Ignoring sea-level uncertainty could make flood protection nearly 70% more costly

Summary

Under a high sea-level-rise scenario, a plan based on one forecast would fail safety standards at more than 80% of the modelled dike segments.

Correcting that plan later could cost nearly 70% more than accounting for uncertainty from the start, or become infeasible under construction limits.

The model covered 150 dike segments and 14 large-scale alternatives in the Rhine-Meuse Estuary-Drechtsteden region.

Why is one sea-level forecast not enough for long-term flood planning?

Dikes and other flood defences are planned decades ahead, while the rate of future sea-level rise remains uncertain. A plan built around one forecast can appear safe today but fall below safety requirements if conditions develop differently.

Testing a fixed plan against a few scenarios after it has been designed also misses an important feature of real infrastructure planning: some investment decisions can be changed later as new information becomes available.

Why is this a prescriptive analytics problem?

Flood planning requires decisions at several points in time: what to build now, what to postpone and how later choices should change when sea-level observations become clearer.

Adjustable robust optimisation models that sequence directly. Instead of selecting one fixed plan, it identifies a strategy that remains safe across a defined range of futures while preserving the option to adapt later.

How did the collaboration model decisions under uncertainty?

Researchers from Tilburg University and Deltares developed the approach and tested it in the Rhine-Meuse Estuary-Drechtsteden region.

The model represented future sea-level rise as a set of possible paths. At defined decision points, the investment strategy could change according to what had actually happened, while each path still had to meet safety requirements.

The case study included 150 dike segments and 14 alternatives to dike raising, including storm-surge barriers and changes to river channels.

What did the findings show?

If sea-level rise reached the high end of the tested range, a plan that ignored uncertainty would fail its safety standards at more than 80% of dike segments.

Correcting that plan after the fact could cost nearly 70% more than accounting for uncertainty from the start, and some corrections could become impossible within construction scheduling limits.

The study also tested uncertainty ranges as large as 40% around sea-level forecasts. Building that uncertainty into the plan increased initial costs only modestly in the model, while the ability to adapt allowed spending to remain lower when conditions developed more favourably. For communities behind the dikes, the central implication is that safety does not have to depend on one forecast being correct.

What can the findings mean for other applications?

The approach can be tested for other infrastructure that must remain reliable for decades while future conditions are uncertain, including coastal barriers, drainage networks and reservoirs.

It is particularly relevant where some investments must be made early, but later decisions can still change. Future applications can incorporate different climate projections, construction constraints and local definitions of acceptable risk.