Commercial fishing boats hovering near the Wood River's mouth capture 60 percent of the migrating sockeye, on average, yet enough make it past the fishermen to form "the Red Wave" – a huge pulse of sockeye in the river and associated streams and lakes that supports the abundance of predators, from bears down to the blowflies that lay eggs in salmon carcasses so their maggots can feast. Scientists have found that a single carcass can support 50,000 maggots, which, in turn, are consumed by other insects, birds and fish.

The first scientists here "developed an integrated view – what we would call 'ecosystem science' today," Schindler told me. "The strength of our program is the long-term measurements," with the continuous studies generating "immense data sets."

The program's current luminaries include gray-haired Thomas Quinn, the author of a definitive textbook, The Behavior and Ecology of Pacific Salmon and Trout. He helped establish how salmon navigate far out in the ocean and then return to their home streams: They sense geomagnetic fields in the ocean and smell distinctive odors in freshwater, which lead them back to the exact spot where they hatched. Some of this information is genetically passed to offspring, Quinn believes.

The salmon face very long odds. The average spawning female lays 3,000 eggs. To maintain a healthy population, at least two of those offspring need to grow up and return to spawn a new generation. That doesn't always happen: Bears kill an astounding number of the Wood River sockeye that make it past the fishermen, ranging from 5 percent on the river itself to as much as 90 percent on the tiniest creeks. Once the bears' initial hunger is sated, they become connoisseurs, often chomping out only the brains of the males and the females' eggs. Scientists calculate that salmon flesh provides a respectable 2 kilojoules of energy per gram, while the eggs and the brains provide 10 kilojoules per gram. "Bears are omnivorous," Quinn said, "but nothing is as predictable and rich as salmon."

As the ripples of the Red Wave spread, the predators transport the dead salmon and their eggs around the ecosystem: Bears carry the carcasses a short distance, spreading nutrients in the riparian area, while gulls take fragments and salmon eggs longer distances to their nesting places, where they enrich little islands and lakes, providing food for snails and Alaska blackfish. Scientists have even found "salmon signatures" embedded in the feathers of songbirds, because nutrients derived from salmon feed the plants that produce berries the birds eat.

Jonny Armstrong, a post-doc working with Schindler, dispensed this expert advice about how it all fits together, pointing to a salmon carcass: "When you find a dead salmon that still has its eyes, you know you're tight on a bear." Translation: Gulls quickly discover salmon carcasses left by bears, and they peck out the fishes' eyes, so if you find a carcass punctured by big teeth marks and its eyes are still intact, you know that you just scared off the bear. It's probably hiding in the bushes waiting for you to move on. The gulls will arrive any second.

The complexity of this wild ecosystem begins with the water in the hundreds of creeks that form the rivers of Bristol Bay. Each creek draws from a different ratio of snowmelt, summer rains and fall rains. When some creeks are low due to a dry summer, others will likely be normal or high, which means the average is more consistent than you would expect looking at just one or two of them. The same goes for the nine rivers: Their average total flow is more reliable than any individual river.

The natural topography and geology are also complex. Some creeks rush down super-steep mountain valleys, while others dawdle down gentler grades from little spring-fed lakes; some drain volcanic soil and others don't; some are colder and some are warmer. This provides a varied habitat for salmon to exploit.

Hiking up Lynx Creek one day, above Lake Nerka, Schindler explained the "hydrological complexity" of just this single creek. It has a cold tributary (roughly 44 degrees Fahrenheit) fed by groundwater, a warm tributary (65 degrees) originating in a small headwaters lake, and a general mosaic of relatively warm pools and cold rapid segments, plus much warmer pools off the main channel, where young fish get stranded in low flows. This complexity provides "refugia" during varying weather and flow conditions, Schindler said. When a tributary floods with rains, for instance, the juvenile salmon attempting to feed in it leave, hanging out in the main channel just above the confluence – "a velocity refuge" – until the flood subsides. If the whole main channel floods, they seek sanctuary in the side pools that are barely connected to the main.

I sat on a bank at the confluence of Lynx Creek and its cold tributary, where bears had flattened the grass and left portions of carcasses, and watched sockeye swim up to me. In many places, the water was so shallow that it didn't even cover their humped backs and dorsal fins. In brief stretches that were merely soggy gravel, the fish wriggled, rather than swam, uphill. Some turned left at the confluence to go up the cold tributary and spawn there, while others kept going up the main channel, heading for higher segments or the headwaters lake, to spawn in warmer water. Those spawning in the cold tributary are genetically distinct from those spawning in the main channel – one data set that supports the conclusion that the locator information is passed on to offspring. "We're discovering the genetic diversity of salmon also benefits the consumers (predators)," Schindler said. If salmon can't spawn in one segment of the creek for some reason, the other segments might be OK, so the predators can find salmon pretty consistently during the run in this creek.

Even the timing of salmon runs is complex. Some creeks have early runs, in June and early July, some have middle-of-the-season runs, and some have late runs, in August. Runs on the spawning sites in the river itself last into September, and runs on the lakeshores, where some fish spawn, last into October. Each run hits its peak for two to three weeks, but because the timing is staggered, the run for the whole Wood River system lasts eight weeks or longer. The predators have learned to roam around the fenceless, roadless miles, hitting each creek, river segment and lake at its peak, Schindler said.

The salmon's varying life cycles also lend resilience to the system. Of the offspring from a single batch of sockeye eggs, some will stay in the freshwater for a year, while others linger for two years. Once they migrate out, some spend two years in the ocean, some three. So one batch of eggs can produce salmon that return to that exact spawning site three years later, four years later, and five years later, providing multiple opportunities to successfully reproduce at that site. Schindler made this point tangible on Berm Creek, which has a persistent sandbar where it meets Lake Nerka. A few years ago, a flood deposited so much sediment on the sandbar that it completely closed off the creek to spawning salmon. No problem for the ecosystem: Other creeks did well that year, and the next spring's runoff was strong enough to blow an opening in the sandbar, allowing the next run to return to every spawning site in Berm Creek. Ultimately, the one-year disaster didn't matter.