The Arctic is too large, dark, cloudy, and expensive to observe continuously from the surface. Satellites solve the coverage problem, but no single instrument can answer every question. Each measurement trades some combination of detail, coverage, penetration, and certainty.

Why photographs are not enough

Winter sea ice covers about $14\text{ million km}^2$. At $80^\circ\mathrm{N}$, the Sun does not rise for 131 days; at the pole, there are 183 days without sunrise. These values describe the absence of sunrise, not complete darkness, and the calculation ignores twilight and atmospheric refraction. The Arctic Ocean is also cloudy on most days, especially in summer when open water adds moisture to the air.

A visible camera needs reflected sunlight and a clear path through the atmosphere. When both are available, it can show metre-scale detail, true colour, and individual melt ponds. When either is missing, it can fail completely.

Thermal infrared removes the sunlight requirement because every surface emits radiation according to its temperature. It can reveal surface temperature, thin ice, and leads losing heat, but clouds still block the view.

The five main observation questions are:

1. Where is the ice? Measure extent and concentration.
2. What kind of ice is it? Distinguish first-year from multi-year ice.
3. How thick is it? This is the hardest quantity.
4. Where is it going? Measure drift and deformation.
5. What is on top? Detect snow, melt ponds, and melt.

No single instrument answers all five.

One electromagnetic toolbox

Remote-sensing instruments differ in two basic choices:

1. which wavelength they observe;
2. whether they receive natural radiation or send a signal and measure its return.

A representative cloud droplet is about $10\,\mu\mathrm{m}$ across. Green light has a wavelength near $0.5\,\mu\mathrm{m}$, so the droplet is about 20 times larger than the wave. At $19\,\mathrm{GHz}$, a microwave has a wavelength near $16\,\mathrm{mm}$. This example rounds the droplet-to-wave ratio to $1/1500$. The rounded dimensions give $1/1600$, while $\lambda=c/f\approx15.78\,\mathrm{mm}$ gives roughly $1/1578$. This scale difference lets selected microwave wavelengths pass through cloud far more effectively than visible light or thermal infrared.

The comparison is qualitative. "Cloud-penetrating" does not mean every microwave measurement is unaffected by the atmosphere.

Passive and active microwave

A passive microwave sensor listens for the faint microwave radiation emitted naturally by the surface. It sends nothing toward Earth. This produces broad, frequent coverage at coarse resolution.

An active microwave sensor, or radar, sends a pulse and measures the returning echo. Supplying its own illumination gives much finer detail, but over less continuous coverage.

The long passive-microwave record begins in 1979. In 2026, it spans 47 years and provides the basis for daily Arctic concentration and extent fields through cloud and polar night.

Brightness temperature separates water and ice

Brightness temperature describes how bright the measured microwave emission is, expressed as a temperature. It is not necessarily the physical temperature of the surface.

Water and ice can have nearly the same physical temperature but very different microwave brightness temperatures. Open water reflects microwaves strongly and emits weakly, with a representative emissivity near 0.5. First-year ice emits more strongly, with a representative emissivity near 0.95. Multi-year ice lies between them because brine drainage leaves air inclusions that scatter radiation inside the ice.

Typical values are:

These are typical values that vary with season. They are not universal calibration constants.

At $19\,\mathrm{GHz}$, first-year ice is $102\,\mathrm{K}$ brighter than open water. The contrast is large enough to estimate how much of a mixed pixel is ice. Multiple frequencies add information about ice type: from 19 to $37\,\mathrm{GHz}$, the multi-year-ice value drops by $25\,\mathrm{K}$, compared with only $8\,\mathrm{K}$ for first-year ice.

The brine and air-inclusion mechanism connects directly to §4, Why brine matters.

From brightness temperature to concentration

Assume the microwave signal from a mixed pixel is a linear blend of its water and ice signals. We can then use the measured brightness temperature to estimate the ice fraction.

A simplified retrieval treats open water and solid first-year ice as two endpoints. Let

• $T_b$ be the pixel brightness temperature;
• $T_w=136\,\mathrm{K}$ be the open-water endpoint;
• $T_i=238\,\mathrm{K}$ be the solid-ice endpoint.

The estimated ice fraction is

$$f_{\mathrm{ice}} =\frac{T_b-T_w}{T_i-T_w} =\frac{T_b-136\,\mathrm{K}}{102\,\mathrm{K}}.$$

For a pixel reading $187\,\mathrm{K}$,

$$f_{\mathrm{ice}} =\frac{187-136}{238-136} =\frac{51}{102} =0.50.$$

The simplified interpretation is half ice and half water.

For a reading of $217\,\mathrm{K}$,

$$f_{\mathrm{ice}} =\frac{217-136}{238-136} =\frac{81}{102} \approx0.79,$$

or about 80% ice.

Real algorithms combine several channels. The endpoint calculation is a teaching model, and it becomes least reliable during melt season. Melt ponds put liquid water on top of ice, so a floe no longer behaves like the solid-ice endpoint. This connects to §3, Melt onset and melt ponds.

Concentration, extent, and area

Sea-ice concentration ($f_i$) is the estimated fraction of pixel $i$ covered by ice.

Ice extent ($E$) counts the full area $A_i$ of every pixel whose concentration is at least 15%:

$$E=\sum_i A_i\,\mathbf{1}\!\left(f_i\ge 0.15\right).$$

Ice area ($A$) weights each pixel area $A_i$ by its estimated ice fraction $f_i$:

$$A=\sum_i A_i f_i.$$

Here, $A_i$ is the area of pixel $i$, and $f_i$ is its estimated ice fraction.

Suppose four neighbouring pixels are each $25\,\mathrm{km}\times25\,\mathrm{km}=625\,\mathrm{km}^2$, with concentrations of 95%, 80%, 40%, and 10%.

Extent counts the first three pixels:

$$E=3(625)=1875\,\mathrm{km}^2.$$

Area includes every fractional contribution:

$$\begin{aligned} A&=(0.95+0.80+0.40+0.10)(625)\\ &=2.25(625)\\ &=1406.25\,\mathrm{km}^2. \end{aligned}$$

The same ice therefore gives an extent of $1875\,\mathrm{km}^2$ and an area of about $1406\,\mathrm{km}^2$. Comparing extent in one year with area in another compares different quantities. Extent is less sensitive than area to melt-pond errors, but it is not immune: a biased concentration can still cross the 15% threshold.

The 25 km blind spot

Broad daily coverage costs spatial detail. A typical passive-microwave pixel is about 25 km across. It may correctly report 96% concentration over $625\,\mathrm{km}^2$ while hiding the location and shape of the remaining 4%.

That missing geometry matters. Leads range from metres to kilometres wide, while pressure ridges can be only tens of metres wide. Narrow open water can carry a large share of the heat loss even when it occupies little area. The average is not wrong; it answers a different question.

For the physical role of these features, see §4, Divergence: leads and polynyas and §4, Convergence: rafting and ridging.

Imaging radar reveals structure

Imaging radar supplies its own pulse, so it does not need sunlight. It looks sideways across a swath: echoes from the near edge return before echoes from the far edge, giving the range dimension of the image.

A physical antenna able to resolve metres from roughly 700 km altitude would be impractically long. Synthetic aperture radar, or SAR, combines coherent echoes collected as the satellite moves, producing the effect of a much longer antenna. The resulting pixels can be about 10-40 m across, compared with about 25 km for passive microwave.

Radar return depends strongly on:

• roughness at the wavelength scale;
• the electrical properties of the target, including salinity, wetness, and internal structure.

A simplified reading key is:

Radar brightness is return strength, not visible colour. These signatures are useful classroom patterns, not unique answers under every viewing geometry, wavelength, polarization, or surface condition.

Repeated radar images also turn texture into motion. If the same floe moves by $\Delta\mathbf{x}$ during a time interval $\Delta t$, its velocity is approximated by

$$\mathbf{v}\approx\frac{\Delta\mathbf{x}}{\Delta t}.$$

Matching many floes produces a drift map rather than a single buoy track.

Wavelength changes what radar can see

Different radar bands probe different roughness scales and depths:

Shorter waves emphasize the snow surface and fine roughness. Longer waves penetrate farther into snow and ice but usually give less spatial detail. The values are approximate, not exact band boundaries or universal penetration depths.

Thickness comes from freeboard

Satellites do not measure sea-ice thickness directly. An altimeter measures the height of a reflecting surface above sea level. The part of floating ice above the water is its freeboard.

A simple rule of thumb is

$$H_i\approx10f,$$

where $H_i$ is ice thickness and $f$ is ice freeboard in the same unit. A freeboard of 20 cm therefore suggests roughly 2 m of ice. This is an approximation, not a complete hydrostatic calculation.

The accompanying example is not internally exact: it labels 10 cm of snow, 20 cm of freeboard, and 2.21 m of ice, while the plotted point at 20 cm lies near 1.9-2.0 m. The mismatch is unresolved, so the 2.21 m column and the ten-to-one rule should be treated as separate approximations rather than one exact conversion.

A radar and a laser do not see the same surface:

• a laser reflects from the top of the snow at elevation $h_{\mathrm{laser}}$;
• a radar return comes from roughly the snow-ice boundary at elevation $h_{\mathrm{radar}}$.

In an idealized case, their difference would estimate snow depth $h_s$:

$$h_s\approx h_{\mathrm{laser}}-h_{\mathrm{radar}}.$$

The problem is that radar does not stop at one clean interface, especially when snow is warm or salty. Snow depth remains one of the hardest quantities to retrieve, and that uncertainty feeds directly into thickness.

Small elevation errors become much larger thickness errors:

• $1\,\mathrm{cm}$ of freeboard error can become $10\,\mathrm{cm}$ of thickness error;
• $10\,\mathrm{cm}$ of snow-depth error can become $30\,\mathrm{cm}$ of thickness error.

The tenfold factor belongs to the freeboard example, not to every source of uncertainty. The importance of ice and snow as insulation connects to §2, The surface energy balance and §4, Snow increases the insulation.

Match the instrument to the question

Resolution and coverage pull in opposite directions

• Imaging
• Broad coverage
• Profile altimeter

Approximate positions used in the resolution-versus-coverage plot

Instrument	Approximate placement
High-resolution SAR	about 10 m resolution and roughly 80 km swath
Wide-swath SAR	about 40 m resolution and several hundred kilometres of swath
Laser altimeter	about 15–20 m resolution along a narrow profile
Radar altimeter	about 300 m resolution along a profile a few kilometres wide
Optical / thermal	about 500 m resolution and coverage of a few thousand kilometres
Passive microwave	near 25 km resolution with daily, broad Arctic coverage
Scatterometer	near 25 km resolution and roughly 1,000 km coverage

The empty upper-left region is the combination scientists would like: fine resolution and broad coverage. In practice, instruments occupy different parts of the plot.

The point is instrument complementarity. Passive microwave supplies the long, broad record. SAR resolves the geometry, type, deformation, and motion hidden inside a coarse pixel. Altimetry measures freeboard and surface elevation, from which thickness is inferred. Snow depth, thin-ice thickness, melt-season retrievals, and small-scale deformation remain difficult.