PART 1

Data about Earth’s 78 most important mountaintops foretell changes in the amount and timing of snowmelt.

Scientific American Walter Immerzeel

The nights are long inside a tent 5,300 meters above sea level at the snout of Nepal's Yala Glacier. At 8:00 P.M., after a meal of Nepali dal bhat (lentils and rice), the 10 members of our expedition take refuge against the cold in sleeping bags inside the small tents that make up our temporary camp. Falling asleep is tough because the low oxygen concentration fools our bodies into increasing their heart rates. As a consequence, I spend many overnight hours listening to distant sounds of thundering avalanches and cracking ice, contemplating whether to leave the sleeping bag to pee outside and what not to forget the next day. As soon as the sun rises, the camp is bustling, and we are on our way up the steep glacier to install special instruments at 5,600 meters.

Our team, which includes colleagues from the International Center for Integrated Mountain Development in Nepal, has been conducting field expeditions biannually in this place, called the Langtang catchment, since 2012. We have erected automated weather stations at the base camp and at higher elevations that measure precipitation, snow depth, radiation, temperature, relative humidity and wind, making Langtang one of the best-monitored high-altitude catchments in Asia. We need to visit the stations every six months to maintain the instruments and to download their data; there is no cellular network to transmit readings automatically, and the mountains tend to block satellite signals. On the current ascent we will mount new sensors on a metal frame three meters high that we will drill into the ice. The sensors will measure sublimation—the phase transition of ice directly to water vapor—by sampling temperature and vapor 10 times a second.

These expeditions help us gather the information necessary to understand the high-altitude water cycle: snow falls on mountaintops and gradually turns into glacier ice, which slowly flows downhill and melts. The water cascades down into growing rivers that supply numerous high-elevation settlements, as well as agricultural terraces, hydropower stations, forests, valley agriculture fields, and large cities and industries below. The high-altitude water cycle was largely a mystery when we started. We did not know how much rain and snow actually falls, how much water flows into and out of the snowpack, or why glaciers blanketed in detritus eroded from surrounding slopes seem to melt as fast as their debris-free counterparts. We need to know these details to determine how much water ultimately flows out of the snowpack and glaciers and how the volume and timing of flow may change in the future.

The Langtang catchment is a small river basin that drains several mountain peaks and glaciers. It feeds the Trishuli River, an important water source for recently constructed hydropower dams halfway down the mountains and the irrigated fields farther downstream. The amassing flow leads to the Ganges Delta hundreds of kilometers away, which provides water to 400 million people as it empties into the Bay of Bengal east of India. There are hundreds of catchments like Langtang across the Himalayas.

The same dynamics play out across many other mountain ranges, such as the Alps, the Andes and the Rockies. A study we published in Nature in December 2019 revealed that 78 of these “water towers,” from the Tarim in China to La Puna in Peru, supply the bulk of fresh water for almost two billion people globally. Computer models show that climate change could threaten these vital water supplies. Changes in temperature, precipitation patterns, the accumulation and melting of snow, and the distribution of airborne particulates all help to determine how much water descends. Yet most plans and policies related to climate change and sustainability ignore the role of mountain water. Now that we have systematically studied the world's top water towers, that may begin to change.

Monsoon Dominance

Mountains function like water towers because it rains and snows more at higher elevation than it does in lower surrounding terrain and because much of the precipitation is stored temporarily as snow and glacier ice. This snow and ice melts slowly and steadily, providing a reliable, predictable supply of water and acting as a natural buffer against dry spells.

The water stored high in Nepal's mountains is crucial for the people who live below because monsoon winds between June and September provide 70 to 80 percent of the Himalayas' annual precipitation. Data from our weather stations, in conjunction with weather models, reveal how the monsoons interact with the mountains. Even across a small catchment such as Langtang, the rains vary strikingly. The valley runs east to west, and its elevation rises in that direction. Much of the warm, moist air precipitates between the entrance of the valley, at around 1,350 meters, and the village of Lama Hotel at 2,480 meters. We have measured around 250 centimeters of annual rainfall at this site, the wettest in the valley. The village of Kyanjin, at 3,900 meters, is only another 15 kilometers to the east, but there we measure around 80 centimeters of annual rainfall—more than three times as dry yet so nearby.


The position and shape of the mountains create other local effects. If we ascend from Kyanjin to the Yala Glacier at 5,300 meters, precipitation increases again by 40 percent on average. Together the large, medium and valley-scale processes shape the distribution of rain and snow throughout the region. If most precipitation falls as rain at lower altitudes, for example, the rivers below will flow differently than they would if most precipitation fell as high-altitude snow.


To understand the dynamics of water towers, we have extensively investigated the amount of water stored in the snowpack close to Yala Glacier. This amount, known as the snow-water equivalent, is influenced by how much snow falls, how much melts and refreezes, how much sublimates and how much is distributed by wind. Our instruments at the 5,300-meter base camp measure many of these kinds of variables. The conditions can be severe: extreme cold has caused batteries to explode, high winds have twisted sensors, and avalanches have knocked over the scaffolding holding our instruments.

It is particularly tricky to measure sublimation, a turbulent process that occurs under cold, sunny, windy and dry conditions, which are common during the Himalayan winter. Instruments such as the ones we erected at 5,600 meters have quantified sublimation by measuring variations in air humidity and temperature. We find that at windy, exposed locations, about 21 percent of the snowfall never ends up in a river because it sublimates into the atmosphere. This phenomenon suggests that even at temperatures well below zero degrees Celsius, the snowpack can shrink.

We have also found that when the air temperature two meters above the snow is below zero, there is enough energy on the snow's surface to melt it. The energy is the net result of the shortwave radiation from the sun, the longwave radiation emitted by the surface and the atmosphere, and turbulent heat fluxes. There are interesting dynamics inside the snowpack as well. At least 30 percent of the snow that melts into water during the day refreezes at night. Much more energy is required to melt snowpack than what would be expected purely on the basis of its mass.

Another instrument, which we install above the snow, measures changes in gamma radiation as a proxy for snow-water equivalent. Rock under the snow naturally emits gamma rays, and the degree to which their signal is attenuated is related to how much water is actually stored in the snowpack.

Sometimes we hit on unexpected ways to collect information. Around 2013 a colleague working in Tasmania sent drones up over a landslide there and came back with intriguing data about the landslide's volume and velocity. We realized we could use the same approach to obtain data on glaciers covered in debris, which are hard to access. Our first outing was in 2013 on the glacier tongue just above Kyanjin.

The drone takes overlapping pictures of the glacier's surface. Software determines the surface elevation to a remarkable resolution of about 10 centimeters. We repeated the surveys every six months or so until 2019. We found that the glacier's edge is receding by about 40 meters a year, that the surface is thinning by about 80 centimeters a year and that the dwindling ice hardly flows anymore. It will not be long until the cold mass no longer qualifies as a glacier. Instead it will be a piece of dead ice slowly withering away like a heap of dirty snow plowed to the end of a parking lot. Theoretically, a debris-covered glacier should melt much more slowly than a debris-free glacier at the same altitude because the debris acts as insulation. But we identified hotspots of melt that amplified the overall process. We would never have found them with traditional satellite imagery because its resolution is too coarse.

We put all these puzzle pieces into our models to learn how much water will flow in the rivers in the future. One last big piece, though, is the amount of water flowing in the rivers today. The water height alone cannot indicate the volume of flow. We need a so-called rating curve—a relation between the water level and the river's discharge. And this curve has to represent both high monsoon flows crashing downhill and meager winter runoff. Generating reliable data is a challenge here, too. In the rivers, we either lodge a pressure transducer housed in a steel pipe at the bottom of the river or mount radar sensors on frames jutting up a few meters above the river's surface. We also throw salt into the river upstream of the gauges and measure changes in electrical conductivity at the gauge site; the extent of salt dilution can help us determine the discharge. Even though the conditions are more hospitable here than at 5,300 meters, sensors get washed away by monsoon floods. But after years of effort, we now have a fairly good idea of water flow.

Continued in Part Two...