The Ganga treaty's sharing formula, set beside thirty years of modelled flow at Farakka.
The 1996 India-Bangladesh treaty shares the lean-season Ganga at Farakka by 10-day periods, for a thirty-year term that runs from 12 December 1996 to 12 December 2026. Sandhi encodes the treaty's text and formula as data and shows how the formula plays out against modelled river flow for every lean season since 1997, with each source cited and each limit stated.
Article XII sets a thirty-year term from signature, renewable by mutual consent, which runs to 12 December 2026. Renewal has been discussed in public by both governments. India's Ministry of External Affairs told the Lok Sabha on 13 February 2026 that discussions for renewal were yet to commence (MEA). Bangladesh's water resources minister said on 18 August 2026 that Bangladesh would seek a guarantee clause in a renewed treaty (AFP via Deccan Chronicle).
The Joint Committee's daily flow records at Farakka and Hardinge Bridge are not published as data. We found no public tool that sets the treaty's own formula and schedule against flows across all thirty lean seasons, in one place, for readers in both countries.
- Encodes the treaty's articles, the Annexure I formula and the Annexure II schedule as data, from two published copies compared word by word. Where the copies differ, both readings are shown.
- Reproduces all 30 share values printed in Annexure II from the formula, in Python and in the browser.
- Turns modelled daily flow at Farakka into the treaty's fifteen 10-day periods for every lean season since 1997, beside the 1949 to 1988 schedule.
- Applies the formula to those flows to show the split it would imply, and counts periods under 75,000, 70,000 and 50,000 cusecs, the last being the level at which Article II (iii) calls for consultations.
- Tests lean-season flow for a trend and states the 95% interval, which for 1997 to 2026 includes zero.
- Shows the outlook for the coming lean season from the 51-run GloFAS forecast, a dated and sourced timeline, and a short Teesta panel: no treaty exists there, so flows are shown as context only.
| Availability at Farakka | Share of India | Share of Bangladesh |
|---|---|---|
| 70,000 cusecs or less | 50% | 50% |
| 70,000 to 75,000 cusecs | Balance of flow | 35,000 cusecs |
| 75,000 cusecs or more | 40,000 cusecs | Balance of flow |
Summary of Annexure I. Each side is also guaranteed 35,000 cusecs in alternate three 10-day periods from 11 March to 10 May, as marked in Annexure II. The verbatim text, with both copies' readings, is on the console's treaty page.
- Encode the textRead both published copies in full, record every difference, and write the formula as code that must reproduce Annexure II.
- Find the riverUse the GloFAS cell on the Ganga's main channel, 14 km below Farakka Barrage, checked against its 48 neighbours.
- Build the periodsAverage daily flow over each treaty period, 1997 to today, and convert m³/s to cusecs (× 35.3147).
- Apply and compareRun the formula on each period, set it beside the schedule, rank seasons, and test for a trend.
- PublishConsole pages, a methods page and CSV, each number with its source, retrieval time and caveats.
Formulas, thresholds, the comparison with the schedule and the known failure modes are in the methods note and on the console's methods page.
| Source | What | Licence | Status |
|---|---|---|---|
| GloFAS v4 via Open-Meteo Flood API | Modelled daily discharge from 1997 at the Farakka and Teesta cells; 51-run forecast | CC BY 4.0; Copernicus EMS | Live |
| Treaty text, two copies | FAOLEX (FAO) copy and the Joint Rivers Commission, Bangladesh copy | Official text | Encoded |
| Official statements and news reports from both countries | Dated statements and meetings for the timeline, each checked against its source | Cited | Cited |
| Joint Committee daily records | Observed flows at Farakka and Hardinge Bridge | Governments of India and Bangladesh | Wanted |
Nothing in Sandhi is simulated. All flows are model output, labelled as such wherever they appear.
- Researchers and journalists in both countries: the same numbers, the same text and the same caveats, whichever side of the border they write from.
- Officials and their technical staff: a quick way to see how the formula behaves in wet and dry seasons, and which periods fall under the treaty's thresholds.
- Basin and disaster authorities: lean-season context for Farakka, and the outlook for the coming season.
- Neutral between the two countries. No finding about compliance, no estimate of releases, no characterisation of either government's position.
- Model flows are never presented as measurements.
- Statements on the timeline are attributed and summarised without adjectives; both treaty readings are shown where the copies differ.
- Every number carries its source and retrieval time; source PDFs are fingerprinted with SHA-256.
- Rivers and records only. Sandhi holds no data about people.
- All flows are modelled. Across 1997 to 2026 the model's median is 66% of the schedule from 11 January to 20 February and 135% of it from 21 April to 31 May, so absolute comparisons carry model error.
- The schedule is a 1949 to 1988 average; a gap between it and the model mixes model error with real change, and Sandhi cannot separate the two.
- The model does not divert water into the Feeder Canal, and may under-represent upstream withdrawals and groundwater use.
- Seasonal forecast skill for dry-season flow at Farakka has not been measured.
- The free Open-Meteo API is for non-commercial use. A deployment needs a subscription or a self-hosted copy.
Work with us to put official gauge records beside the model under a data agreement, or to add other rivers with published sharing schedules. contact@trinetra.in