Water management plans are built around futures. Every design standard, every infrastructure investment, every land-use decision encodes an assumption about what is coming — a flood of a certain magnitude, a population of a certain size, a rainfall of a certain intensity. This pattern describes what happens when that assumption hardens into a single authoritative forecast, and the system is optimised around it. It connects backward to No Fat Modelling, where the problem begins in how models are built, and upward to Governance Cannot Reflect on Itself, where the institutional failure to learn from surprise is described.

When planners optimise for a single scenario, the system performs well under that scenario and fails badly under all the others.

The single scenario is attractive for understandable reasons. Decisions require commitment. Infrastructure must be sized. Budgets must be allocated. A range of futures feels like an admission of ignorance, and decision-makers under political and financial pressure tend to prefer a number to a distribution. The model obliges. A design flood is chosen, a planning horizon fixed, a population projection adopted. The plan is written around these figures and the figures gradually become the future — not a guess about it, but a substitute for it.

The brittleness this produces is not immediately visible. The system performs as designed under the conditions it was designed for. It is only when reality diverges — a flood exceeding the design standard, a city growing faster than the projection, a drought arriving in a system optimised for surplus — that the cost of the single scenario becomes legible. And by then the infrastructure is built, the land is occupied, and the institutions have organised themselves around assumptions they no longer examine.

There is a compounding effect. A plan built around one future tends to crowd out the adaptive capacity that might have handled a different one. Resources committed to the design case are not available for the unexpected case. Communities settled behind the design-standard embankment have stopped developing the local knowledge and informal practices that preceded it. The single scenario does not just fail to anticipate alternatives — it actively erodes the system's ability to respond to them.

How institutions avoid learning from this failure is the territory of Governance Cannot Reflect on Itself. How the model that generates the single scenario should be built is addressed in No Fat Modelling.

Plan explicitly across multiple scenarios rather than optimising for one. Choose infrastructure and governance responses that perform acceptably across a range of futures, even if they are not optimal for any single one. Make the scenarios and their assumptions visible and revisable as conditions change. Treat the design case as one plausible future among several, not as a prediction — and preserve the adaptive capacity that allows the system to respond when reality arrives differently than expected.

Connected patterns: No Fat Modelling — Governance Cannot Reflect on Itself — The Safety Paradox — Relief Crowds Out Prevention — Science Waits Its Turn