When the Glaciers Go: Water Security in the Tropical Andes

1Background: glaciers as a seasonal reservoir

Tropical glaciers store water in the wet season and release it in the dry season, when rainfall alone cannot sustain agriculture, drinking water or hydropower. Peru holds around 70% of the world’s tropical glaciers.^[2] A review of the tropical Andes concludes that glaciers have been retreating for several decades, which has temporarily increased dry-season water supply downstream, and that projected shrinkage will lead to a long-term reduction in dry-season river discharge from glacierized catchments.^[1]

The retreat is closely related to the observed increase in high-elevation air temperature in the region.^[1] Globally, a 2023 projection finds that glaciers will lose 26% of their mass (at 1.5°C of warming) to 41% (at 4°C) by 2100 relative to 2015, and that mass loss is linearly related to temperature increase, so every reduction in warming reduces loss.^[6]

This brief asks three questions. How fast are Andean glaciers disappearing? How much of the region’s water do they actually supply, and when? And what does the evidence support as a response?

2The record of loss

Peru’s National Inventory of Glaciers quantified a 48% decline of glacial area, from 2,041.85 km² in 1962 to 1,058.37 km² in 2021.^[2] In the southern Peruvian Andes, a satellite analysis found that the glacierized area has shrunk rapidly since the 1970s, with losses of 54%, 56% and 64% in the Cordilleras Vilcanota, Vilcabamba and Urubamba respectively. The Cordillera Vilcanota lost 3.18 ± 0.44 gigatonnes of ice between 2000 and 2020, and low-lying glaciers receded the most.^[2]

Responses differ by climate. Using a simple scaling approach, Vuille and colleagues suggest that glaciers in the inner tropics, such as Antizana in Ecuador, may be most vulnerable to future warming, while glaciers in the more arid outer tropics, such as Zongo in Bolivia, may persist, albeit in a smaller size, throughout the 21st century regardless of emission scenario. They also list major uncertainties, including snowfall measurement in the accumulation zone, glacier thickness, and the role of black carbon and groundwater.^[1]

3Peak water: why the supply can look stable before it falls

The best-documented case is La Paz, Bolivia. A mass-balance study of 70 glaciers in the city’s drainage basins found that glaciers contributed about 15% of water resources at an annual scale between 1963 and 2006, 14% in the wet season and 27% in the dry season (Figure 1). Despite the loss of 50% of the glacierized area during the study period, runoff at La Paz did not change significantly, because increased ice melt rates compensated for the smaller surface area.^[3]

That compensation has an end. The same study estimates that, if the glaciers disappeared completely and precipitation did not change, runoff would diminish by about 12% at an annual scale, 9% in the wet season and 24% in the dry season.^[3] This is the timing problem in concrete form: stable flows can mask a shrinking reservoir, and the loss then arrives in the season when water is scarcest.^[1,3]

4Who depends on glacier melt

Reliance on melt is concentrated and rises in drought. A regional mapping of water demand downstream of major tropical glaciers found that the number of users relying continuously on water with a high (above 25%) long-term average contribution from glacial melt is low, at 391,000 domestic users, 398 km² of irrigated land and 11 MW of hydropower. That reliance increases sharply in drought, to up to 3.92 million domestic users, 2,096 km² of irrigated land and 732 MW of hydropower in the driest month of a drought year. Many of the affected domestic and agricultural users live in rural regions where climate adaptation capacity tends to be low.^[4]

The lesson is that a glacier’s importance is not measured by its average contribution. In drought months, the buffer carries far more of the load, and it is rural, low-capacity users who depend on it.^[4]

5What the evidence supports as a response

Natural storage. Pre-Inca cultures in the tropical Andes developed nature-based water harvesting technologies to manage drought risk. An evaluation of a 1,400-year-old infiltration system that diverts headwater stream water onto mountain slopes in the wet season found that infiltrated water is retained for an average of 45 days before resurfacing, confirming the system’s contribution to dry-season flows. The authors estimate that upscaling the system to the source-water areas of Lima could delay 99 million cubic meters of streamflow a year and increase dry-season flows by 7.5% on average, which they describe as a potentially critical complement to conventional engineering solutions.^[5]

Storage and regulation more broadly. The mapping study suggests that adaptation should focus on increasing natural and artificial water storage and regulation capacity to bridge dry periods.^[4]

Governance and culture. The Vuille review argues that local water management practices reflect cultural belief systems and spiritual values, that comprehensive adaptation needs to consider science, policy, culture and practice, and that social and political problems over water allocation for subsistence farming have led to conflicts through lack of adequate water governance.^[1]

6Implications for disaster risk management

Analysis. This section is Disaster Science’s interpretation of the evidence above. It is not drawn from the cited sources.

Plan from projections, not from the recent record. The La Paz case shows that recent flows can look stable while the reservoir shrinks. Water plans and rights agreements that use historical averages will overstate future supply in the dry season.

Design for the drought month. Averages hide the months and the users for whom melt matters most. Risk assessments should map reliance in dry months and in drought years, and prioritize rural users with low capacity.

Start now because the lead time is long. Storage, infiltration works and allocation rules take years to build or agree. Because glacier loss unfolds over decades, the sensible moment to start is while flows still look normal.

Treat natural storage as one tool among several. Infiltration systems add modest, seasonal buffering and can sit beside reservoirs and efficiency measures, but the evidence does not show they replace a glacier’s role at scale.

7Uncertainty and limitations

Different inventories. Glacier area losses are reported for different regions, baselines and periods, and are not directly comparable: 48% for Peru nationally from 1962 to 2021, and 54% to 64% for three southern cordilleras since the 1970s.^[2]

Single-city evidence. The seasonal contribution figures come from La Paz in 1963 to 2006 and may differ for other cities, basins and climates. Soruco and colleagues also report uncertainty in their estimates, including about 12% of total discharge from the mass-balance uncertainty.^[3]

Upscaling is modeled. The Lima figure for infiltration systems is an estimate that extends findings from one catchment, and the system works on wet-season streamflow, not glacier melt.^[5]

Scope. This brief synthesizes peer-reviewed studies and does not include original data or a review of Bolivian and Peruvian water law. It is an editorial synthesis that has not undergone peer review.