A seasonal climate forecast and a narrow-therapeutic-index drug level pose the same problem: a number that changes the odds, routinely misread as a number that settles the outcome. This project applies clinical epidemiological discipline to that problem — and derives, rather than assumes, the quantities involved.
One frames the mechanism. One computes the risk. Both are open; the tool is intended for professional use.
Ten structural capabilities a coupled ocean–atmosphere anomaly must acquire before it produces remote climatic harm — each with its marker, its failure mode, and its clinical-pathology analogue. Modelled on the hallmarks framework used in cancer and immunology.
Read the doctrine → Professional useEnter the ocean state and the decision you face. Returns pre-test and post-test risk, likelihood ratio, ARI, NNE, attributable fractions and an action threshold — each explained against its clinical analogue, with confidence intervals throughout.
Open the tool →In 1997 the strongest El Niño of the twentieth century coincided with an Indian monsoon 2% above normal. In 2002 a merely moderate event produced a national drought, with July rainfall 49% below normal — the worst in the recorded observational history.
Neither outcome refutes the risk factor. Both refute reading it as a prophecy. Medicine solved this communication problem decades ago and built a vocabulary for it: absolute risk increase, number needed to treat, likelihood ratio, attributable fraction, decision threshold. That vocabulary transfers cleanly, and almost nobody has transferred it.
A risk factor changes the odds; it does not determine the outcome. Everything else here is an attempt to make that statement operational rather than merely true — to turn it into numbers a decision-maker can act on, with the uncertainty attached rather than stripped.
The core likelihood ratio is computed from primary sources, not assumed. The full derivation is public.
| Element | Source |
|---|---|
| Exposure classification | NOAA CPC Oceanic Niño Index, ERSSTv5 — monsoon-season ONI, 1950–2015 |
| Outcome census | All-India JJAS rainfall below −10% of mean — the IMD drought threshold |
| Derived result | LR 8.06 (95% CI 2.11–30.78), Fisher exact p = 0.0026 |
| Robustness | Five exposure definitions, leave-one-year-out, era split — all reported |
| Recorded falsification | The intensity ladder used in v1.0 was refuted by the data and removed |
That last row matters more than the rest. A first version of the risk tool graded risk by El Niño strength, on the assumption that stronger events carry higher drought risk. The census refuted it. The control was deleted rather than retained with a caveat, and the refutation is documented in full on the methods page.
Nothing here issues a seasonal forecast. Outputs are population-level rates across many seasons, never a prediction for a specific season, district, or holding. The India Meteorological Department is the operational authority for Indian monsoon forecasting. Full limitations are stated on the scope page and inside the tool itself.