Skip to content

The Example Problem

The training problem is:

Choose priority ecological monitoring sites that maximize biological value and environmental coverage while minimizing redundancy and cost.

This is intentionally familiar. Environmental data scientists and ecologists already make decisions like this when designing monitoring networks, prioritizing field campaigns, or choosing conservation areas.

The structure echoes systematic conservation planning: choose a subset of candidate places that represents important ecological features while accounting for cost and practical constraints. The demo keeps this idea small enough for a training workflow, but the conceptual ingredients are the same ones used in larger spatial prioritization systems.

Candidate Sites

The demo creates 100 synthetic candidate sites. Each site has:

  • a location (lat, lon)
  • species_richness
  • climate_refugia_score
  • habitat_connectivity
  • cost
  • region
  • environmental features such as temperature, precipitation, and elevation

The default task is to choose 12 sites.

In optimization language, these are candidate planning units. Each site can be selected or not selected. That binary choice is what makes the problem easy to translate into a QUBO.

Decision Tension

The best network is not simply the 12 richest sites. A useful monitoring network also needs environmental coverage, reasonable cost, and low redundancy. Two near-identical sites may both look strong on their own, but selecting both can add less information than selecting complementary sites.

That tension is what makes the problem useful for optimization practice.

A strong result is therefore not simply "the highest score." It is a defensible tradeoff. One solution might favor biological value, another might favor environmental coverage, and a third might reduce cost. The scientific task is to make those choices explicit enough that a working group can debate them.