Sea ice does not simply freeze in place and stay there. Cold air first removes enough heat for crystals to form near the ocean surface. Once those crystals join into a sheet, more water freezes onto its underside. Summer warmth melts the ice from above and below. Snow slows winter growth by insulating the ice and can delay summer melt by reflecting sunlight. Winds and currents also move the ice across the Arctic, pulling floes apart in some places and piling them on top of one another elsewhere. Thermodynamics describes the heat transfer that freezes and melts the ice; dynamics describes how forces move and deform it.

Why seawater freezes differently from a lake

A freshwater lake can begin freezing at the surface before all the water below has cooled to $0\,^\circ\mathrm{C}$. Fresh water is densest near $4\,^\circ\mathrm{C}$. Water that cools below that temperature becomes lighter, so it stays at the surface instead of sinking. This surface layer can then cool from $4\,^\circ\mathrm{C}$ to its freezing point at $0\,^\circ\mathrm{C}$ without being replaced by warmer water from below.

Ordinary seawater behaves differently because it does not reach a density maximum before it freezes. At salinities above about $24.7\,\mathrm{g\,kg^{-1}}$, seawater keeps getting denser as it cools toward its freezing point. The cooled surface water sinks and mixes with warmer water below, so the atmosphere must remove heat from the whole mixed layer before ice can form. Cooling only a thin surface layer is not enough.

Dissolved salt also lowers the freezing point. Seawater with a salinity of roughly $33$ to $35\,\mathrm{g\,kg^{-1}}$ begins freezing near $-1.8\,^\circ\mathrm{C}$ rather than $0\,^\circ\mathrm{C}$. A nearby lake can freeze weeks before the sea under similar weather because the ocean has to cool farther while sinking surface water keeps bringing up heat from below.

From crystals to an ice sheet

Once the upper ocean reaches its freezing point, tiny plates and needles of ice called frazil begin to form in the water. Waves and currents then determine how those crystals come together.

In calm water, frazil collects in a soupy layer called grease ice. It then freezes into nilas, a continuous, elastic crust no more than about $10\,\mathrm{cm}$ thick that bends with waves and swell. In rough water, waves jostle frazil and grease ice into circular pieces called pancake ice. As the pieces collide, ice builds up around their edges and forms raised rims. Nilas can also break into pancakes if waves return.

Either route can eventually close the gaps and form a continuous sheet. The sheet shields the water below from most wave motion. Ice then thickens mainly as crystals grow downward from its underside, producing the vertical structure called congelation ice.

From open water to growing sea ice

Why ice growth slows

When water freezes onto the bottom of the ice, it releases latent heat. More water can freeze only after that heat travels upward through the existing ice and escapes into the cold atmosphere. A simple conductive flux has the form

$$q_c\approx k_i\frac{T_f-T_s}{H_i},$$

where $k_i$ is the thermal conductivity of ice, $T_f$ is the temperature at the freezing interface, $T_s$ is the ice-surface temperature, and $H_i$ is ice thickness. As $H_i$ increases, heat has farther to travel through the ice. The upward heat flux drops, so water freezes onto the bottom more slowly.

This is a negative feedback. Thin ice loses heat quickly and grows quickly, but the new ice then acts as insulation and slows further growth. Under the idealized assumptions behind Stefan's law, thickness increases approximately with the square root of time rather than by the same amount each day.

The lecture's constant-temperature curves show how large this slowdown can be:

• after $180$ days at $-30\,^\circ\mathrm{C}$ with no snow, the idealized thickness is roughly $2.5\,\mathrm{m}$;
• at $-20\,^\circ\mathrm{C}$, it is about $2\,\mathrm{m}$;
• in one example, roughly $16\,\mathrm{cm}$ forms on the first day, while the increment on day 180 is less than $1\,\mathrm{cm}$.

These model results do not set a universal $2\,\mathrm{m}$ growth limit. Natural ice thickness also depends on the length and severity of winter and on how much snow covers the ice. Ocean heat, summer melt, and the piling up of floes change it too. Historically, old undeformed Arctic ice could approach an annual equilibrium thickness near $3\,\mathrm{m}$, where winter growth roughly balanced summer loss. Ridging can make the ice much thicker than that.

Snow increases the insulation

Snow slows the loss of heat through the ice. It conducts heat roughly seven times less effectively than sea ice, so a shallow snow layer can add as much thermal resistance as a much thicker layer of ice. In the lecture comparison, $10\,\mathrm{cm}$ of snow insulates about as much as $67\,\mathrm{cm}$ of ice. On the same scale, $15\,\mathrm{cm}$ of snow adds about as much resistance as $1\,\mathrm{m}$ of ice.

When the snow falls matters too. Snow that arrives after a thick ice sheet has formed cannot reverse that earlier growth. If the same snow falls just after freeze-up, it insulates the thin ice for most of the winter, so much less ice forms underneath by spring. In summer, snow can delay melting because it reflects sunlight and shields the ice. The same layer that slows winter growth can protect the ice from summer melt.

Sea ice is not simply frozen seawater

A growing ice crystal cannot easily fit dissolved salt into its crystal lattice. It rejects most of the salt into the ocean as it freezes. Some concentrated seawater remains trapped between the crystals as liquid brine, held in pockets and channels. Sea ice is not fresh ice filled with solid salt crystals. It is a mixture of nearly fresh ice crystals and liquid, concentrated brine, with air inclusions becoming more common as the ice ages.

As the surface water cools and ice forms, rejected salt makes the water beneath the ice saltier and denser. This dense water can sink and mix the upper ocean, so sea-ice formation also contributes to water-mass formation and larger-scale ocean circulation. A smaller share of the brine remains trapped inside the ice.

The amount of liquid inside the ice changes with temperature even before the ice loses any of its total salt:

• during cooling, more water freezes out of each pocket, making the remaining brine saltier and smaller in volume;
• during warming, some surrounding ice melts into the brine, so the liquid volume grows;
• once enough pockets connect, brine and surface meltwater can drain through channels.

In typical first-year ice, warming toward about $-5\,^\circ\mathrm{C}$ can raise the liquid fraction into the range where separate pockets begin to connect. The exact threshold depends on the ice's bulk salinity and structure, so there is no single temperature that applies to every floe. What matters physically is that warming enlarges the liquid inclusions until they connect into drainage paths.

Why brine matters

Strength

Liquid brine pockets break up the solid framework that carries a load through the ice. As the ice warms, the brine volume usually increases, so the ice becomes less stiff and weaker. Two floes with the same thickness can support different loads. Their strength also depends on salinity, cracks, snow loading, and deformation history. Older ice is often fresher and mechanically different from first-year ice, but age alone cannot show whether ice is safe for travel.

Radar response

Liquid brine absorbs microwave energy, so radar penetrates salty first-year ice only weakly and much of the echo comes from the surface. Ice that survives several summers loses brine through drainage and flushing and develops more air inclusions. In cold, dry conditions, radar can travel farther into this older ice, where boundaries between air and ice return energy through volume scattering.

These physical differences can help separate ice types, but brightness alone is not enough. Wavelength, polarization, viewing angle, surface roughness, snow, temperature, and liquid water also change the return. §5 explains how those factors can make similar ice look different or different surfaces look similar.

Age and salinity

Sea ice loses salt through gravity drainage, pressure-driven expulsion, and summer flushing. First-year ice commonly has a bulk salinity of several grams per kilogram. Ice that survives repeated summers is usually much fresher, often around $1$ to $3\,\mathrm{g\,kg^{-1}}$. Multi-year ice can still contain some brine, but it generally contains much less liquid brine than first-year ice.

A first-year ice core often has a C-shaped salinity profile. The upper ice retains salt that was trapped during rapid early growth. The middle is fresher, while the bottom remains relatively salty because it is still growing. In summer, meltwater moves through connected channels and flushes out more brine, especially from the upper ice.

The table describes common tendencies, not fixed identification rules. A deformed first-year ridge can be thicker than nearby level multi-year ice. Warm or wet surfaces can also reverse the radar contrast that is common under cold, dry conditions.

The loss of old ice

The lecture compares 1985 with 2023. Over that period, ice older than four years fell from about 33% of the Arctic pack to only about 3%, while first-year ice came to dominate. The exact percentage changes with the week chosen and the geographic mask, but very old ice now covers only a small fraction of its former area.

Ice age records how many summers a floe has survived. A floe that lasts for several years can grow during several winters, become thicker when other floes press against it, and lose salt as summer meltwater flushes its brine channels. When old ice disappears, the pack loses ice that is usually thicker, fresher, and mechanically different from young ice. The change is not just a different colour on an age map.

How the pack moves

Sea ice is a collection of floes, not a rigid shell. Individual floes can move, rotate, separate, and collide. Fridtjof Nansen demonstrated that motion with the Fram. In 1893, the specially rounded ship entered the ice north of Siberia. It emerged near Svalbard about three years later after drifting across the Arctic with the pack.

Wind provides much of the force over short periods. Ocean currents, internal stress within the pack, coastlines, and Earth's rotation change how the ice responds. Under free drift, a common rule of thumb is that ice moves at about 2% of wind speed and, in the Northern Hemisphere, roughly $20^\circ$ to $40^\circ$ to the right of the wind. Those numbers are not universal. Closely packed floes resist one another, coastlines redirect motion, and ocean drag changes both the speed and direction.

Two broad circulation patterns move much of the Arctic pack:

• the Beaufort Gyre circulates ice clockwise on the Canadian and Alaskan side, sometimes retaining it for several years;
• the Transpolar Drift carries ice from the Siberian side across the central Arctic toward the Atlantic.

The principal exit from the Arctic is Fram Strait, between Greenland and Svalbard. Long-term estimates put the annual exported area near $880{,}000\,\mathrm{km^2}$, about one tenth of the ice-covered Arctic Basin. Nares Strait and channels through the Canadian Arctic Archipelago export additional ice. This export removes ice from the Arctic pack even if the floes do not melt until they reach the Atlantic.

Convergence: rafting and ridging

When winds and currents push floes together, the solid ice cannot compress enough to absorb all the motion. Thin sheets can slide over one another in rafting. Thicker floes break into blocks, and the blocks pile above and below the waterline to form a pressure ridge.

The visible part of a ridge is its sail, while the much larger submerged part is its keel. As the blocks settle into buoyant balance, most of the displaced ice remains below the water. In a field example from the lecture, first-year ice near Cambridge Bay piled into an approximately $8$ to $9\,\mathrm{m}$ deformed feature even though the undeformed first-year ice nearby was much thinner. Large keels can reach $10$ to $30\,\mathrm{m}$.

Cold weather adds ice slowly to the bottom of a level floe and usually produces a sheet only a few metres thick. Convergence can instead stack several floes and broken blocks in one place, producing the thickest ice in the pack without freezing that full thickness from seawater. A thick ridge may survive summer melt, but it can also block travel routes and create a serious navigation hazard.

Divergence: leads and polynyas

When floes move apart, they expose a crack or channel called a lead. In winter, water near $-1.8\,^\circ\mathrm{C}$ may suddenly be exposed to air near $-30\,^\circ\mathrm{C}$. Heat and moisture then escape quickly. New frazil forms, and tens of centimetres of thin ice can grow across the opening in one day before the insulating ice slows further growth.

Leads may cover only a few percent of the winter pack, but open water or newly formed ice can lose heat roughly one hundred times faster than nearby thick, snow-covered ice. Even a small lead can release a large share of the local heat when a coarse map shows almost complete ice cover. Leads may offer ships a route through the pack, but they move, close, and refreeze.

A polynya is a non-linear opening enclosed by ice. Many polynyas return in the same region without staying ice-free all the time. In a latent-heat polynya, wind repeatedly pushes newly formed ice away from a coast, exposing more water to freeze. In a sensible-heat polynya, heat supplied by the ocean below prevents or slows freezing. A polynya that repeatedly forms and exports new ice acts as an ice factory. Each round of freezing rejects brine, stirs the water, and exposes a new surface to the atmosphere. The recurring open water and mixing also supply productive habitat, making many polynyas biological hotspots.

Thermodynamics and dynamics together

Cold air removes heat, so seawater freezes onto the underside of a floe; summer warmth and ocean heat melt ice away. Winds and currents then move the existing floes. They expose water when the floes separate and stack ice into deep ridges when the floes converge.

A recently opened lead may have only a thin layer of new ice because freezing has just begun. Beside it, convergence may have stacked first-year floes into a ridge many metres thick. Older ice is thicker on average because it has survived more growth seasons, but age alone does not reveal whether that ice is level, rafted, or ridged.

Why the oldest ice is disappearing

Several physical changes reduce the number of floes that survive long enough to become old ice:

1. warmer summers melt more ice, so less first-year ice survives to enter the multi-year category;
2. later autumn freeze-up shortens the growth season, leaving the new winter cover thinner;
3. thinner floes respond more readily to wind and can reach export gateways before they have time to age;
4. loss through Fram Strait removes ice from the Arctic even when that ice survives local summer melt;
5. a younger, thinner pack breaks and opens more readily, while melt ponds and exposed ocean lower the summer surface albedo and increase absorbed solar energy.

More summer melt leaves fewer floes to survive into the next year. Later freeze-up also gives their replacements less time to thicken, and those thin floes are easier for wind to break apart and carry toward Fram Strait. When leads, melt ponds, and open water replace bright snow-covered ice, the darker surfaces absorb more sunlight and promote further melt. The result is not only a smaller ice-covered area. The remaining pack has a different age distribution, salinity, strength, thickness, mobility, and pattern of leads and ridges, so an area map alone cannot describe how the Arctic ice cover has changed.