Do Colombia's regions offset each other during El Niño?

No. Measured on IDEAM's series (2003–2026, 33 departments), water anomalies move nearly in phase: median correlation +0.12, and only 0.2% of department pairs are strongly anticorrelated. When ENSO dries one region, it tends to dry its neighbours too.

So a water-transfer network is useless?

As the core of the solution, no: the scenario it would exploit —one region dry while its neighbour is wet— happens only ~7% of months. What does the work is temporal storage (holding the wet pulse for the slow deficit); transfer links are few and short.

Where does the data come from?

From IDEAM's official series (2003–2026, all 33 departments), public domain, via its official connector. Water quality uses the IRCA index (INS/MinSalud). The computation is reproducible and visible at agua.xiliux.com.

Colombia's water and El Niño: I measured the design's bet, and the data refuted it

The design's bet, and why it had to be measured

Colombia swings between drought (El Niño) and flooding (La Niña). The intuitive idea for a national water network is a grid that moves water from where there is too much to where there is too little: if ENSO dries some regions while flooding others —spatial anticorrelation— moving water between them cancels the blow.

That is a hypothesis, not a fact. And before proposing billions in infrastructure on top of it, you measure it.

What I measured

I built a regional water-balance simulator on IDEAM's real series (2003–2026, all 33 departments), with its official connector, and computed the correlation of water anomalies between departments.

The evidence refutes the bet. The spatial anticorrelation is weak: the median anomaly correlation is +0.12 —regions move nearly in phase, not against each other. Only 0.2% of department pairs are strongly anticorrelated. The scenario a grid could exploit —one region dry while its neighbour is wet— happens only ~7% of months. When ENSO dries one region, it tends to dry its neighbours too: the country dries together and floods together.

The consequence for the design

If regions do not offset each other in space, the transfer grid is not the heart of the solution: it is an expensive ornament. What works is holding the wet pulse for the slow deficit — temporal storage: restored páramos and wetlands, reservoirs, buffers for both extremes. Transfer links are worth few and short; storage does the work.

Measuring changed the product: from "move water between regions" to "store across time". That is the difference between proposing on a hunch and proposing on data.

The number that does sting

The same work carries the quality map: by the IRCA index (INS/MinSalud), 9 of the 33 departments are at risk on water quality. The problem is not only how much water there is, but which of it is drinkable.

What it is, and what it is not

The first deliverable —the water-anticorrelation map on real data— is already built and runs in the browser: agua.xiliux.com. It is an analysis-and-proposal piece, written in Rust, on IDEAM public-domain data. It does not replace official basin studies: it uses them as a source and shows its working.