Ice–albedo feedback

• 1. Introduction

  ›

  Ice-albedo feedback is one of the most consequential self-reinforcing loops in Earth's climate system, and understanding it begins with a single physical fact: ice is extraordinarily reflective. Open ocean absorbs the sunlight that falls on it. Ice sends most of that same energy straight back to space. Change which one covers the surface, and you change the planet's temperature. Do that at scale, and the planet changes itself.

  The mechanism runs in both directions. Warm the planet enough to melt ice, and the darker surface left behind absorbs more heat, which melts more ice. Cool the planet enough to grow new ice, and the fresh reflective surface bounces away more solar energy, which cools the planet further, which grows more ice still. Each change feeds the next. That is the feedback.

  This loop has shaped Earth's history across timescales almost too large to hold in mind. It was active during the Sturtian glaciation roughly 717 million years ago, when ice crept toward the equator and a runaway freeze nearly swallowed the entire planet. It drove the repeated advances and retreats of ice sheets throughout the Pleistocene, the geological period that ended only around 10,000 years ago. And it is operating right now, in the Arctic, where sea ice has been shrinking for decades in ways that are measurable from space.

• 2. Early Science

  ›

  In the 1950s, climatologists such as Syukuro Manabe were already trying to account for the role of ice cover in Earth's energy budget, working to quantify how the planet actually balances incoming and outgoing radiation.

  The pivotal year was 1969. That year, the Soviet scientist Mikhail Ivanovich Budyko and the American William D. Sellers each independently published papers introducing some of the first energy-balance climate models. Both showed that the reflectivity of ice had a substantial effect on Earth's climate, and that changes to snow and ice cover in either direction could act as a powerful feedback. Two scientists, two countries, one conclusion.

  By 1974 the process was recognized as a crucial part of climate modelling in a formal review. A year later, in 1975, Manabe and Richard T. Wetherald used what they called "snow cover feedback" in a general circulation model designed to measure the effects of doubling atmospheric CO2 concentration. That doubling scenario became a standard benchmark for climate sensitivity, and the ice-albedo mechanism was baked into it from the start.

  Since those foundational papers, the feedback has been included in every generation of climate models. Researchers also apply the same calculations to paleoclimate studies, using modern physics to reconstruct the ancient climates of the Pleistocene.

• 3. The Poles Today

  ›

  Continuous satellite readings of Arctic sea ice began in 1979, and the data collected since then tells a consistent story: the Arctic is warming nearly four times faster than the global average, a phenomenon scientists call Arctic amplification.

  Modelling studies make the connection explicit. Strong Arctic amplification appears only during the months when significant sea ice is being lost. When simulated ice cover is held artificially constant in models, the amplification largely vanishes. The ice loss is not a consequence of the warming alone; it is an active contributor to it.

  Antarctica presents a contrast that illuminates how geography shapes feedback. The East Antarctic ice sheet is so massive it rises nearly 4 kilometers above sea level. That elevation keeps the continent cold, and Antarctica has experienced very little net warming over the past seven decades. What warming has occurred is concentrated in West Antarctica, and it is driven not by ice-albedo feedback but by the warming of the Southern Ocean, which absorbed 35-43% of all the heat taken up by the world's oceans between 1970 and 2017.

  The global accounting of ice loss between 1992 and 2018 is stark. The combined warming effect of declining Arctic sea ice and more recent Antarctic sea ice loss over that period equals roughly 10% of all the greenhouse gas emissions produced during the same years.

• 4. The 2023 Antarctic Shock

  ›

  In 2023, Antarctica broke every record for low sea ice extent in the 45 years of satellite observation. February's summer minimum fell by 1.79 million square kilometers compared to the 1981-2000 climatological mean, a drop of 38%. The September winter maximum reached just 16.96 million square kilometers, breaking the previous record by more than one million square kilometers, an area roughly the size of Texas and California combined.

  Researchers traced the trigger to a deepened Amundsen Sea low in the preceding spring. That atmospheric pattern drove anomalous northward transport of sea ice, leaving behind large expanses of open water. The darker water absorbed more solar energy, and the ice-albedo feedback accelerated the melt from there.

  2023 was also the year global mean temperatures reached 1.5 degrees Kelvin above pre-industrial levels. Scientists found an unexpected gap of about 0.2 K that could not be explained by anthropogenic warming and El Nino events alone. That gap was attributed to the record planetary lows in albedo, which contributed approximately 0.22 plus or minus 0.04 K of warming on their own.

  The 2023 anomaly demonstrated that the feedback is not a distant threat. It is already producing measurable, recordbreaking temperature effects in the present climate.

• 5. Ice-Free Arctic Futures

  ›

  Under every climate change scenario modelled, the Arctic Ocean is projected to experience at least one ice-free September before 2050. Under the scenario of continually accelerating greenhouse gas emissions, that first ice-free September is expected around 2035.

  September marks the annual minimum of Arctic sea ice, the end of the summer melt season. Losing it entirely would be a historic threshold, but its effect on the ice-albedo feedback is actually limited. By September, the sun sits low over the Arctic and the total solar energy arriving is already very low. June is the month that matters more. At the peak of the Arctic summer, even a relatively small reduction in June sea ice extent would transfer far more additional solar energy to the ocean.

  A 2018 paper put the frequency of ice-free Septembers into probability terms. Under 1.5 degrees C of warming, such an event would occur roughly once every 40 years. Under 2 degrees C, once every 8 years. Under 3 degrees C, once every 1.5 years.

  CMIP5 models estimate that complete loss of Arctic sea ice from June through September would raise global temperatures by 0.19 degrees C, with the range running from 0.16 to 0.21 degrees C. Regional Arctic temperatures under the same scenario would rise by more than 1.5 degrees C. Crucially, these figures are already embedded within the warming projections of every CMIP5 and CMIP6 model; they are not additional warming on top of existing forecasts.

• 6. Tipping Points and Ice Sheets

  ›

  There is a threshold beyond ice-free summers: an Arctic that stops reforming sea ice through the winter entirely. Unlike a summer without ice, a year-round ice-free Arctic may represent an irreversible tipping point. Models place the most likely trigger at around 6.3 degrees C of global warming, though it could occur as early as 4.5 C or as late as 8.7 C.

  The warming difference between that state and the Arctic as it was in 1979 is equivalent to roughly a trillion tons of CO2 emissions. That figure is around 40% of the 2.39 trillion tons of cumulative emissions recorded between 1850 and 2019. About a quarter of that impact has already been realized through current sea ice loss. The remaining potential global warming effect, relative to the present, would be 0.6 degrees C globally and between 0.6 and 1.2 degrees C regionally.

  For the large land-based ice masses, the timeline extends far further. Greenland, West Antarctica, and East Antarctica all participate in ice-albedo feedback, but their full disappearance would take centuries or millennia. Climate models do not include them in 21st century projections for that reason. Experiments modelling their total loss assign 0.13 degrees C to Greenland, 0.05 degrees C to West Antarctica, and 0.08 degrees C to mountain glaciers, all at a baseline of 1.5 degrees C average warming.

  The East Antarctic ice sheet, rising nearly 4 km above sea level, would require a minimum of 10,000 years to melt entirely even under very high warming. Its maximum potential contribution to global temperature is estimated at around 0.6 degrees C.

• 7. Darkening Ice and Dust

  ›

  The ice-albedo feedback can be amplified by anything that darkens a reflective surface. Particles deposited on snow and ice lower the albedo directly, and higher concentrations produce larger decreases. Black carbon and mineral dust are the primary culprits from the atmosphere. Living organisms add another layer: snow algae on glaciers and ice algae on sea ice both reduce reflectivity, and the melt they cause generates liquid water that stimulates further algal growth, creating a self-reinforcing biological feedback within the physical one.

  The Siachen Glacier in the Karakoram range of the Himalayas offers a documented case study in human-caused darkening. The glacier has been a military zone disputed between India and Pakistan since 1984, and the sustained presence of troops has measurably affected its reflectivity. Researchers report that the glacier has retreated by several kilometers since the conflict began. Constant use of kerosene for heating and cooking deposits black carbon soot across the glacier's surface. Non-biodegradable waste, including plastics, metal scraps, and discarded equipment, forms structures researchers call "dirt cones" that trap heat and create localized melting points.

  In the Swiss Alps, satellite imagery shows that bare-ice zones on glaciers have darkened in recent decades. Parts of the ablation areas have shown statistically significant albedo declines of about 0.05 per decade since the late twentieth century. Earlier seasonal snowmelt exposes darker underlying ice for longer, and substantial growth of pigmented glacier algae on bare ice has been documented across the European Alps, compounding the darkening and accelerating melt.

• 8. Snowball Earth

  ›

  Paleoclimate evidence places the most extreme expression of ice-albedo feedback at around 717 million years ago, when the Sturtian glaciation began and ice spread toward the equator. The feedback drove temperatures down, and as more ice formed, more solar radiation was reflected away, driving temperatures still lower. Geological evidence of glaciers near the equator survives in the rock record from that period.

  The Sturtian glaciation persisted until about 660 million years ago. A brief warmer interval followed before the Marinoan glaciation began, ending around 634 million years ago. Whether the Earth froze completely into a solid snowball or retained a thin equatorial band of liquid water as a slush ball remains a genuine scientific debate. But ice-albedo feedback is central to either scenario.

  The exit from Snowball Earth conditions involved the same mechanism working in reverse. Dust from erosion gradually accumulated on the snow-ice surface, layer by layer, lowering the albedo. The midlatitudes were the likely starting point: colder than the tropics but receiving less precipitation, which meant less fresh snow to bury the accumulating dust and restore reflectivity. Once enough midlatitude ice was gone, temperatures rose, and the resulting isostatic rebound eventually enhanced volcanism, allowing CO2 to accumulate in the atmosphere in ways that had been impossible while the ice persisted.

  The dust-driven deglaciation of Snowball Earth is an ancient illustration of the same principle documented today on the Siachen Glacier and in the Swiss Alps: darken the surface, and the ice retreats.