A drought index tells you how unusual the weather has been. It does not tell you how much water is left in the reservoir upstream of a city. These are different questions, they can disagree in both directions, and confusing them is one of the most common mistakes we see when someone puts a risk map in front of an operational decision.
What a drought index actually measures
SPEI, the index underneath much of what we publish, compares accumulated precipitation and evaporative demand against the same location's own history. SPEI-3 asks whether the last three months were wet or dry for this place. A province at −1.5 is not "drier than its neighbour at −0.5" in absolute terms; it is further from its own normal. That framing is a feature, not a limitation: it is what lets a map cover a rainforest and a dry plain at once.
But everything in that sentence is about the atmosphere. Rainfall arrived, or it did not. Evaporative demand was high, or it was not. Nothing in the index knows whether anyone captured the water that fell.
What sits between the rain and the tap
Between a rainfall anomaly and a water shortage there is a catchment, a set of reservoirs with their own operating rules, an allocation policy that decides who is served first when storage falls, and a distribution network with its own losses. Each of those is a choice made by people, and each can absorb or amplify what the weather did.
This is why a meteorological drought and a water-supply crisis are not the same event, and why they do not have to happen in the same season. Storage carries the past forward. A reservoir filled by a strong wet season can serve a dry one without anybody noticing; a reservoir already drawn down for irrigation can fail in a season the index calls normal.
The disagreement runs in both directions
The direction people expect is a dry index and no shortage: the rain failed, storage was healthy, and nothing happened to the customer. That case is common and mostly benign, though it does erode trust in the map when it repeats.
The direction that costs money is the other one. The index reads normal, and supply is still short — because the previous year's deficit was never made up, because the wet season arrived as a few intense bursts that ran off instead of infiltrating, or because demand grew faster than anyone re-planned for. A monitoring process that only escalates when the index goes red will be silent through all three.
Neither direction is a defect in the index. Both are the index being asked a question it was not built to answer.
How to read the two together
Treat the index as a statement about input, and storage as a statement about state. Input tells you what is arriving; state tells you what you have to work with. An operational picture needs both, and the useful cases are the ones where they disagree, because a disagreement is where a decision actually lives.
In practice this means three habits. Look at storage on the same screen as the anomaly rather than in a separate report nobody opens. When the two disagree, say which one you are acting on and why, in writing. And when you escalate, name the quantity that crossed a line — "storage at the serving reservoir", not "the drought map turned orange" — so that the next person can check the same thing you checked.
None of this requires a more complicated index. It requires being explicit about which question is on the table.
What we are not claiming
GlobMaps publishes a meteorological signal. We do not measure reservoir storage, and a risk score from us is not a statement about anyone's water supply. Where a decision turns on storage, the number has to come from the operator who reads the gauge, and our map is context around that number rather than a substitute for it.
We would rather say that plainly than let a clean-looking score be read as more than it is. A map that is trusted for the wrong thing is worse than one that is trusted for nothing, because the failure only shows up on the day it matters.