ATOP MAUNA KEA, Hawaii — Two and a half miles above the Pacific, with the combined exhalations of a vast swath of humankind and its cars and factories blowing toward him, Aidan Colton looked out over the volcano’s snow-streaked summit and lifted up a glass flask the size of a coconut.

He held his breath — even the carbon dioxide from his lungs might corrupt the sample. After a moment, he opened the valve.

The air he is collecting at Mauna Kea is feeding the world’s longest-running record of direct readings of heat-trapping gases in the atmosphere. The measurements constitute the most complete body of firsthand evidence for how Earth’s chemistry has changed since the mid-20th century, contorting the global climate. They represent a triumph of long-haul scientific commitment. And they were thrown into crisis late last year.

For six decades, scientists had been taking their air measurements from a cluster of squat buildings on Mauna Loa, another giant volcano on Hawaii’s Big Island. Then, in November, Mauna Loa erupted for the first time in almost 40 years. No one was hurt, but lava flows up to 30 feet deep toppled the observatory’s power lines and buried a mile of the main road up the mountain. The facility was paralyzed.

It took a transoceanic scramble, and a dose of luck, for scientists with the Mauna Loa observatory to restart their readings — by taking them, for the first time, on Mauna Kea, the next volcano over.

The interruption highlights the careful planning and delicate labor that goes into gathering this data, plus the obstacles, both human and natural, that can get in the way. It shows how the simple-seeming task of measuring the air is anything but.

After Mauna Loa started belching lava, technicians with the National Oceanic and Atmospheric Administration, which runs the observatory, flew in and installed instruments on Mauna Kea shortly before a fierce winter storm brought hurricane-force winds to the summit, which could have delayed the work. They got it done so quickly because, months earlier, NOAA had already started exploring putting a backup site there, at a telescope run by the University of Hawaii.

“It’s certainly lucky timing,” said Brian A. Vasel, the director of observatory operations for the NOAA Global Monitoring Laboratory. But “it’s certainly not a coincidence.”

The agency ended up going a little more than a week without measurements. The Scripps Institution of Oceanography, which had also been monitoring the atmosphere at Mauna Loa for decades, began collecting data at Mauna Kea a week after NOAA did. Scripps, which is part of the University of California, San Diego, uses instruments that are separate from NOAA’s but housed at the agency’s facilities.

NOAA is now ferrying solar panels and batteries up Mauna Loa by helicopter to restore power at the stranded observatory. The plan is to collect parallel measurements at both volcanoes for a year to see how they compare, and to evaluate whether Mauna Kea, which last erupted 4,600 years ago, might become a long-term backup for Mauna Loa, the world’s largest active volcano.

Once the lava cools enough to put in a new road, which could be by early summer, the agency also plans to begin upgrading its aging Mauna Loa observatory with revamped laboratory space, improved electricity and fiber-optic connectivity, and its first septic system. The site currently relies on a cesspool.

“We had obsolete facilities,” Mr. Vasel said. The goal now is “building the site that’s going to support the Mauna Loa mission for the next decade, and decades to come.”

As a place for sampling air that best represents the planet’s as a whole, Mauna Loa has several unique things going for it. So NOAA officials faced a tough question last year as they scouted for a backup: How could they possibly do this work anywhere else?

It helps, first, to be surrounded by thousands of miles of empty ocean. It helps to be very high up, away from nearby towns, cars, and forests, and in the path of air masses that travel the globe at great altitudes, swirling together the gases above many of Earth’s most developed regions.

Both Mauna Loa and Mauna Kea check these boxes. But the air at the two sites most likely blows through from somewhat different heights in the atmosphere, which could influence the greenhouse-gas levels they detect. The Mauna Loa facility is half a mile lower in elevation than the Mauna Kea one. It also sits below the volcano’s summit, which allows it to catch clean, high-altitude air that cools and flows downhill upon hitting the cold nighttime slopes. The Mauna Kea site is at the mountain’s peak, where there isn’t this effect.

To see how much these differences might affect NOAA’s readings, Mr. Colton, an atmospheric technician with the agency, has been making weekly trips up both volcanoes, flying by helicopter at Mauna Loa and navigating a treacherously rocky road at Mauna Kea.

At each volcano, he fills glass flasks with long sips of the mountain air, in a ritual that has changed little since Charles David Keeling, a scientist with Scripps, began sampling the atmosphere at Mauna Loa in the 1950s. Some of the flasks are even the same ones from decades ago.

The analog methods help ensure the measurements can be compared over time. But it’s still up to Mr. Colton to figure out how to collect his samples in conditions that are as consistent as a possible week to week. At Mauna Loa, he worked out long ago where on the volcano to stand, and at what time of day, to capture the air at its cleanest. He’s still figuring that out at Mauna Kea, where the 13 stargazing stations bend the wind and tourist traffic throws off carbon levels.

One area on the volcano’s western flank became inaccessible after it snowed. Another time, a snowplow spewing fumes showed up while Mr. Colton was sampling.

“Every time you change something, you can create another anomaly, something that can affect the outcome,” he said.

It is not a glamorous setting. The ceiling is crumbling. The internet cables are taped to the wall.

Mr. Colton peered at the analyzer’s screen. Two valves in the equipment were stuck. Again. This seemed to be happening more here than at Mauna Loa, he said, and he wondered whether cold temperatures and aging equipment might be to blame.

NOAA hopes to complete the first stage of its upgrades at Mauna Loa by fall 2024, Mr. Vasel said. The price tag: $5.5 million.

Scraping together resources for the Mauna Loa operation has long been a battle. A few years ago, the road up the volcano needed maintenance, said Darryl Kuniyuki, who manages station operations. The federal government provided some money, he said, but not enough to pay contractors to repaint the lines.

“I had to get creative,” Mr. Kuniyuki said, “so I hired the Boy Scouts.” He and other observatory employees did most of the painting. Local boys did the rest as an Eagle Scout project.

For funding agencies, long-term atmospheric monitoring is a tough sell, said Ralph Keeling, a scientist at the Scripps Institution and Charles David Keeling’s son.

“Climate change is playing out decade by decade; you don’t know what’s happening unless you’re observing decade by decade,” the younger Dr. Keeling said. “That means measurements on a time frame much longer than a typical science project.”

“At some point, the agencies say, ‘Well, why are we paying for this?’” he said.

Nowadays, the Big Island isn’t the only place where scientists are tracking global carbon levels. With newer methods, researchers can estimate emissions down to the individual factory, power plant, and oil field. As Mr. Colton rattled down Mauna Kea’s gravel road in a truck, he explained that the observatory’s readings still provided an all-important baseline for understanding other data on emissions.

They are “the foundation,” he said, “the pillars at the bottom that everybody comes back to.”