Conservation Decision Making Under Uncertainty

Real species. Real trade-offs.
Partial information.

Each game here is based on a published POMDP or MDP model for wildlife management. You make the decisions. The hidden dynamics unfold. See how natural resource models can help inform real-world decisions.

Interactive Simulations

Choose a conservation problem to manage

Partially Observable Problems
POMDP Problems — Hidden State, Imperfect Observation
Species Detection
Sumatran Tiger — Managing an Invisible Population
After Chadès et al. (2008, PNAS)
You are the conservation manager for a critically endangered tiger population. Survey detection is imperfect — a missed sighting doesn't mean extinction. A hidden catastrophic threat (disease outbreak, prey collapse, poaching surge) can activate silently and drive the population to collapse. You must choose across three actions — Monitor, Act, or Do Nothing — over 10 time steps.
2 States 3 Actions Imperfect detection Hidden threat 10-step horizon
Play Now →
Invasive Species Management
Weed Control — The Seeds You Can't See
After Regan, Chadès & Possingham (2011, J. Applied Ecology)
An invasive plant infestation starts as an adult population — visible, detectable. But fumigation is catastrophic if seeds are present in the soil and you can't tell seeds from an empty field. One wrong action and you poison your own crop. Navigate three management strategies across three hidden states over 10 steps.
3 States 3 Actions Imperfect detection Latent seed bank 10-step horizon
Play →
Multi-Population Management
Tiger Two-Population — Splitting the Budget
After Chadès et al. (2008, PNAS)
Two isolated tiger subpopulations, each at risk of collapse. You can protect one, monitor one, or neglect both — but resources are limited. Imperfect observation means you can't always tell which populations are present. The interaction between local extinction and metapopulation rescue makes this a classic conservation triage problem.
4 States 4 Actions Two subpopulations Imperfect observation 10-step horizon
Play →
Model Uncertainty · Expert Disagreement
Gouldian Finch — Which Expert Is Right?
After Chadès et al. (2012, AAAI-12)
Four competing ecological models explain Gouldian Finch decline: some emphasizes trophic cascades (dingoes suppress cats), others emphasize direct predation. Each model implies a different optimal intervention. Manage the species under model uncertainty — and watch which model's predictions prove correct as evidence accumulates across 10 decision steps.

This game introduces POMDP with model uncertainty (also known as Adaptive Management): the belief state spans not just population states but competing explanations of the world.
4 Expert Models 4 Actions Model uncertainty Expert disagreement 10-step horizon
Play →
Fully Observable Problems
MDP Problems — Known State, Resource Trade-offs
Fisheries Quota Management
Fisheries — Fishing the Right Amount
After Clark (1990); Quota-based MDP formulation
A commercial fishery at the edge of viability. You set the annual harvest quota — too little and the fishing industry collapses economically, too much and the stock crashes. Population dynamics are stochastic but fully observed each season. Find the sustainable yield policy that keeps the fishery profitable for 10 years without driving the stock to collapse.
Continuous stock Quota actions Stochastic growth 10-step horizon
Play →
Predator Population Management · SDP
Wolf Population Management — Balancing Conservation & Livestock
After Marescot et al. (2012, Methods Ecol Evol) — SDP primer
Wolf populations are recovering in a region where conservation goals and livestock losses conflict. Use Stochastic Dynamic Programming to find the optimal management policy: when to cull, when to protect, and when to do nothing.
10 Population states 4 Actions Discounted (γ=0.95) 10-step horizon Policy comparison
Play →
Two-Species Interaction · MDP · Chadès 2012
Sea Otter × Abalone — Managing a Trophic Conflict
After Chadès et al. (2012, Conservation Biology 26(4):720–730)
Sea otters are recovering — and competing with abalone. Both species are of conservation concern, but they are in direct trophic conflict. Managing one without accounting for the other leads to suboptimal outcomes for both. Navigate difficult trade-offs and find a solution that works for both species.
9 Joint states 4 Actions Trophic interaction Discounted (γ=0.95) 10-step horizon Policy comparison
Play →