Indian morphometry is published one watershed at a time. Each study picks its own elevation model and its own stream threshold.
So the numbers cannot be added up, ranked or regressed. They measure different things under the same name.
One elevation model, one channel network, one threshold, one projection — applied to all 212 catchments alike.
Every figure is printed beside the published figure it should match — including where the two disagree.
A uniform description of India's drainage, built so the numbers can be compared.
NADI delineates 212 named river catchments, nested four levels deep across 124 basins, from the HydroSHEDS v1 conditioned elevation model. For each one it computes the classical morphometric suite — drainage density, stream frequency, bifurcation ratio, hypsometric integral, relief ratio, ruggedness and the rest — using the same inputs and the same threshold throughout.
It is not a survey document and not a source of record. It derives; it does not observe. What it offers is that a figure for the Chambal and a figure for the Kaveri were produced the same way, and so can be set side by side.
Open a river. Or open all 212 at once.
Any river, in one click
Catchment area, discharge, river length and people — each printed beside the published figure it should match, with a measured band for how far to trust it.
Four nested levels
Basin, sub-basin, tributary, sub-tributary. Twenty parameters with their formulas, a hypsometric curve, a long profile and a 3-D terrain model.
All 212 side by side
Any measurement, every catchment, on one chart. The thing a one-watershed-per-paper literature has never been able to offer.
Which catchments come first
The compound-parameter ranking Indian watershed planning uses, set against the number of people inside the modelled flood footprint — on separate axes, deliberately.
Eight validation checks
Most of them against evidence from outside the atlas entirely. Where NADI disagrees with the published figure, both numbers are shown.
Every table, no account
Your own browser writes the files. Nothing is uploaded, which is why the whole thing still works with the wifi off.
Three things that were not known before this was built.
Not claims. Measurements, each reproducible from the published tables.
The variable nobody reports matters eight times more than the one everybody does.
Thirteen catchments were recomputed on the 30 m Copernicus model with the method held fixed. Drainage density rose by a median of 16 per cent. Moving the channel-initiation threshold by one decade moves the same quantity by a factor of three.
Resolution is stated in almost every morphometry paper. The threshold is stated in almost none. It is the one that decides the answer.
A correction the atlas published against itself.
An earlier release removed structures outside India using a bounding box, which cannot separate India from its neighbours. 441 barriers survived that filter and should not have. Every barrier was re-tested point-in-polygon against the national outline the atlas itself draws.
The count went from 7,543 to 7,102, of which 6,895 fall inside a traced catchment. The earlier figure was wrong, and it is corrected in public rather than quietly replaced.
Where India actually watches its groundwater.
40,100 Central Ground Water Board observation wells, carrying 1,653,432 water-level readings, located inside the traced catchments. 208 of the 212 have at least one.
The four with none are Tso Moriri, Zanskar, Dibang and Tons — all Himalayan or Ladakh. Dibang is the clearest case: fourteenth on the priority ranking, and not one observation well in it.
Including the places it failed.
A number without a test beside it is an assertion. These are the tests, with the results they actually gave.
| Check | Against | Result |
|---|---|---|
| Modelled discharge | 4,101,735 daily observations at 499 Central Water Commission stations | ρ 0.842 340 of 499 within ×2 |
| Catchment area | All 21 CWC basins, their own published areas | median 0.978 17 of 21 within 10% |
| Basin runoff | The CWC 2024 reassessment | median 0.874 19 of 20 within ×2 |
| Threshold scaling | Six thresholds, each catchment fitted separately | median r² 0.987 144 of 212 at ≥0.98 |
| Elevation model | 30 m Copernicus GLO-30, 13 catchments, method fixed | +16% median |
| Himalayan discharge | Nothing. No public daily record exists for the Ganga, Brahmaputra or Indus | unvalidated |
The last row is the one that matters most. Satellite altimetry cannot currently close that gap either: its own bias is characterised only where gauges already exist.
One person, no institution, and everything checkable.
NADI was built by Dr Pradeepika (Praddy) Kaushik, working independently. There is no lab behind it and no grant.
That is stated plainly because it should not matter to whether the numbers are right. Every figure here is checkable by a stranger, without permission and without anyone's letterhead: the computed tables and the full validation record are archived on Zenodo under CC BY 4.0, and the atlas writes every table out of your own browser.
The build code is held privately until the method is published, and is provided to editors and referees on request.
Click any river. See its entire story.
Nothing is held back and nothing is asked of you. No account, no sign-up, no email.