Small changes in Earth’s orbit tweak how sunlight is distributed across latitudes and seasons, but the globally averaged change in incoming energy is modest. Yet the planet’s response across glacial–interglacial cycles is anything but modest: continental ice sheets grow and collapse, oceans reorganize, and ecosystems migrate. That mismatch—small orbital nudge, huge global rearrangement—demands an explanation. CO₂ is the loudest supporting actor: not usually the first shove, but the amplifier that helps a subtle push become a sustained, planet-wide shift.
To be clear about roles, we need two plain-English definitions. A forcing is an external push on Earth’s energy budget—something that changes the balance between energy coming in and energy going out without needing the climate to change first. Orbital variations are forcings in this sense: they alter the geometry of sunlight delivery whether or not ice, oceans, or clouds respond. A feedback is an internal response that depends on the initial climate change and then loops back to strengthen or weaken it. If warming melts ice, lowering reflectivity and causing more warming, that extra warming is a feedback effect. CO₂ can be either, depending on timescale and context: in the industrial era, we add CO₂ directly and it acts primarily as a forcing; in glacial cycles, the initial trigger is typically orbital (especially summer insolation at high northern latitudes), and CO₂ largely acts as a feedback that amplifies and prolongs the change. Same molecule, different job description.
One physical fact is non-negotiable and purely mechanical: more CO₂ in the air traps more outgoing heat. CO₂ molecules absorb infrared radiation in specific wavelengths that Earth’s surface and lower atmosphere emit as heat. When CO₂ increases, the atmosphere becomes more opaque in those bands, so infrared photons that used to escape to space from lower, warmer layers get absorbed and re-emitted many times. The net effect is that the effective “radiating level”—the average altitude from which those wavelengths finally escape to space—moves upward to colder air. Colder air emits less infrared radiation, so at the top of the atmosphere the outgoing heat temporarily drops. Energy then accumulates in the Earth system until temperatures rise enough that outgoing heat once again matches incoming sunlight. That is the greenhouse effect in working parts: changed infrared transparency, reduced heat loss, then warming until balance is restored.
Now the puzzle pieces fit. Orbital forcing is a precision tool, not a sledgehammer. It doesn’t add much to the planet’s annual average sunlight, but it can strongly alter where and when sunlight hits—most importantly, the intensity of summer sunlight over regions where ice sheets can melt or persist. A slightly cooler northern summer can let winter snow survive; surviving snow brightens the surface, reflecting more sunlight; that local persistence can seed broader ice growth. But to turn a regional seasonal imbalance into a global, multi-millennial climate state, you need levers that operate everywhere, all year. CO₂ is exactly that kind of lever.
Quantitatively, glacial–interglacial CO₂ changes are large enough to matter and consistent across multiple cycles. Ice-core records show atmospheric CO₂ cycling roughly between ~180–280 ppm (range) across the last several hundred thousand years. Over the same swings, global mean temperature changes are on the order of ~4–6 °C (range), and global mean sea level shifts by roughly ~120–130 m (range) as ice sheets lock up or release water. Those are not subtle changes. If you only look at the orbital “nudge” in globally averaged watts per square meter, it seems too small to plausibly deliver the full temperature and sea-level response on its own. The climate system’s internal amplifiers—ice albedo, water vapor, clouds, ocean circulation, and CO₂—are how the nudge becomes a global transformation.
This is where two common wrong beliefs need to be punctured in passing because they sound plausible. You’ll often hear: “If CO₂ lags temperature, it’s irrelevant.” That confuses sequence with influence. In many glacial terminations, Antarctic temperature begins rising before CO₂ rises, implying the initial push came from elsewhere (orbital changes filtered through ocean circulation). But a lagging feedback can still do enormous work—like a microphone that starts amplifying after the singer begins. The singer initiates the sound; the amplifier controls whether the whole stadium hears it and how long it sustains. Once CO₂ rises, it alters the planet’s heat-loss rate everywhere, reinforcing warming, encouraging further ice retreat, and helping the system cross thresholds that a seasonal regional nudge alone might not maintain.
The opposite misconception also circulates: “CO₂ alone causes ice ages.” If that were true, we’d expect glacial cycles to be paced primarily by internal carbon fluctuations without the striking alignment to orbital periodicities and without the strong seasonal, high-latitude fingerprints that ice-sheet physics demands. In reality, ice ages are not simply “CO₂ goes down, therefore ice grows.” Ice sheets care intensely about summer melt, which is governed by regional insolation. Orbital variations provide the timing and the geography of vulnerability—when summers are cool enough for snow to persist. CO₂ then helps determine how far the system runs once it starts: how cold the mean state becomes, how much ocean heat is withheld from high latitudes, and how stable an icy world remains over thousands of years.
A useful way to keep the logic straight is to imagine the climate as a heavy door with a spring. Insolation changes are the hand that pushes the door at just the right angle—small force, but applied where it matters (summer at high latitudes). The door starts to swing; that initial motion changes the room’s airflow and pressure. CO₂ is like the spring tension adjusting as the door moves: it is not the first push, but once engaged it changes how strongly the door returns or continues, and it can lock the door into a new open or closed position for a long time. In glacial cycles, orbital geometry times the push; ice and oceans begin to respond; CO₂ responds to those changes (through ocean solubility, biological pumps, and circulation) and then feeds back, modifying the whole system’s energy balance so the new state is larger and longer-lived than the orbital nudge alone would sustain.
So CO₂ is neither the puppet master that single-handedly summons ice ages nor a bystander dismissed because it “lags.” It is a powerful climate lever whose role depends on the story’s timescale: in glacial–interglacial cycles, CO₂ is typically the amplifier and sustainer of change that was initially nudged by insolation, helping explain why small orbital forcing can accompany large global shifts in temperature and sea level.
So where does CO₂ live, and how does it move?
“CO₂ in the air” is the part of the carbon cycle we can measure most directly, but it is not the whole carbon system—it’s a small, mobile slice that sits on top of much larger reservoirs. Thinking only in terms of the atmospheric number is like judging a bank’s solvency by the cash in one ATM: it matters, it moves quickly, and it affects what people can do today, but most of the money is elsewhere, and small reallocations from the big vaults can swing what you see in the ATM by a lot. The key idea is inventory: where the carbon is stored, in what form, and how easily it can move between stores.
Start with the main reservoirs. In the atmosphere, carbon is mostly present as CO₂ gas mixed through the air, a thin film compared with the planet’s other carbon stores. In the surface ocean, carbon is mostly “dissolved” in seawater as a mix of forms that behave like a chemical buffer—carbon that can exchange with the air on timescales of months to years. In the deep ocean, carbon is again dissolved, but it’s isolated from the surface by slow circulation, so it functions like a huge warehouse that trades with the atmosphere on centuries to millennia. On land, the biosphere and soils store carbon as living tissue (wood, leaves, roots) and dead organic matter (litter, humus), essentially carbon bound into biological molecules. In sediments and rocks, most carbon is locked into carbonates and organic-rich deposits—geologic storage that is enormous but usually exchanges with the surface system very slowly.
Once you see those inventories side by side, the “small slice” point becomes unavoidable. The ocean holds vastly more carbon than the atmosphere. You don’t need the ocean to change by a lot in percentage terms to change the atmosphere by a lot in absolute terms, because the ocean’s reservoir is so large. If the atmosphere is a small checking account and the ocean is a huge savings account, then a tiny transfer out of savings can double what’s in checking without meaningfully denting the savings balance. This is why glacial–interglacial CO₂ swings can be driven largely by ocean rearrangements: the atmosphere is sensitive to how the ocean partitions carbon between surface and depth, even if the ocean’s total carbon inventory barely changes in percentage.
To understand that partitioning, you need one concept in plain English: dissolved inorganic carbon in seawater is not just CO₂ molecules floating around unchanged. When CO₂ enters water, much of it converts into other dissolved forms that don’t behave like a gas anymore. You can think of seawater as having a “carbon handling system” that rapidly shuffles carbon between forms that are easier or harder to vent back to the air. Some of the dissolved carbon is present in a form that can readily turn back into CO₂ and escape; much of it is stored in forms that are more stable in water. Because the water can convert CO₂ into these other forms, the ocean can take up a lot more total carbon than you’d expect if CO₂ simply sat there as a dissolved gas. That conversion also means that changing temperature, circulation, or acidity shifts the balance among those forms, which changes how strongly the surface ocean “wants” to hold onto carbon versus release it to the atmosphere.
With that in mind, “rearrangement” becomes the right word. You don’t always need new carbon coming in from volcanoes or leaving into rocks to change atmospheric CO₂; you can shift carbon among reservoirs and forms. One concrete example is simple warming of surface waters. Warm water holds less gas than cold water—like a warm soda going flat faster—so when the surface ocean warms, it tends to release CO₂ to the atmosphere. The ocean hasn’t created new carbon; it has changed the balance between dissolved forms and the gaseous exchange, and because the atmosphere is a small reservoir, that outgassing can noticeably raise atmospheric CO₂. Flip the sign and you get another example: when surface waters cool, they can absorb more CO₂ from the air, and because cold, dense water is more likely to sink, some of that carbon can be carried into the deep ocean where it’s stored away from the atmosphere for long periods.
Biology provides a second kind of rearrangement that doesn’t require the total carbon inventory to change much. In the sunlit surface ocean, phytoplankton take up dissolved carbon to build organic matter. Some of that organic carbon sinks as particles or is transported downward as organisms die and clump, effectively exporting carbon from the surface layer to the deep ocean—a process often called the “biological pump” in carbon-cycle shorthand. If export becomes more efficient, the surface ocean is left with less dissolved carbon available to equilibrate with the atmosphere, so atmospheric CO₂ can fall. If export weakens—because nutrients are redistributed, ecosystems shift, or stratification cuts off nutrient supply—the surface retains more carbon, making it easier for CO₂ to remain in or return to the atmosphere. Again, the move that matters is not necessarily “more carbon exists,” but “where is it parked, and how connected is it to the air?”
These rearrangements also clarify two common failure modes in how people think about long-term CO₂ changes. One is treating CO₂ like a dial controlled only by volcanoes, as if the atmosphere rises and falls mainly because volcanic sources crank up or down. Volcanic outgassing is real, and over geologic times it matters enormously, but in many climate swings the dominant story is redistribution within the ocean–atmosphere–biosphere system rather than a sudden change in the planet’s carbon income. If you assume volcanoes must be the driver, you miss how a tiny shift in ocean storage or circulation can swing the atmospheric “meter reading” without requiring a dramatic change in total carbon entering the surface system.
The other failure mode is treating the biosphere as irrelevant, as if land plants and soils are too small or too passive to matter compared with the ocean. Land reservoirs are not the biggest in total mass, but they are dynamic and responsive. They can gain carbon when climate conditions favor growth and soil accumulation, and lose carbon through respiration, fire, decay, and thaw. Over glacial–interglacial transitions, vegetation zones shift, deserts expand or retreat, peatlands grow or drain, and permafrost stores can become more or less stable. Those are not minor details; they alter how much carbon is held on land versus available to the ocean–atmosphere system. Ignoring the biosphere is like ignoring all the small businesses in an economy because they’re individually smaller than the biggest corporation—collectively, their behavior changes the flow.
The punchline is that atmospheric CO₂ is a messenger, not the warehouse. It responds quickly to exchanges with the surface ocean and biosphere, and it reflects slower reorganizations in the deep ocean and, over longer spans, the geologic reservoir. Because the atmospheric inventory is small relative to the ocean’s, the atmosphere is exquisitely sensitive to how the ocean partitions carbon between the surface layer that “talks” to the air and the deep layer that does not. That’s why understanding past CO₂ swings means tracing pathways, not just looking for a single knob.
Start with the fastest exchange partner: the surface ocean.
CO₂ moves between air and ocean on years-to-centuries timescales through a fast “skin” exchange at the surface and a slower conveyor that connects that skin to the deep ocean. If you picture the atmosphere as a thin blanket and the surface ocean as a shallow sponge underneath it, CO₂ is constantly being breathed back and forth across the boundary. What determines the direction and strength of that breathing is partly temperature, partly the ocean’s chemical buffering, and partly the ocean’s circulation—how quickly the surface sponge is squeezed, refilled, and connected to the huge reservoir below.
Henry’s law is the plain-English starting point: colder water holds more gas; warmer water holds less. You already know this from everyday life: a cold fizzy drink stays carbonated; warm it up and it goes flat. The same physics applies at the sea surface. When surface waters cool, they can dissolve more CO₂ from the air before they become “satisfied,” so CO₂ tends to move from atmosphere to ocean. When surface waters warm, the surface ocean’s capacity to hold dissolved CO₂ drops, so it tends to release CO₂ back to the atmosphere. This gives an immediate glacial–interglacial intuition: colder glacial oceans generally favor drawing down atmospheric CO₂, while warmer interglacial oceans favor higher atmospheric CO₂—everything else equal. The important caveat is that “everything else equal” is almost never true for the real ocean.
Temperature is powerful, but it is not the whole story because CO₂ does not simply dissolve and sit there as a gas. Seawater has a built-in chemical buffer that shuffles carbon into different dissolved forms, and that buffering depends on alkalinity—a measure of seawater’s ability to neutralize acids and, practically, its capacity to store carbon in non-gaseous forms. Here’s the plain-English version without a chemistry lecture: when CO₂ enters seawater, much of it is converted into other dissolved forms that are less ready to pop back into the air. That conversion means the ocean can store far more total inorganic carbon than you’d predict from “gas dissolves in water” alone. It also means two surface waters at the same temperature can have different “appetites” for atmospheric CO₂ depending on their alkalinity and existing dissolved carbon inventory. If you like mechanical analogies, think of temperature as changing how stretchy the sponge is, while alkalinity changes how many internal pockets the sponge has to tuck CO₂ into once it’s inside. A stretchy sponge helps, but pocket count matters too.
Circulation adds the second reason temperature alone can’t determine uptake. The atmosphere only exchanges directly with the surface ocean, not the deep ocean. If the surface layer takes up CO₂ but then just sits there, it will eventually approach equilibrium and stop absorbing much more. To keep drawing CO₂ from the air, the surface must be connected to a place where that carbon can be stored away from the atmosphere for a long time—meaning the deep ocean. That connection is controlled by mixing, sinking, upwelling, and the large-scale overturning circulation. In other words, surface exchange is the doorway; circulation determines whether there’s actually a hallway leading to storage, or just a closet that fills up quickly.
These ideas come together in two complementary mechanisms people call the solubility pump and the biological pump. The names sound technical, but the underlying stories are simple.
The solubility pump is the physically driven pathway: cold high-latitude surface waters absorb CO₂ efficiently and then, because they become dense when cooled (and often made saltier by sea-ice processes), they sink and carry that dissolved carbon into the deep ocean. The deep ocean then holds that carbon out of contact with the atmosphere for decades to centuries (and longer, depending on where it ends up). A clear example is wintertime high-latitude cooling: surface water chills, takes up CO₂, becomes denser, and sinks into intermediate or deep layers. The “pump” is not a machine with a piston; it’s the fact that cold water both takes up more CO₂ and is more likely to subduct or sink, physically exporting carbon downward. In glacial climates, expanded sea ice, colder surface waters, and altered stratification can change both the uptake and the export parts of this story, which is why the solubility pump is often invoked in explaining lower glacial atmospheric CO₂.
The biological pump is the life-driven pathway: in the sunlit surface ocean, phytoplankton use dissolved carbon to build organic matter. Some fraction of that organic matter sinks as particles or is transported downward as it is eaten and repackaged, effectively moving carbon from the surface layer into the deep ocean as organic debris and dissolved products. A clear example is a productive region where a bloom occurs: surface carbon is converted into biomass; a portion sinks below the surface mixing layer; as it decomposes at depth, it enriches deep waters in dissolved carbon. If the biological pump strengthens—say, because nutrient supply increases or ecosystems shift toward more efficiently sinking particles—surface waters can end up with less dissolved carbon available to equilibrate with the air, which can lower atmospheric CO₂. If it weakens, more carbon stays near the surface, making it easier for CO₂ to remain in or return to the atmosphere.
Both pumps interact with the concrete physics of how the ocean is stirred. A useful physical picture is wind-driven mixing and upwelling. Winds can push surface waters away from a coastline or along the equator in such a way that deeper water rises to replace them. That upwelled water often carries a large load of dissolved carbon because it has accumulated the products of respiration and decomposition at depth. When this carbon-rich water reaches the surface, it is suddenly exposed to the atmosphere. If its internal “CO₂ pressure” is higher than the air’s, CO₂ will outgas—like opening a bottle that has been shaken and warmed. You can watch this logic play out in major upwelling zones: they are frequently net sources of CO₂ to the atmosphere even while other regions are net sinks. The point is not that the ocean is uniformly absorbing or emitting; it is a patchwork where physics brings different water masses to the surface with different carbon loads and chemical capacities.
Quantitatively, it helps to pin down at least two anchors. First, the surface ocean equilibration timescale with the atmosphere is relatively fast: on the order of months to a few years (range) for the surface mixed layer to substantially exchange CO₂ with the air, depending on winds, temperature, and how deep the mixed layer is. That doesn’t mean the whole ocean equilibrates quickly—only the surface skin that is directly ventilated. Second, across a glacial cycle, atmospheric CO₂ changes by a large, repeatable amount: roughly ~80–100 ppm (range) from glacial lows to interglacial highs, commonly framed as about ~180 to ~280 ppm (range). That magnitude is far too large to treat as a minor chemical detail, yet it is also small enough, relative to the ocean’s total carbon inventory, that it can plausibly arise from modest percentage shifts in how carbon is partitioned between surface and deep.
Put the pieces together and you get a coherent years-to-centuries story. The atmosphere and surface ocean are in a relatively quick tug-of-war set by temperature (Henry’s law intuition), modified strongly by buffering (alkalinity and carbon forms), and constantly rearranged by circulation and biology. Cooling tends to pull CO₂ into the surface ocean; warming tends to push it out. But whether that CO₂ stays near the surface—able to leak back to the air—or is exported into the deep depends on the solubility pump’s physical sinking pathways, the biological pump’s export of organic carbon, and the circulation “plumbing” that can either ventilate the deep ocean or keep it sealed off.
The deep ocean is the real storage vault, but it only matters if circulation opens the vault door.
The deep ocean is the planet’s main carbon reservoir on climate timescales because it is both enormous and, most of the time, poorly connected to the atmosphere. The surface ocean trades CO₂ with the air quickly, but the deep ocean is where carbon can be stored away from the atmosphere for centuries to millennia—long enough to matter for glacial–interglacial cycles. The basic trick is simple: once carbon-rich water is moved out of the surface “conversation layer” and into the abyss, it can’t easily leak its CO₂ back to the air unless circulation later brings it back up.
Deep water forms through a very physical process: water becomes dense and sinks. Cold water is denser than warm water, and salty water is denser than fresh water. In a few key regions—especially high latitudes—surface waters lose heat to the atmosphere, become colder, and sometimes become saltier when sea ice forms (ice excludes salt, leaving the surrounding water saltier). Cold, salty water can become dense enough to sink beneath lighter waters and spread through the deep ocean. When that sinking happens, it drags surface properties downward: dissolved gases, including CO₂, and dissolved carbon that has been converted into non-gaseous forms by seawater’s buffering. It also carries oxygen downward. The crucial connection to carbon storage is that deep waters are cut off from direct contact with the atmosphere; they can accumulate dissolved carbon over time as sinking organic matter decays and respiration products build up. So deep-water formation is like sealing a container and sliding it onto a warehouse shelf: once stored, its contents don’t exchange rapidly with the outside.
During glacials, multiple mechanisms could increase deep-ocean carbon storage (and therefore lower atmospheric CO₂) by making the deep ocean more isolated or by changing how carbon is partitioned between surface and depth. One mechanism is stronger stratification, meaning the ocean becomes more layered and resistant to vertical mixing. If the surface is fresher (from altered precipitation or meltwater routing) or colder in ways that create stable density layering, the boundary between surface and deep can become harder to cross. A more strongly stratified ocean is like a building where the stairwells are locked: carbon that gets into the deep stays there longer, and carbon-rich deep water is less likely to be mixed back to the surface where it could outgas.
A second mechanism centers on the Southern Ocean, particularly how winds and sea ice influence “ventilation”—the exposure of deep waters to the atmosphere. Ventilation is not a mystical term; it means that deep or intermediate waters reach the surface (or near-surface), exchange gases with the air, and then sink again. In today’s climate, the Southern Ocean is a unique region where deep waters can be brought back toward the surface by upwelling and mixing, making it a major gateway for carbon stored in the deep ocean to communicate with the atmosphere. During glacials, expanded sea ice can act like a physical lid over parts of this gateway: reduced open-water area limits air–sea gas exchange, and altered surface buoyancy can suppress the mixing that would otherwise bring deep carbon to the surface. Meanwhile, changes in wind patterns can alter how strongly surface waters are pushed around and how much deep water is drawn upward. Depending on the direction of change, this can either tighten the seal on deep carbon storage or loosen it.
A third mechanism is altered overturning circulation—the global-scale conveyor that connects surface and deep via formation regions, spreading pathways, and return flows. Overturning is not one single loop, but changes in the rate or geometry of deep-water formation in the North Atlantic or around Antarctica can reshape where carbon accumulates and how quickly it is returned. If overturning weakens in a way that reduces the renewal of deep waters with oxygen-rich, low-carbon surface waters, the deep ocean can become older on average, allowing more carbon to build up from the continual rain of organic matter and its decomposition. Conversely, if overturning strengthens or reorganizes so that deep waters are more frequently returned to the surface, stored carbon is more likely to be released to the atmosphere.
This is why the Southern Ocean is often central in discussions of glacial CO₂: it is a major window between the deep ocean and the atmosphere. Most of the deep ocean is insulated from the air by stratification and geography; it doesn’t directly surface and breathe. The Southern Ocean, by contrast, sits astride the pathways that connect multiple ocean basins and hosts strong winds and current systems that can draw old, carbon-rich waters upward. It’s not that the Southern Ocean has magical properties; it’s that it is one of the few places where the ocean’s deep “warehouse” has a relatively accessible loading dock. If you want to change atmospheric CO₂ by changing deep carbon storage, you focus on the main places where deep waters are ventilated and transformed—hence the Southern Ocean’s outsized importance.
A grounded, plausible example of a circulation shift that could raise CO₂ during deglaciation is increased ventilation and upwelling in the Southern Ocean. Imagine a glacial state where deep waters have accumulated extra dissolved carbon, partly because stratification and sea-ice cover reduced their exposure to the air. As deglaciation begins—triggered by insolation changes and amplified by feedbacks—winds can shift and sea ice can retreat, opening more surface area and enabling stronger exchange and mixing. If upwelling intensifies or if the surface becomes less capped, carbon-rich deep water reaches the surface more readily. Once at the surface, the water’s dissolved carbon inventory can translate into a higher tendency to outgas CO₂, because the atmosphere is comparatively low in carbon “inventory” and the surface water is suddenly in contact with air. That outgassing raises atmospheric CO₂, which then amplifies warming, which can further reduce sea ice and change circulation—an intertwined physical story with a clear mechanism: open the gateway, expose the stored carbon, and the atmosphere responds.
A necessary caution belongs here because this is an active research area. What is well-supported is the broad framework: the deep ocean contains vastly more carbon than the atmosphere; exchange between deep ocean and atmosphere is mediated by circulation and surface processes; and changes in ventilation, stratification, and biological export can shift how much carbon is stored at depth versus available to the air. Also well-supported is that the Southern Ocean plays a disproportionate role as a ventilation gateway in the modern ocean and likely did so in the past. What remains debated is the exact recipe of changes during different glacial terminations—how much of the atmospheric CO₂ rise comes from Southern Ocean ventilation versus changes in North Atlantic overturning, how sea-ice extent and winds co-evolved, how biological pump efficiency shifted, and how these pieces varied by region and by termination. The debate is not whether physics allows circulation to move CO₂; it’s about attribution and timing: which levers moved first, how strongly, and how consistently across cycles.
The deep ocean, then, is not just a passive backdrop. It’s a dynamic reservoir whose connection to the atmosphere can tighten or loosen as circulation reorganizes. Glacial climates can, in several plausible ways, increase the isolation of deep carbon—through stronger stratification, more sea-ice capping and altered winds in the Southern Ocean that reduce ventilation, and changes in overturning that lengthen the deep ocean’s “storage time.” Deglaciation can reverse some of these, opening pathways that let deep, carbon-rich waters surface and vent CO₂, boosting atmospheric concentrations and reinforcing the transition.
Oceans store carbon, but life moves carbon too. Enter the land biosphere and soils.
On glacial timescales, land ecosystems and soils matter for atmospheric CO₂ because they are a large, climate-sensitive carbon store that can expand, contract, burn, and be buried. The core physical reason is simple: land carbon is mostly carbon kept out of the air—locked into wood, leaves, roots, and soil organic matter. Change climate, and you change plant growth, decomposition, fire, and where different ecosystems can exist. The atmosphere is the smallest “account” in this system, so when land shifts carbon into or out of storage, atmospheric CO₂ can move noticeably even if the land’s percentage change is modest.
Glacial climates tended to shrink forests and expand steppe and desert in many regions because colder temperatures, altered rainfall patterns, lower CO₂ availability for photosynthesis, and widespread ice-sheet influence all pushed water and energy limits in directions that favor grasslands and sparse vegetation over dense trees. Trees are carbon-expensive structures: they require long growing seasons, enough moisture to support tall canopies, and relatively stable conditions to accumulate wood over decades to centuries. In glacials, growing seasons shortened in mid-to-high latitudes, cold air held less moisture and often shifted storm tracks, and large ice sheets cooled and dried downwind regions. Where water became limiting or seasons became too short, forests retreated and were replaced by open steppe/tundra mosaics. That shift reduces land carbon storage for a mechanical reason: wood is a high-density, long-lived carbon pool, while grasses and sparse shrubs store less carbon aboveground and turn over faster. Less forest area and biomass generally means less carbon held on land and more potentially available to the ocean–atmosphere system.
Pathway (1) is the most direct: vegetation area and biomass changes. Climate sets the boundaries of biomes because plants must balance incoming energy (sunlight and temperature) and water supply (precipitation, soil moisture) against losses (evaporation and plant transpiration). In a glacial world, cooler air reduces potential evaporation, but precipitation patterns often shift and many regions become effectively drier, especially near ice sheets and in continental interiors. Add lower atmospheric CO₂—glacial minima around ~180–200 ppm range—and many plants operate with tighter carbon budgets: with less CO₂ available, plants typically need to open stomata longer to acquire carbon, which increases water loss. That “CO₂ starvation meets water stress” combination favors vegetation that can tolerate dryness and cold rather than dense forests. Physically, when forests retreat, the long-term carbon locked in trunks, branches, and deep roots declines; when they expand again in interglacials, carbon is drawn from the atmosphere and stored back into biomass.
Pathway (2) is soil carbon changes with temperature and moisture. Soil carbon is mostly dead organic matter—partially decomposed plant material protected in aggregates, bound to minerals, or stored in cold/wet conditions that slow microbial activity. Two physical “because” statements govern it. First, decomposition is a biological oxidation process, and its rate generally increases with temperature because microbial metabolism speeds up and enzymes work faster. Colder climates therefore tend to slow decomposition, which can allow soil carbon to accumulate—if plant inputs are maintained. Second, moisture controls oxygen availability: waterlogged soils limit oxygen diffusion, slowing decomposition and favoring carbon storage; very dry soils limit microbial activity too, but they also limit plant growth and litter inputs. Glacial climates often reduce plant inputs in many regions (less biomass), while also cooling soils (slower decay). The net effect on soil carbon can therefore vary by region: in some cold, moist areas, soils can retain or build carbon; in cold, dry steppe regions, reduced plant inputs and thin soils can mean lower soil carbon overall. The key is not hand-wavy “soil stores carbon,” but the balance of inputs (plant litter, roots) versus outputs (microbial respiration), both set by temperature and moisture.
Pathway (3) is wildfire, charcoal, and burial. Fire is a rapid oxidation pathway: it converts living biomass and surface litter into CO₂ on timescales of hours to days. Whether fire increases or decreases in glacials depends on two competing physical controls: fuel availability and dryness. Forests provide lots of fuel, but they may be wetter and less flammable; open steppe can be dry and windy but may have less total biomass. When conditions produce frequent burning, the immediate effect is to push carbon to the atmosphere. But fire also produces charcoal (black carbon)—a more chemically resistant form that can persist in soils or be transported and buried in sediments. Charcoal is essentially carbon moved into a slower pool: it is harder for microbes to decompose, so a fraction of burned carbon can be sequestered longer than the original plant material would have been. This creates a two-sided physical story: more fire can raise atmospheric CO₂ quickly, while simultaneously increasing the production of a long-lived carbon form that, if buried, reduces atmospheric CO₂ over longer timescales. The net sign depends on how much carbon is combusted versus how much is converted to persistent charcoal and successfully buried rather than re-oxidized later.
A concrete geographic example helps anchor ecosystem reorganization. Consider Europe during the Last Glacial Maximum. Large parts of northern Europe were covered by ice sheets, and south of the ice, climates were colder and often drier, with extensive steppe-tundra and sparse tree cover compared with modern interglacial forests. Mechanically, ice cover eliminates vegetation and soil carbon storage by removing the ecosystem entirely; periglacial landscapes support lower woody biomass; and colder, windier conditions can enhance dustiness and reduce soil development in exposed regions. The direction of land carbon storage change is therefore generally down relative to interglacial Europe, because forest biomass shrank and ice-covered land held essentially no living carbon. As deglaciation progressed and climates warmed and moistened, forests recolonized, soils developed, and carbon storage on land increased—drawing down atmospheric CO₂ relative to what it would have been if land had remained sparse.
It’s important to state the limits explicitly. Land carbon matters, but the deep ocean is bigger. The deep ocean contains vastly more carbon than the entire atmosphere and typically more than land biomass and soils combined, so on glacial timescales the largest potential “swing” reservoir is the ocean. That said, both still matter for atmospheric CO₂ because the atmosphere is small and responsive. A change of carbon storage on land that is modest compared with the ocean can still shift atmospheric CO₂ noticeably, especially when it happens in the same direction as ocean-driven changes. Also, land and ocean changes are coupled: altered dust supply from glacial deserts can fertilize parts of the ocean and affect biological carbon export; changes in freshwater input and climate can reshape ocean circulation; and CO₂ changes feed back on plant water-use efficiency and ecosystem boundaries. The deep ocean sets the largest storage capacity, but land ecosystems can change quickly (relative to ocean mixing) and can influence the trajectory and timing of atmospheric CO₂ during transitions.
Every pathway comes back to a physical “because.” Forests contract in cold/dry/low-CO₂ conditions because trees require long seasons, moisture, and sustained carbon uptake; losing forests reduces carbon stored in long-lived wood. Soil carbon changes because it is the balance of plant inputs versus microbial decay, both controlled by temperature and moisture. Fire changes carbon because combustion rapidly oxidizes biomass to CO₂, while charcoal formation and burial can move a fraction into a longer-lived pool. None of these require guessing motives or invoking vague “nature does X”; they follow from energy, water, and oxidation constraints.
Over geologic time, rocks set the background CO₂ level. That background decides whether ice sheets are even possible.
On million-year timescales, Earth’s climate is less like a thermostat you can nudge and more like a bathtub with two slow plumbing lines: a long-term inflow that adds CO₂ to the air–ocean system, and a long-term drain that removes it. When the inflow exceeds the drain for long enough, CO₂ accumulates and the planet tends toward warmer “hothouse” states; when the drain keeps up or wins, CO₂ stays lower and “icehouse” conditions become possible. Glacial cycles play out on top of this baseline, but the baseline decides what kind of game the climate system is even allowed to play.
The drain is silicate weathering, and the plain-English mechanism is surprisingly concrete. CO₂ in the air dissolves into raindrops and soil water, forming a weak acid (carbonic acid—think “slightly fizzy rainwater,” not battery acid). When this weakly acidic water flows over and through silicate rocks (many common crustal rocks), it reacts chemically with minerals and breaks them down. The key products—dissolved ions like calcium, magnesium, and bicarbonate—are carried by rivers to the ocean. In the ocean, those ingredients can be used (often with help from marine life) to form carbonate minerals that get buried as limestone and related sediments. The net effect, when you track all the steps, is that CO₂ is transferred from the atmosphere–ocean system into solid rock storage over long times. It’s not one reaction in one place; it’s a chain: CO₂ → weak acid in water → rock alteration → river transport → ocean chemistry → burial as carbonate.
A helpful physical picture is to imagine atmospheric CO₂ as a fine dust in the air that rain can “scrub” out—except instead of being physically washed out, it is chemically transformed into dissolved components that ultimately get locked into minerals. The scrubbing rate depends on how much “reactive surface” the planet exposes and how much water and heat are available to run the reactions. Warm, wet conditions accelerate many chemical reactions and increase runoff, so weathering tends to be faster in warmer climates. That creates a stabilizing logic: higher CO₂ warms climate, which tends to increase rainfall and chemical reaction rates, which tends to increase silicate weathering, which removes CO₂—like a slow, geological negative feedback.
The inflow is volcanic and tectonic outgassing, the primary long-term source that returns CO₂ to the surface system. CO₂ is stored in Earth’s interior and in carbonate-rich crust that gets recycled. At mid-ocean ridges, mantle material rises and melts; volcanic gases include CO₂ released to the ocean and atmosphere. At subduction zones, oceanic crust and sediments (including carbonates and organic carbon) are carried downward; some of that carbon is released back to the surface through arc volcanism and metamorphic degassing. If weathering and burial are the drain, outgassing is the faucet that refills the tub.
The crucial point is timescale. These are not decade processes; they are the multi-million-year carbon cycle sometimes called the carbonate–silicate cycle. Two quantitative anchors keep intuition honest. First, changes in global weathering and burial that significantly shift atmospheric CO₂ typically operate over ~10⁵ to 10⁶+ years (range) because you are altering large reservoirs through slow chemical processing and sediment accumulation. Second, tectonic changes that reorganize outgassing rates or mountain belts and thereby shift the long-term CO₂ balance often unfold over ~10⁶ to 10⁷+ years (range), because plate motions, seafloor production, and uplift/erosion are inherently slow. You can have faster spikes (large igneous provinces, rapid rifting episodes), but the baseline “set point” for hothouse vs icehouse is a long game.
Tectonics can shift the balance in concrete, mechanical ways. One classic example is mountain uplift increasing weathering. When a large mountain belt rises—think the Himalaya and Tibetan Plateau region as a type case—several things happen that increase the drain. Fresh rock is exposed by uplift and erosion, providing new mineral surfaces that react more readily than already-weathered material. Steep slopes and high relief accelerate physical erosion, continually revealing unweathered rock. Mountains also influence rainfall patterns; orographic uplift can increase precipitation on windward slopes, raising runoff and the delivery of weathering products to rivers. Put simply: uplift can turn on a bigger, better-supplied weathering machine. Over millions of years, that can pull down atmospheric CO₂, nudging the climate toward cooler states.
A different tectonic lever acts on the faucet: changes in seafloor spreading and volcanic outgassing. Faster seafloor spreading generally implies more volcanic activity at ridges and more CO₂ delivered to the surface system. Slower spreading implies less ridge outgassing. Over geologic time, periods associated with high spreading rates and extensive volcanism tend to coincide with warmer climates, while reduced outgassing can help allow CO₂ to fall—again, not as an instant cause, but as a background shift in the long-term balance between source and sink.
This slow baseline matters because ice sheets require a permissive boundary condition: CO₂ must be low enough that high-latitude summers can stay cool enough for snow to persist and for ice to grow, and geography must allow the necessary circulation patterns and land configurations. The last ~34 million years are central here because that’s when Earth entered the modern “icehouse” era with large Antarctic ice sheets (the Eocene–Oligocene transition, around 34 million years ago). The essential connection is: the long-term CO₂ baseline had declined enough—via the slow source–sink balance—to make sustained polar ice feasible, and geography was supportive. Antarctica’s position over the South Pole and its isolation by surrounding oceans helps; once a strong circumpolar flow develops, it can reduce poleward ocean heat transport to Antarctica, making it harder to melt ice. High Antarctic topography also matters because higher elevations are colder, so once ice begins to accumulate, it is easier to maintain.
Notice how this frames glacial cycles without confusing levels of causation. The orbital pacing of ice ages can only produce large ice sheets if the “background climate” is already in a regime where ice is stable. That regime is set by the multi-million-year carbon cycle and plate-tectonic geography. If CO₂ were much higher, the same orbital nudges would still happen, but they would act on a warmer planet where summer melt overwhelms snow survival—no big ice sheets, no dramatic 120-meter sea-level swings. The slow carbon cycle is the stage crew that builds the set and chooses the props; orbital cycles and fast feedbacks are the actors moving around on that set.
That sets the stage. Now explain the “lag” issue that people misunderstand.
Ice cores give us two parallel time series from the same cylinder of ancient snow: (1) CO₂ in trapped air bubbles and (2) a temperature proxy from the ice itself. The CO₂ number is measured directly from the composition of the ancient atmosphere sealed in bubbles (or in some cases in the air trapped in firn). Temperature is inferred indirectly from the ratio of stable isotopes in the water molecules that make up the ice (commonly δD or δ¹⁸O): in plain English, colder conditions change how heavy vs light water isotopes get transported and precipitated, and the ice preserves that fingerprint. These two records can be precisely measured, but their timestamps are not perfectly identical by default, because the air gets sealed after the snow fell.
That mismatch has a physical name: gas-age vs ice-age. Snow accumulates at the surface and compresses into firn (a porous, compacted snowpack) before turning into impermeable ice. While it’s still firn, air can mix through the pores, so the “air sample” is not locked in yet. Only when the firn densifies enough does it close off bubbles and trap air. That means the ice at a given depth is older than the gas trapped within it by some offset—the enclosure time—which depends on temperature, accumulation rate, and firn properties. Even after careful modeling and cross-checking, this introduces an uncertainty envelope and can create apparent leads/lags of order decades to centuries, especially in low-accumulation, cold sites. So when you hear “CO₂ lags temperature,” part of what’s being discussed is the climate system’s dynamics, and part is the measurement reality that the two signals are recorded and sealed in slightly different ways.
Now to the bigger logical point: even if a lag is real, it does not mean CO₂ is irrelevant. The embedded wrong belief here is “lag disproves causation,” corrected immediately by distinguishing initiation from amplification. In many deglaciations, the initial warming begins with changes in orbital geometry that alter seasonal and regional sunlight—especially summer insolation at high northern latitudes—setting off ice melt and circulation shifts. CO₂ can respond after that initial nudge (so it “lags”) and still be a major causal factor in the subsequent warming because it changes Earth’s heat loss to space. A feedback can be causal without being first. If you push a swing with your hand, your hand acts first; the swing’s motion then changes the tension and timing of the chains, which then affects how the swing continues. The chain tension didn’t initiate the motion, but it still shapes the motion in a causal way once things are moving.
Here’s the clean feedback logic. Step 1: an external forcing (orbital insolation changes) nudges regional temperature and ice melt. Step 2: that climate shift changes internal carbon reservoirs—especially the ocean—causing CO₂ to rise. Step 3: rising CO₂ increases greenhouse trapping, reducing outgoing infrared radiation until the system warms enough to restore balance. Step 4: that extra warming reinforces the original direction, helping melt ice, shift winds, and reorganize circulation further. A lag is exactly what you’d expect in a coupled system where the first domino (insolation) knocks over slower dominoes (ocean circulation and carbon partitioning), and those slower dominoes then feed back strongly.
Quantitatively, the CO₂ signal in ice cores across glacial–interglacial cycles is large and repeatable: roughly ~80–100 ppm (range) from typical glacial minima to interglacial maxima (often framed as about ~180 to ~280 ppm (range)). Lags between Antarctic temperature proxies and CO₂ during deglaciations are not a single fixed number; they vary by event, by site, and by how the gas-age/ice-age correction is handled. A defensible way to state it is that inferred lags are often from near-zero up to several hundred years (range), sometimes discussed on the order of ~0–800 years (range) rather than thousands. Meanwhile, the deglaciation itself—the transition from full glacial conditions into an interglacial—typically unfolds over ~5,000–10,000 years (range). So even a few-hundred-year lag is small compared with the multi-millennial duration of the transition that CO₂ helps amplify and sustain.
A concrete deglaciation narrative, stated cautiously, goes like this. Orbital variations shift the seasonal pattern of sunlight, and at some point summer insolation in the Northern Hemisphere becomes favorable for net ice loss: summers are intense enough (or long enough) that winter snow and ice don’t fully survive. That initiates regional warming and melt, which reduces surface reflectivity (less bright ice, more dark land/ocean), adding an ice-albedo boost. Meltwater and changing winds can then alter ocean circulation—especially how the Southern Ocean and other key regions ventilate the deep ocean. As circulation and surface conditions change, the ocean’s carbon partitioning shifts: warmer surface waters can hold less CO₂ (a Henry’s-law intuition), and changes in upwelling and ventilation can expose deep, carbon-rich waters to the surface where they can outgas. Land ecosystems also reorganize as climate warms: forests expand into formerly cold/dry regions, soils thaw and change respiration rates, and the net land carbon balance can shift. These ocean and land responses release some carbon to the atmosphere, raising CO₂. Once CO₂ rises, it exerts a global radiative effect: it reduces outgoing heat loss and contributes additional warming that is not confined to a particular season or latitude. That broader warming helps continue ice-sheet retreat and reinforces the transition. The evidence-based caution is that the relative contributions—ocean vs land, which ocean pathways dominate, how winds and sea ice evolved—are still active research topics, but the coupled sequence “orbital nudge → initial warming/melt → carbon-cycle response → CO₂ amplification” is the coherent physical framework that fits the main observations.
A second embedded wrong belief is “CO₂ changes are purely volcanic,” corrected immediately by scale and timing. Volcanoes and tectonic outgassing matter enormously on million-year baselines, but the characteristic ~80–100 ppm swings on glacial timescales align with reorganizations of ocean–ice–biosphere dynamics and occur on millennial pacing that is hard to attribute primarily to large, systematic volcanic rate changes. That doesn’t mean volcanoes are irrelevant; it means they are not the best single-knob explanation for these repeated, orbit-paced swings recorded in ice cores.
So what should you do with “lags” when you see them? Treat them as diagnostic, not dismissive. A lag can tell you which part of the machine moved first in that episode—often the insolation-driven regional changes—but it cannot tell you that the later-moving part had no causal power. In a coupled system, the whole point of a feedback is that it responds to a change and then changes the outcome.
So CO₂ is a memory and an amplifier. Put it back into the full machine: orbit + feedbacks + oceans + ice.
CO₂ is the loudest supporting actor in the ice-age story because it usually doesn’t deliver the opening line, but it does make the scene make sense. Orbital shifts redistribute sunlight in ways that can favor ice growth or melt in particular places and seasons, especially summer at high northern latitudes. Yet the outcome is a planet-scale reorganization—global temperatures shifting by several degrees, sea level moving by roughly a hundred meters, ice sheets expanding and collapsing across continents. CO₂’s role is to magnify and stabilize a climate shift that was initiated elsewhere, turning a regional, seasonal nudge into a global, persistent state change with the right amplitude and coherence.
The key is that CO₂ is a global lever. Insolation changes are geographically picky: they push hardest in certain latitudes and seasons, and the global annual-mean energy change is small. CO₂, once changed, is not picky at all. It mixes through the atmosphere and alters the rate at which the entire planet loses heat to space. That matters because it links regions that orbital forcing affects very unevenly. A local summer cooling that helps snow survive in Canada or Scandinavia can start ice expansion, but it cannot by itself guarantee a global response unless the rest of the system is recruited. When CO₂ falls, nights, winters, tropics, and oceans all feel a shift in the background heat budget—not equally, but everywhere. That is how a climate signal that begins as “summer melt changed here” becomes “the whole planet’s energy balance changed,” which is what you need to explain why glacial–interglacial transitions look like synchronized global reorganizations rather than isolated regional quirks.
This also dissolves a false choice people cling to: “CO₂ is either everything or nothing,” corrected immediately by basic systems logic. In glacial cycles, CO₂ is rarely the sole trigger, but it is also not optional decoration; it is the amplifier and stabilizer that helps determine how far and how long the climate moves once the first push happens. A forcing can start the motion, a feedback can govern the magnitude. The climate system does not award causality points only to whoever moved first.
Another confusion comes from language rather than physics: “CO₂ is a pollutant only,” corrected immediately by the fact that CO₂ is first and foremost a radiatively active gas in an energy-budget problem. It can be discussed in policy contexts, but the glacial story doesn’t require politics to be true. More CO₂ increases the atmosphere’s infrared opacity in key wavelength bands, reducing outgoing heat until temperatures adjust. That mechanical relationship is why CO₂ is a powerful climate lever whether it is being changed by human emissions, by ocean circulation, or by slow geologic processes. Calling it a “pollutant” or a “plant food” doesn’t tell you what it does to the planet’s heat loss; radiative transfer does.
A third wrong belief flips the logic backward: “ice ages prove CO₂ doesn’t matter,” corrected immediately by noticing what ice ages actually show. They show that CO₂ and temperature move together with large, repeated swings—roughly ~80–100 ppm (range) for CO₂ between glacial and interglacial states—while the climate flips between very different equilibria. That co-variation is exactly what you expect if CO₂ is a strong feedback in a system paced by orbital forcing and mediated by oceans and ice. If CO₂ truly didn’t matter, you would need other feedbacks alone to explain why small orbital changes are accompanied by global, sustained changes in temperature and sea level. You might still build a story with ice albedo and water vapor, but you would be throwing away a globally acting lever that the records show changed dramatically at the right times.
CO₂’s “supporting actor” role is inseparable from the ocean’s role because most of the carbon isn’t in the air—it’s in seawater and sediments—so the next misconception to retire is “oceans just absorb CO₂ passively,” corrected immediately by how dynamic the ocean is. The surface ocean is not a static sponge; it is a chemically buffered, wind-stirred interface that can either draw down or release CO₂ depending on temperature, alkalinity, and how quickly surface waters are connected to the deep reservoir. The deep ocean is an active warehouse whose doors are opened and closed by circulation. During glacials, stronger stratification, more sea ice capping in key regions, and altered overturning can keep carbon sequestered at depth; during deglaciations, shifts in winds, sea ice, and upwelling can ventilate carbon-rich deep waters and raise atmospheric CO₂. “Passive” would mean the ocean merely follows the atmosphere. In reality, the ocean often leads the carbon bookkeeping because it holds the inventory that the atmosphere samples.
Biology is part of that bookkeeping too, so the claim “biology is irrelevant” needs correcting immediately: living systems move carbon between reservoirs by controlling how much carbon is stored as biomass and soils, and by altering how much carbon the ocean exports to depth as sinking organic matter. Glacial climates tend to reorganize ecosystems—forests shrink in many regions, steppe expands, soils change with temperature and moisture, fire regimes shift—changing land carbon storage. In the ocean, plankton and the “biological pump” affect how much carbon leaves the surface layer and is stored in the deep sea. These are physical “because” stories: biology matters because it is a mechanism for moving carbon into forms and places that exchange slowly with the air.
What, then, does this chapter not claim? It does not claim CO₂ is the sole trigger of ice ages. Orbital changes set the pacing and provide the geographically specific push that matters for ice-sheet mass balance, and CO₂ typically responds as part of the system’s internal adjustment. It does not claim CO₂ alone explains the timing of abrupt events—sharp climate jumps on century scales involve thresholds in ocean circulation, sea ice, and ice-sheet dynamics that CO₂ can influence but does not uniquely schedule. And it does not claim CO₂ acts in isolation: CO₂ interacts with oceans (as the main carbon reservoir and as a heat transporter), with ice (through albedo and meltwater feedbacks), and with water vapor and clouds (which respond to temperature). CO₂ is powerful precisely because it is embedded in a network: it’s a global knob that the rest of the system can turn and that then turns the rest of the system in return.
If you want the cleanest summary, it is this: orbital forcing often supplies the initial asymmetry—where melt wins or loses—while CO₂ supplies the global coherence—why the whole planet shifts into a new, sustained energy-balance regime. CO₂ spreads through the air on timescales short compared with ice sheets and deep-ocean turnover, so it can take a regional, seasonal perturbation and make it “felt” everywhere by changing the planetary heat budget. That is what an amplifier does: it does not invent the signal, but it determines whether the signal becomes a whisper that fades or a sustained change that reorganizes the entire system.
Next, we follow the biggest amplifier’s partner: the oceans, Earth’s heat bank, whose circulation decides whether warmth and carbon stay locked away or get spent at the surface.