After an oil spill

A spill has a first week and a long tail. In the first week the question is where the slick will be tomorrow, and responders answer it with trajectory models. The long tail is where models of places are thin. Sixteen years after Deepwater Horizon, a study of the Gulf of Mexico's midwater found its fish, krill and squid down 60 to 85% and not recovering, a loss nobody saw from the surface. The Gulf's deep-diving whales fell further than the models of the spill's damage had predicted.

The people who carry the long tail are concrete: a fisheries council setting catch limits, a coastal town whose charter boats and processors depend on those fish, a trustee or agency deciding where restoration money goes, a court or an insurer weighing damage.

What you are really asking

• Where did it go? On the surface, onto the shore, down through the water and onto the sea floor. Each path needs a different kind of evidence.
• What did it reach? Which habitats and which animals, at which depths, and through what: the oil, the dispersant, or the two together.
• Is it recovering, and how would we know? This needs a baseline from before, and counting that continues after the news has moved on.
• What carries over? Fewer prey in the midwater may mean thinner whales, smaller fish catches and less carbon carried down by the animals that migrate each night. Each of these is a separate claim to test.
• What would have changed the outcome? A different response, a faster cap, less dispersant, a closed fishery. The counterfactual is where a model earns its keep, and where it is least certain.

What kind of model answers it

Trajectory and weathering models move oil with the wind and the currents and show how it changes on the surface. NOAA's GNOME, built by its emergency responders, is free on the web. Ocean circulation models supply the currents, including the deep ones a surface trajectory model does not cover. Food-web models follow what a loss at one level does to the levels above and below. Ecopath with Ecosim is a free one, and its Ecotracer module follows a contaminant through the web. Monitoring is the part no model replaces: the Gulf study rests on more than 2,000 net samples over thirteen years, and its counts are public in GRIIDC, so a model can be tested against them.

A whole-system model of a spill joins these: oil and currents feeding exposure, exposure feeding the food web, the food web feeding fisheries and the people who depend on them. Each link is a published kind of model; the joins between them are where the work is. The Gulf adds one more link that simple models leave out: below some density, sparse animals stop finding mates, so a population knocked down can stay down. Ecologists call it the Allee effect.

What you need

• Where and how much was released, and over how long, with the uncertainty stated. Official estimates and press accounts can disagree.
• Currents and winds for the period, surface and deep.
• A baseline: counts of the animals and habitats from before the spill. Without one, recovery cannot be measured.
• Repeated surveys afterwards, by the same method.
• Catch and landing records, and who depends on them.

Without a model

Find out who is still counting. For the Gulf, the study's authors and its funders, NOAA Research and the Gulf of Mexico Research Initiative, are the place to start. For a spill near you, ask whether anyone recorded what lived there before.