 to 2025, sourced individually. Where the record carries gauge levels we cite levels (2023's 208.66 m, the highest ever recorded, and 1978's 207.49 m benchmark); where the Central Water Commission's own case study tabulates peak discharges rather than levels, we state discharges - the 1995 flood was the largest flow between 1978 and 2023. The chronic-waterlogging register carries Delhi's named perennial hotspots (Minto Bridge, Pul Prahladpur, Zakhira and the rest) with the official count hierarchy alongside: 448 points mapped from traffic-police data, 169 locations identified for 2025, 71 with a nodal officer. Those full lists are referenced in reporting but never published as data, so the register carries the named sites only.

Water bodies & restoration

The polygon base is OpenStreetMap - 1,845 water bodies totalling about 5,805 hectares, the largest of which is the 601-hectare remnant of the Najafgarh Jheel. That single polygon carries the city's sharpest loss statistic: the jheel spread over roughly 226 sq km before colonial-era drainage turned the Sahibi into the Najafgarh drain, and its wetland notification has been stuck between Delhi and Haryana since 2014.

Onto that we join the first Census of Water Bodies ("Jal Dharohar", enumerated 2022) - all 893 Delhi records with ownership, capacity, depth and khasra number. Only 234 match an OSM polygon (31 inside one, 203 within 150 m); the remaining 659 are recorded as unmatched rather than force-fitted, because the census enumerates johads and recharge pits that OSM has never mapped. Restoration priority currently scores the 12-body flagship register (the hauz-and-baoli chain plus the modern lakes) rather than all 1,845 polygons, so most wards show a count without a scored list - a documented limitation, not an omission.

My Ward

Wards are the 250 of the post-2022 unified MCD - the current delimitation, not the pre-merger 272 or the commonly-cited 270. Each ward joins the water bodies and census records inside it, the lost-water register, chronic waterlogging hotspots, mapped drain length, the nearest DPCC river station, its delimitation population, and its councillor with party and seat reservation from the December 2022 election (the party split validates against the published result: AAP 134, BJP 104, Congress 9, Independent 3).

Assembly and parliamentary representatives are not shown: the ward-to-constituency mapping exists in the ward geometry, but we have not ingested the assembly and parliamentary result sets, and a half-filled card is worse than an honest one. Sewerage is marked unavailable - DJB's network dataset was delisted from its host and no copy survives.

Each ward also carries its JJ bastis. DUSIB publishes the household roster and the coordinates as two separate PDFs, linked from the same page and numbered differently, so the two are joined on location name rather than serial: 642 of the 675 clusters fall inside an MCD ward, carrying 252,833 households across 142 wards. Where a cluster's name could not be matched to the roster with confidence, it keeps its location but carries no household count, so a ward's total is a floor and is labelled as one. The 33 clusters that sit in no ward are not an error: 21 are in the NDMC area and 11 in Delhi Cantonment, neither of which is part of the MCD.

Allocation Ledger & Commitments Register

Delhi is the strongest case on the platform for both surfaces. The Allocation Ledger tracks five instrument-governed arrangements - the 1994 MoU's 0.724 BCM/yr, the Munak carrier fixed at about 1,050 cusecs by the 2018 Standing Committee, Bhakra's share set meeting-by-meeting in BBMB minutes, Tehri's 300 cusecs, and unregulated groundwater - and grades each on how solid the paper is. The received side is mostly null on purpose: that is what the public record actually contains.

The Commitments Register holds eight dated promises with their original citations, and never overwrites a date. The lead entry is already overdue: the government's target to trap all 39 major drains by 30 June 2026 passed without a public completion statement. DPCC's monthly drain readings are the instrument that will confirm or refute it - a trapped drain shows up as "no flow".

Origins

A four-chapter long-read on how a city that engineered its own water for a thousand years - Anangpur's gravity dam, the Sultanate hauz, the stepwells, Shah Jahan's Nahar-i-Behisht - became one that waits for water from five other states. Every claim is drawn from the same sourced corpus as the data pages; paraphrased quotations are marked as such rather than presented as verbatim. Images are Wikimedia Commons (public domain or CC), with per-file provenance recorded alongside them in the repository.

How we classify river health

CPCB publishes two parallelriver-water-quality classification systems, and they don't always agree. Knowing which one we use - and why - matters for reading our river status badges honestly.

Our status badges ("dead", "severely degraded", "degraded", "stressed", "healthy") are computed from current readings via the Designated Best-Use thresholds- not from the PRS Priority list. We take the worst classification across a river's monitored stations and surface that as the river-level status.

Practical consequence: a river on CPCB's PRS Priority list (e.g. the Madurai-Manamadurai stretch of the Vaigai is Priority III) won't automatically render as "severely degraded" here. If the underlying NWMP readings show only Class C/D conditions, the badge reflects that. The PRS designation belongs in the river description as historical context, not as the live status.

Status thresholds follow CPCB's published Designated Best-Use criteria; readings are from CPCB NWMP annual River Water Quality reports.

Known Limitations

• Estimates are approximations. Actual water availability depends on factors not modeled (groundwater extraction, Krishna water transfer, distribution losses, industrial use).
• Utility-published feeds may occasionally be stale (weekends, holidays, portal outages). Each surface shows a freshness indicator, and the freshness registry names any feed that has stopped updating.
• Groundwater data from OpenCity may lag by months. The map always shows the most recent available period.
• Forecasts use ARIMAX (AutoARIMA with inflow/outflow as exogenous regressors) and work best with 2+ years of daily data.
• Risk scores are relative indicators for comparison between wards, not absolute measures of water safety.
• Satellite spread is a summary of surface water extent, not a direct measure of storage volume, water quality, or inflow source. A lake can look broad and still hold less usable water than expected.
• Reservoir catchment polygons are reviewed operational geometries for rainfall context, not official legal boundaries. This matters especially in Chennai's managed canal and transfer system.
• Current satellite context relies on optical Sentinel-2 observations. During persistently cloudy periods, some water bodies may temporarily lose this insight until a radar fallback is added.

Disclaimer

Open Source

Neer Vazhvu is fully open source. The code, data pipeline, and methodology are transparent and available on GitHub. Contributions, bug reports, and data corrections are welcome.

Support this project

Neer Vazhvu is free and open source. If you find it useful, consider supporting us on Patreon to help cover satellite data, hosting, and API costs.