Clad in waders and armed with bear spray, besieged by mosquitoes and yelling, "Hey, bear! Hey, bear!" to warn any lurking bruins that I was coming, I slogged with the scientists up cobble-bottomed creeks where thousands of big mature sockeye glowed psychedelically red. They turn that color for their final days, as if they're flaming out at the end of their exhausting migration from the ocean back to their home waters, intent on spawning where they were born years ago. Some of the streams were so shallow and narrow, they seemed to contain more fish than water.

We counted salmon, not only in the creeks but also in areas along the lake shores where some preferred to spawn. We netted and measured and tagged them as they thrashed in our hands, and recorded their eventual deaths from predators or exhaustion. A typical narration to the note-taker, by a scientist examining a single female carcass left on a streambank: "AW" ... "K2" ... "BK" ... "belly 10 percent!" (Translation: AW is the tracking number on the plastic tag they'd attached to this salmon the previous day, when it was alive; K2 indicates the creek segment where the carcass was found; BK means "bear kill," as shown by teeth marks. And check it out: The bear chomped out only the eggs, the richest nutrition, and discarded 90 percent of the fish.)

We also netted different species of fish and recorded their basic data, including stomach contents. The method for that: Grab the slippery fish, stick the tube from a squeeze-bottle of water into its mouth and flush the stomach contents into a pan; then release the fish back into the stream. "Grayling" ... "182" ... "62" (species, length and weight) ... "17 black fly adults, 10 caddis fly larvae, one stone fly nymph, 10 midge pupae, and salmon eggs" (what this particular grayling had eaten lately).

The scientists climbed trees to retrieve memory cards from cameras positioned to shoot photos and video of predators gorging on the salmon (when reviewing the images, scroll past the vegetarian moose that wandered through the frame). They clipped fins off fish so the DNA could be recorded, to track not only individual fish, but also generations of offspring over the years. They used tweezers to extract otoliths – the tiny stones in a salmon's skull cavity, formed by the minerals it swims through – which can show not only where a salmon traveled in fresh- and saltwater, but also the timing of its movements.

This kind of science is incredibly laborious, requiring high tolerance for immersion in cold water, hours in the rain punctuated by blasts of sunburn, swarms of stinging and biting insects, and the constant risk of irritating a bear, even in camp. (One evening we watched a bear swim across the bay and pad ashore into the brush on our side; a few weeks earlier, within sight of our camp, a bear killed a moose calf while the mother moose circled helplessly.) You also have to put up with fragrant outhouses and the lack of electricity, as a generator and solar panels provide only a few hours of juice per day. The fieldwork is less Herculean than Sisyphean; day after day, the scientists push the boulder up the hill, gathering precious data for later analysis and re-analysis and re-re-analysis.

All this work reveals an underlying truth: The sheer complexity of inter-related species and habitat is essential for an ecosystem's health, particularly for resilience to stresses like climate change. And the inverse is also true: When you "coarsen" an ecosystem (Schindler's term) by introducing roads that carve up the habitat, channeling streams, erasing wetlands, and bringing in mining projects, new transmission lines, more buildings, more traffic and so on – as we've already done in most of the West, and as is proposed for the headwaters of part of this area – the ecosystem weakens and may even collapse.

The University of Washington's Alaska Salmon Program is billed as "the world's longest-running effort to monitor salmon and their ecosystems." UW scientists began working here back in 1946, and generations of them have returned every year since. They have six camps in the area, concentrating on the Wood River, which they consider a proxy for the eight other major rivers that also flow into the saltwater of Bristol Bay – a megasystem collectively hosting the world's best sockeye runs.

The only system comparable in the Lower 48 is the Columbia River and its tributaries, stretching from the Oregon coast to headwaters in Idaho and Canada. Coarsened by more than a hundred big dams and vast modifications to floodplains and wetlands, the Columbia supports roughly 1 to 2 million migrating salmon in a good year – 10 percent of the number it used to support – and most of them are raised in hatcheries and injected into the system like a shot of methamphetamine. In Alaska, most salmon runs are still near the historical highs. The Wood River system alone – an area less than one-one-hundredth the size of the Columbia – averages 2 to 3 million sockeye per year, and sometimes hits 10 million. And they're all wild.