A Systematic, Regional Assessment of High Mountain Asia Glacier Mass Balance

Every major river system draining into South and Central Asia begins as ice. The Ganges, the Indus, the Brahmaputra, and the Mekong all originate in High Mountain Asia, a region that holds the largest concentration of glacier ice found anywhere outside the poles. Hundreds of millions of people downstream depend on that ice for irrigation, drinking water, and hydropower. Until very recently, no one had actually measured how fast it was disappearing across the whole region, at high resolution, all at once. This paper by David Shean and colleagues provides that measurement. High Mountain Asia — the Tibetan Plateau and its surrounding ranges, including the Himalaya, Karakoram, Tien Shan, and Pamir — contains 95 thousand five hundred thirty-six individual glaciers covering roughly 97 thousand six hundred square kilometers. That's the Randolph Glacier Inventory count, and it's a staggering number. These glaciers act as both climate indicators and seasonal water reservoirs. Shean and colleagues note that beyond supplying water, they can also pose hazards like glacial outburst floods. The long-term picture is clear and grim: Himalayan glaciers have been retreating since roughly eighteen hundred fifty, with sustained mass loss documented through the mid-twentieth century and accelerating losses since the mid-1990s. Climate projections calibrated against these observations suggest substantial fractions of High Mountain Asia glacier mass may be gone by twenty-one hundred. The problem is that before this study, the numbers didn't agree. Different satellite methods provided wildly different answers for how much mass was being lost each year. Gardner and colleagues, comparing methods for two thousand three to two thousand nine, found that ICESat laser altimetry reported a regional loss of about twenty-nine gigatons per year, GRACE satellite gravimetry reported nineteen gigatons per year, and extrapolations from traditional field measurements reported eighty-six gigatons per year. That's not a small spread — that's a factor of four between the low and high estimates. The disagreement had real methodological causes. ICESat provides accurate elevation changes but along sparse ground tracks, so spatial sampling is patchy. GRACE detects mass changes integrated over entire basins — it can't distinguish one glacier from another, and it mixes glacier melt with groundwater changes and other terrestrial water signals. Earlier digital elevation model differencing studies had coarser resolution and limited spatial coverage. The region also has sharp spatial contrasts — severe loss in most ranges, with near-zero or positive mass balance in others — that coarse methods simply can't resolve. Shean and colleagues addressed this by building a dataset at a scale that had never been attempted. They processed sub-meter commercial stereo imagery from DigitalGlobe and Maxar — WorldView-1, WorldView-2, WorldView-3, and GeoEye-1 — to generate five thousand seven hundred ninety-seven high-resolution digital elevation models, concentrated between two thousand thirteen and two thousand seventeen. They also reprocessed twenty-eight thousand two hundred seventy-eight ASTER stereo pairs spanning two thousand to two thousand eighteen. Combined, that's roughly thirty-four thousand digital elevation models covering ninety-nine percent of High Mountain Asia glaciers. Stereo imagery works similarly to how your two eyes work: two photos of the same mountain taken from slightly different angles allow you to compute a three-dimensional surface elevation map. The team used the NASA Ames Stereo Pipeline for elevation model generation and co-registered every model to the TanDEM-X global reference, which has absolute vertical accuracy around three point five meters. For each glacier, they stacked all available elevation models through time and fit a linear elevation change trend using the Theil-Sen estimator — a robust statistical method that isn't influenced by outliers. The conversion from height change to mass is conceptually straightforward: change in mass equals density multiplied by change in volume, using a bulk ice density of eight hundred fifty kilograms per cubic meter. The verdict from all of this: High Mountain Asia lost ice at a rate of negative nineteen point zero plus or minus two point five gigatons per year between two thousand and two thousand eighteen. In specific terms — normalized by glacierized area — that's negative zero point one nine plus or minus zero point zero three meters water equivalent per year. To put a gigaton in physical terms: one gigaton of water equals a cubic kilometer. High Mountain Asia was losing roughly nineteen cubic kilometers of ice annually. That's the headline, and it falls right between the previous satellite estimates, with substantially better spatial coverage and resolution than any prior study. Where the loss occurs is where this study really advances the science. The spatial pattern is heterogeneous in ways that matter. The greatest losses concentrated across the Western, Central, and Eastern Himalayas, Nyainqentanglha, and the Central and Eastern Tien Shan. At the basin level, the Brahmaputra lost four point eight seven plus or minus one point zero one gigatons per year, the Indus lost three point five three plus or minus zero point nine seven, and the Ganges lost three point one nine plus or minus zero point five eight. Meanwhile, the Western Kunlun Shan and Eastern Pamir showed slightly positive mass balance — the ice there was actually growing or holding steady. The Karakoram, long noted for anomalous behavior, showed a west-to-east gradient: near balance or slightly positive in the west, slightly negative in the east. This spatial nuance — positive pockets embedded in a mostly negative regional picture — was only resolvable because of the fine spatial aggregation that thirty-four thousand elevation models made possible. Glacier size also mattered: very small glaciers under zero point one square kilometers were on average less negative than the regional mean, while glaciers in the ten to thirty square kilometer range tended to be more negative. Now, take that mass loss and translate it into two consequences that touch human lives directly. The first is sea level. The total High Mountain Asia loss of nineteen gigatons per year corresponds to a sea-level contribution of zero point zero five two plus or minus zero point zero zero seven millimeters per year. Counting only exorheic basins — those that drain to the ocean — the rate was zero point zero three seven plus or minus zero point zero zero six millimeters per year, accumulating to roughly zero point seven millimeters of sea-level rise over the eighteen-year study period. That's about four to six percent of the total contribution from all global glaciers and ice caps combined. Not the dominant driver of sea-level rise, but a real and measurable portion of it. The second consequence is more immediate for the people who live downstream. Shean and colleagues distinguish between total glacier meltwater runoff — which includes normal seasonal melt from a glacier in equilibrium — and what they call excess meltwater runoff, which is the water released specifically because the glacier is shrinking. That excess is a one-way withdrawal from a finite account. Across High Mountain Asia basins, this excess component ranged from roughly twelve to fifty-three percent of each basin's total glacier meltwater runoff. Interior endorheic basins on the western Tibetan Plateau showed the highest percentages, around twenty-five to fifty-three percent. The Indus, Ganges, and Brahmaputra fell in the eighteen to twenty-nine percent range. What this means in practice is that a meaningful fraction of the river water those basins are currently receiving — water used for farming, drinking, and power — is not coming from sustainable seasonal melt. It's coming from the glacier itself disappearing. That's a short-term water bonus with a long-term reckoning built in. The dataset Shean and colleagues assembled is the foundation that makes future work possible. Because it covers ninety-nine percent of High Mountain Asia glaciers at this resolution, it can calibrate and validate glacier mass-balance models, cross-check GRACE gravimetry trends, and feed the hydrological models that governments use to plan water infrastructure. Ongoing satellite tasking campaigns — new WorldView and GeoEye stereo imagery, continued ASTER acquisitions, and laser altimetry from NASA's ICESat-2 — will extend the record forward in time. The ice is shrinking at nineteen gigatons per year. We now know, at high resolution and across nearly the entire region, where and how fast. That knowledge is the prerequisite for every serious water-management decision that follows across the most glacierized non-polar region on Earth. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.