A map of arrows over polar sea ice does not show how much of the ocean is frozen, how thick the ice is, or how the wind blows. It usually describes the displacement of recognizable ice patterns between observations. Sea-ice drift is a motion estimate built from images, defined over a time interval and a coordinate system. Translating the arrows into meaningful language requires protecting the difference between displacement, velocity, concentration, growth and loss.
Imagine a data file giving 8.64 kilometres of eastward ice displacement over 48 hours, a northward component, and an uncertainty value at a grid point. The first arithmetic step is to turn kilometres and hours into an average speed if that is what the reader needs. The harder step is to establish what ice pattern was tracked, whether that grid location was observed or filled, and whether the product’s reported uncertainty applies to a displacement component rather than a complete trajectory.
This tutorial uses first-party EUMETSAT Ocean and Sea Ice Satellite Application Facility documentation, including its recent near-real-time sea-ice drift product and longer-term climate record, to explain satellite tracking from first principles. Every numerical exercise is a constructed teaching case. It is not a navigational ice chart or a forecast of safe polar travel.
Sea-ice drift is not sea-ice concentration
Sea-ice concentration is commonly expressed as a fraction or percentage of a grid cell covered by ice. Drift describes how an identifiable ice pattern moves between two observation times. An ice field can have high concentration while moving slowly, high concentration while moving quickly, or lower concentration with a fast-moving remnant. These variables answer different questions. A chart indicating 90 percent ice cover must not be translated as 90 percent drift or a motion direction.
Thickness is yet another property. Two floes of similar surface appearance can have different thickness, age and mechanical strength. Radar or microwave observations of ice motion do not automatically determine these properties. Surface melting, freezing, deformation and ridging can change the pattern itself between satellite looks, which complicates tracking. A moving pattern can be measured while its material properties remain poorly known.
For a reader, label each map by the quantity it actually contains: displacement during a stated interval, its components, average motion if derived, valid-vector coverage and uncertainty. Avoid the vague term ‘ice activity’ when a numerical variable has a precise meaning. A translation that separates concentration, motion and thickness can immediately prevent a mistaken scientific or operational conclusion.
How paired satellite images produce an arrow
A satellite provides a spatial pattern of brightness, microwave response or texture. A drift processor compares observations separated in time and searches for a corresponding pattern in a later image. Cross-correlation is one method: the algorithm moves a window through candidate offsets and assesses how closely its appearance matches. The offset that best explains the change becomes an estimated displacement vector. This is a measurement of pattern translation, conditioned on the ability to recognize the pattern again.
A good correlation match does not guarantee that every ice floe in the window moved identically. Rotation, deformation, breaking, melting, sensor noise and changing viewing conditions can reduce matching reliability. A large window may average different movements; a small window may contain too little texture to match. Some processors use continuous maximum-cross-correlation variants and combine observations from several sensors. The chosen method and reference times belong in a responsible summary.
The resulting arrow is a displacement between two image states, not a movie of the full trajectory. Ice could have moved, curved and partly returned in the intervening time. Dividing net displacement by elapsed time gives average vector velocity, which is useful but cannot recover the path length or instantaneous acceleration. Describe what the algorithm estimated, not what one imagines the ice did between image acquisitions.
Active and passive microwave inputs
Active microwave instruments transmit radar and receive returns; passive microwave radiometers detect naturally emitted microwave energy. Both can contribute information about sea ice, but their image textures, footprints, atmospheric sensitivities and usable conditions differ. The OSI SAF near-real-time low-resolution drift product uses multiple satellite inputs and correlation methods. A merged output may combine information of different sampling geometries and timing, so it should not be mistaken for a single sharp photographic exposure.
Synthetic aperture radar can reveal relatively fine spatial features and leads, making it valuable for some tracking applications. Passive microwave instruments often offer broad coverage and recurring measurements but coarser footprints. Which product is appropriate depends on the scale of motion, region, revisit interval, sensor quality and purpose. More pixels on a displayed map do not prove that each vector has independent fine-scale information.
During melt seasons and in wet atmospheric conditions, the surface pattern can change or become harder to match. EUMETSAT’s operational sea-ice drift documentation explicitly highlights larger uncertainties and more interpolated vectors in challenging summer periods. A translation should not call these regions ‘observed with the same quality’ simply because the software exported a complete rectangular grid.
Displacement versus average velocity
Let delta x and delta y be the eastward and northward displacement components in metres, and let delta t be the elapsed time in seconds. Average velocity components are u = delta x / delta t and v = delta y / delta t. Speed is the square root of u squared plus v squared. The direction of drift normally points toward the destination, unlike meteorological wind direction, which usually names where air comes from. Keeping this convention explicit avoids an accidental 180-degree reversal.
A constructed example gives eastward displacement of 8.64 km over 48 hours. Converting units, 8,640 m divided by 172,800 seconds equals 0.050 m/s eastward. That is five centimetres per second, not five metres per second. It is also 0.18 km/h. The calculation is a time-average of the eastward component, and says nothing by itself about the northward component or total speed.
Now let the northward displacement be 6.48 km during the same interval. The northward average is 0.0375 m/s. The net displacement magnitude is the square root of 8.64² plus 6.48², which is 10.8 km; the average speed magnitude from the net vector is 10,800 / 172,800 = 0.0625 m/s. This does not mean that the ice travelled only 10.8 km along its actual, potentially curved path.
Product time stamps are not decorative
A near-real-time product may report motion estimated across a 48-hour window. Its file date can refer to a central or nominal product time rather than the beginning or end of the displacement. Some climate records use a 24-hour interval with a specified central time. Always read start, end and reference timestamps from the product definition or variables. Using the download time as the drift time changes the observation into an invented event.
Comparing a vector field valid across two days with a current forecast valid at an instant requires care. The average drift may combine wind shifts, tides and ocean forcing from several periods. A daily ice edge map may represent another observation time entirely. A sentence such as ‘ice is currently moving east at 0.05 m/s’ is stronger than the evidence if the number is a two-day historical displacement average.
When mapping multiple products, note the temporal support for each data layer. An animated display that shows a two-day vector on every hourly frame can falsely imply hourly measurements. A professional translation says ‘estimated net displacement over the preceding 48-hour pair’ unless the actual interval differs. Clear time references are essential for both field planning and scientific comparison.
Coordinate systems at the poles
Polar sea-ice products commonly use specialized map projections such as polar stereographic grids. A file can contain x and y displacements in grid coordinates that are not identical to local east and north at every point. Grid orientation, convergence of meridians and units must be examined before translating projected components into cardinal direction language. Near the pole, familiar north-up map intuitions can fail.
A trajectory shown as a straight line on a projected map may not be straight on the globe, and equally sized-looking grid cells need not represent equal distances in every coordinate system. Check the coordinate reference system, projection parameters and grid mapping attributes. If a product supplies earth-relative east/north components directly, use those rather than applying an unneeded rotation. If it supplies projected components, document any transformation into geographic bearings.
An arrow that points right on a display is not automatically eastward. The plot may be rotated, mirrored, transposed or reprojected. A simple known test vector and coordinate transformation check is often more valuable than a beautiful basemap. The purpose of translation is to carry the physical direction across displays and languages without changing it.
Cross-correlation quality and pattern persistence
A correlation coefficient or match score assesses how well a shifted observation resembles its reference window under an algorithm’s assumptions. High similarity is helpful, but a repeating texture may produce several plausible offsets. Low texture, cloud-related issues in optical products, changing radar roughness, deformation and rotation complicate feature correspondence. Product QC can reject poor matches or mark regions requiring more interpolation.
A particular strength of comparing many windows is spatial coherence: neighbouring estimates that agree can reinforce confidence. But physical deformation and shear can make real gradients. A quality controller must not erase all discontinuities merely because smooth maps are easier to read. Compare the match field, sensor coverage, deformation context and producer uncertainty estimates before deciding whether a sharp change is noise.
When an algorithm returns no valid vector, a blank cell means insufficient usable evidence or coverage, not stationary ice. A gridded product may later fill that gap from nearby information or a model. If so, distinguish ‘estimated from observation’, ‘spatial interpolation’ and ‘model-assisted gap fill’. The words matter because they disclose how strongly a position is constrained by actual satellite data.
Uncertainty may refer to a component
An ice-drift data file may provide an uncertainty as the expected standard deviation of displacement components, for example expressed in kilometres over the tracking interval. That is not necessarily the uncertainty of total speed, direction, or an entire drifting floe’s future position. Product documentation may define separate x and y errors or a common measure. Read the metadata closely and propagate it only under justified assumptions.
Suppose the eastward 48-hour displacement is 8.64 km with an illustrative one-standard-deviation component uncertainty of 1.0 km. The eastward average velocity is 0.050 m/s, with a corresponding component standard deviation of 1,000 / 172,800, approximately 0.00579 m/s. That is about 0.58 cm/s. Translating the 1.0 km displacement error directly as a 1.0 m/s velocity error would be off by orders of magnitude.
If independent displacement-component errors are supplied, speed uncertainty depends on both components and their covariance. Angle uncertainty can grow sharply when displacement magnitude is small. If errors are correlated, a root-sum-square assumption may be inappropriate. Present the component uncertainties and their reference period rather than inventing a neat single confidence radius. The file’s uncertainty definition should travel with every derived metric.
Observed, interpolated and model-assisted fields
Gridded sea-ice motion products strive for spatial coverage, but full-looking fields can incorporate interpolation. The OSI SAF climate data record explicitly describes model-assisted gap filling in some summer periods. A vector in a model-filled region is a useful estimate, but it is not independent satellite tracking of a recognized pattern at that exact point. Colour scales that ignore the provenance field create a false sense of uniformly observed evidence.
For studies of a storm or deformation event, a few directly observed local vectors may be more informative than a smooth but heavily interpolated field. For large-scale climate statistics, a carefully documented homogeneous, gap-filled record may be preferable. The proper choice depends on the scientific question. One product’s strength does not make another product defective; they are designed for different inference jobs.
Report the fraction of observation-supported, interpolated and missing cells where the product permits it. If an analysis masks to only directly observed data, explain the resulting coverage bias. If it uses fills, explain what independent information is no longer available. A useful translation helps the reader understand the trade-off rather than hiding it behind a polished map.
Drift, deformation and divergence
A field of ice-motion vectors can be used to examine deformation. If neighbouring parcels move apart, the field has divergence; if they move together, it has convergence; differential motion may indicate shear. Yet finite differences of noisy or coarse vectors can amplify uncertainty. An apparently large deformation at a missing-data boundary or an interpolation seam should be checked before being interpreted as opening a lead.
Consider two hypothetical floes 10 km apart along a line. If their relative separation increases by 1 km during 48 hours, the fractional separation change is 10 percent over that interval under the simplified geometry. Dividing by time gives an approximate average normal strain rate of 0.1 / 172,800 = 5.79 × 10^-7 per second. This toy calculation assumes identifiable material parcels and a constant reference separation; a gridded data product may not satisfy those assumptions exactly.
Deformation is not the same as drift speed. Two adjacent floes can move quickly together with little local deformation, while slow but opposing motions can create high local strain. Translating an image of a fractured ice pack requires examining gradients, time interval and scale, not just the colour of arrows. A detailed local ice safety assessment requires more than a coarse global drift field.
Wind forcing, ocean currents and the Coriolis effect
Sea-ice drift responds to atmospheric stress, ocean currents, internal ice forces, boundaries and Earth’s rotation. These influences vary with ice concentration, thickness, compactness, coastline geometry and weather. Wind direction often correlates with ice motion, but the two are not identical. Ice can keep moving after winds change, rotate under internal stress or be constrained by surrounding packed ice.
A satellite scatterometer wind vector normally describes air motion at the sea surface under a wind convention. An ice-drift vector describes the estimated movement of an ice pattern toward its new location. Comparing them without correcting the from/to convention can make them appear opposite. Even after convention correction, an angular offset may reflect real forces rather than a measurement error.
Do not infer that a drift speed of 0.05 m/s proves a particular 10-m wind speed. That would require a dynamical model and local conditions. Similarly, wind forecasts alone are not validated ice trajectories. A responsible explanation describes the observed or retrieved quantities separately and interprets their relationship as a physical hypothesis, with appropriate model and observational evidence.
Seasonal limitations and long-term records
Melting, wet snow and variable surface properties can change the microwave signature of ice, making correlation-based tracking more difficult. OSI SAF specifically warns about larger uncertainty and more interpolation in difficult seasonal periods. A time series that includes summer should therefore examine quality metrics and gap filling before describing a seasonal shift in drift speed. A processing change can also alter a trend if it is not accounted for.
Near-real-time and climate data record products serve different jobs. An operational stream emphasizes timely estimates for models and monitoring, while a climate record prioritizes consistent historical processing and documented changes. Combining them in a single trend plot requires a careful compatibility audit. A product issued recently is not necessarily the right one for multi-decade climate inference.
The word ‘trend’ demands a statistical time span, uncertainty analysis, and knowledge of coverage changes. One winter’s fast drift is not proof of a climate trend. Conversely, a long record with seasonal model fill must disclose how its values were generated. Good translation protects the link between observed motion, dataset construction and broader conclusions about polar change.
Worked example: net vector versus travelled path
A constructed floe position moves 6 km east and 8 km north over two days. Its net displacement magnitude is 10 km. The corresponding average velocity magnitude is 10,000 / 172,800 = about 0.0579 m/s. The displacement direction is northeast, specifically arctan(6/8) about 36.9 degrees east of north. This is a bearing toward the final position, not a meteorological source-direction label.
Imagine instead the ice first moves 8 km east, then 6 km north, while total elapsed time remains 48 hours. The actual path length is 14 km, but the net displacement is still 10 km. Two image endpoints alone cannot distinguish these trajectories. If a report says ‘the floe travelled exactly 10 km’ it silently substitutes endpoint separation for path length. Repair it with ‘net displacement of 10 km between tracked positions’.
Now assume the observed endpoint has a horizontal displacement uncertainty. That uncertainty describes confidence about the estimated match, not the unobserved route. Even a perfectly known endpoint does not recover the full path. The difference between net movement and route travelled is central when interpreting every long-baseline tracking product.
Worked example: uncertainty and dimensional checks
Suppose a hypothetical dataset gives x displacement 12.0 km, y displacement 0.0 km, and a 48-hour interval. The average x velocity is 12,000 / 172,800 = 0.0694 m/s; y velocity is zero. If the x-displacement uncertainty is 2.0 km as one standard deviation, its velocity-component uncertainty is 2,000 / 172,800 = 0.0116 m/s. The numerical ratio of signal to component uncertainty is six, under that simplified characterization.
Writing the uncertainty as 2.0 km/h would be dimensionally wrong. A distance uncertainty over an interval must be divided by elapsed time to become a velocity-component uncertainty. If the product defines uncertainty in pixels or projected grid units, first convert using the grid geometry and projection. If instead it already supplies m/s, dividing again would be another mistake.
An uncertainty interval should not be presented as a guaranteed bound on every floe. One standard deviation is a statistical descriptor under a model, not a hard safety margin. Navigational operations may need conservative, locally verified ice information beyond a broad drift product. Scientific precision is not the same as an operational safety guarantee.
A field-reading workflow for researchers and learners
Begin with product identity and provenance: provider, collection, release, sensor inputs, grid mapping, spatial resolution, tracking window, valid time and licence. Examine units of x and y displacement or velocity, missing-value codes, flags, expected-accuracy variables and any fields describing interpolation. Do not guess whether a vector is earth-relative or projection-relative. If the name is ambiguous, consult the producer’s user guide.
Next test a simple numerical sample using an independent calculation. Convert displacement to average speed only when elapsed time is known; verify components and bearings. Plot vector length at a scale that does not obscure uncertainty. Keep direct observation and filled data visually distinguishable. Where comparisons are useful, collocate with buoys, ice drifters or alternate products using appropriate time and position tolerance.
Finally translate the result with the correct epistemic status: ‘satellite pattern matching suggests a net eastward displacement of about 8.6 km over 48 hours’ is precise about both method and support. If a product uses gap filling, label that explicitly. A clear result should tell the reader what moved, over what period, by what evidence, and what prevents an exact local interpretation.
Communicating ice information responsibly
Sea-ice motion matters to climate science, ecosystems, marine operations and community safety. Yet a low-resolution drift map does not identify every lead, ridge, crack or rapidly evolving hazard. The tracked pattern may represent a broad area rather than a particular named floe. Do not turn a historical average arrow into a precise travel instruction or promise that ice will remain stable near a route.
Reports should separate observed evidence from decisions. An ice analyst can report the movement estimate and its quality; an operational authority can issue location-specific advisories based on multiple information sources. A public explanation should never translate a past displacement directly into ‘safe passage’ or ‘unsafe passage’ without the appropriate current expertise and observations.
Across languages, preserve verbs that distinguish ‘drifted’, ‘is drifting’, ‘expected to drift’ and ‘may drift’. These tenses imply different time references and evidence. Keep units, central time, reference interval, projection and uncertainty together. When the words are careful, an abstract vector plot becomes a useful explanation without exceeding what the satellite and processing chain can show.
Diagnostic cases: ice-drift evidence and common repair decisions
Case 1: The arrows show motion, not ice amount
A high-concentration region has short drift arrows. It is incorrect to read the short arrows as thin or sparse ice. Motion and coverage are different variables. State the net displacement over its reference interval, then cite concentration data separately if needed. A single graphic may combine both layers, but the underlying measurements must remain distinct.
Case 2: Two-day motion is called instantaneous
A vector encodes displacement from Monday morning to Wednesday morning. Dividing by time gives average velocity, not an instantaneous Wednesday wind or ice speed. If motion reversed overnight, net displacement may be small despite substantial travel. Translate ‘net drift during 48 hours’ rather than ‘ice is now travelling at’ without current evidence.
Case 3: The map arrows point right
A polar projection rotates geographical directions across the scene. Rightward on the screen may not mean eastward in geographic coordinates. Check projection and vector rotation before choosing cardinal words. A strong visual impression is no substitute for coordinate metadata. The error becomes especially serious when maps from different pole-centred projections are compared.
Case 4: A vector is blank near the ice edge
Pattern tracking may fail where ice breaks up, melts or has poor image texture. The missing cell does not prove zero drift. Describe data unavailability and examine a separate product if useful. Never fill blanks with stillness in a scientific chart or public communication. Preserve the missing-value flag rather than converting it to a numeric zero.
Case 5: Summer has many smooth filled arrows
A merged field remains complete through a difficult melt period. Inspect how many vectors came from tracking and how many from interpolation or model-assisted fill. A smooth-looking map can be useful for large-scale analysis yet contain fewer independent observations. Report the provenance and avoid claiming every arrow was measured by satellite image correlation.
Case 6: Pattern matching has two peaks
A repeated ice texture gives two possible correlation maxima. The algorithm may select one by neighbourhood coherence or prediction, but ambiguity remains. Examine match quality and independent nearby observations. Do not describe the chosen offset as uniquely observed if the raw image evidence supports alternatives. The same logic applies to other remote-sensing inversions.
Case 7: Large floes rotate while drifting
A tracking window contains an ice floe that both translates and rotates. Pure translational correlation may fit poorly or yield an approximate centre-of-pattern motion. Rotation can also change apparent texture. The retrieved vector should not be interpreted as the exact movement of every corner of the floe. Distinguish rigid displacement from deformation and angular motion.
Case 8: Near-zero displacement has wild bearings
A cell reports a displacement magnitude close to its uncertainty. Small changes in x and y then create large changes in bearing. The angle should be treated as weakly constrained rather than numerically precise to one decimal degree. The estimated small net motion may still be useful. State the signal-to-uncertainty context before interpreting direction.
Case 9: Wind points against the drift arrows
Meteorological wind direction often means ‘from’, while drift arrows mean ‘toward’. Correct this convention first. If the apparent disagreement remains after conversion, investigate ocean currents, ice stress and temporal mismatch. A difference is not automatically a sensor error. Translation should preserve vector semantics and physical complexity together.
Case 10: A drift product is used as a ship route
A broad global vector map is not a current, local sea-ice navigation assessment. It may miss leads, ridges and sharp boundary changes. Warn that its resolution and time interval limit operational use and direct safety decisions to authoritative ice services. A scientific estimate can be valuable without being a safe passage instruction.
Case 11: A monthly average hides storms
A monthly mean is calculated from daily or two-day motion fields. Opposing displacements can cancel, producing a small mean vector even when individual events were fast. Distinguish mean vector magnitude from mean of individual speeds. The former can be smaller. Report the statistic and denominator used rather than simply calling both ‘average ice speed’.
Case 12: Uncertainty is listed in kilometres
A file gives one-kilometre standard deviation for a displacement component over 48 hours. Converting it to a component velocity standard deviation requires 1,000 / 172,800, or 0.00579 m/s. Writing ±1 m/s is dimensionally incorrect. A component error also does not automatically represent radial position confidence or future track uncertainty.
Case 13: The drifting path curves
The floe moves 8 km east then 6 km north. Path length is 14 km, but endpoint separation is 10 km. A two-image displacement retrieval sees only the latter under ideal conditions. This distinction matters for travel and material transport. Translate ‘net displacement’ rather than ‘total path length’ whenever intermediate positions are unknown.
Case 14: Image times cross midnight
A tracking window begins at 18:00 UTC and ends after midnight two days later. A local-time caption using only calendar dates can distort the interval. Preserve full timestamps, time zone and elapsed seconds. If an endpoint is converted into local time, convert both endpoints consistently. A correct distance with a wrong interval yields a wrong average velocity.
Case 15: Products use different grid spacing
One sea-ice motion product posts vectors on a coarse global grid while another uses finer regional imagery. Do not assume the finer grid independently resolves every physical process. Compare effective footprint, sensor input, match window and processing. A difference in reported boundary position may reflect representation and filtering rather than true contradictory ice motion.
Case 16: A climate record changes method
A product history switches sensors or gap-filling method. An apparent jump in motion statistics may be methodological. Consult release notes and assess continuity or reprocessing before claiming abrupt climate change. Near-real-time and climate data records are not always exchangeable. A transparent report identifies dataset version and treats possible inhomogeneity as a limitation.
Case 17: A convergence map highlights a seam
A large negative divergence occurs at the boundary between observed and interpolated vectors. Finite differences amplify abrupt changes, including artificial ones. Check flags and smoothing sensitivity before interpreting the seam as a new lead or compressive ridge. A structural gradient in a plotted field is not necessarily a physical deformation rate.
Case 18: A floe is near grounded ice
Landfast ice and mobile pack ice have distinct motion constraints. A coarse tracking cell near the boundary may mix stationary and moving patterns, affecting correlation. Do not attribute its vector to a specific floe without local evidence. State the grid-based pattern estimate and the physical mixture. A static coastline is not automatically a reliable reference for every mixed pixel.
Case 19: A time series uses different time bases
A 24-hour climate record and a 48-hour near-real-time product are compared as raw kilometres per interval. Their displacement magnitudes cannot be compared without normalizing elapsed time, and even velocity comparisons need method and coverage checks. Convert dimensions carefully and disclose product differences. Similar units are necessary but not sufficient for scientific comparability.
Case 20: A headline calls ice faster than wind
An author compares a 12 km drift over 48 hours to a 10 m/s wind and decides ice is faster because 12 is larger than 10. These numbers have different units and dimensions. Ice’s 12,000 / 172,800 is about 0.069 m/s. The units repair the comparison immediately, though different physical quantities still need suitable interpretation.
Case 21: A missing interval is invented
A file reports drift components but no documented start or end time. Assuming a 24-hour period solely because the filename includes a date can halve or double the inferred speed. Consult metadata, product documentation or the record’s time bounds. If the interval remains unavailable, report displacement and refrain from converting to velocity.
Independent transfer problems with checked answers
Problem 1: Convert 8.64 km eastward in 48 h to average velocity. Answer: 8,640 / 172,800 = 0.0500 m/s eastward, or 5 cm/s.
Problem 2: Convert 6.48 km northward in 48 h. Answer: 6,480 / 172,800 = 0.0375 m/s northward.
Problem 3: Combine the preceding perpendicular displacements. Answer: net displacement is sqrt(8.64² + 6.48²) = 10.8 km and average net speed is 0.0625 m/s.
Problem 4: A field reports 12 km motion over 48 h. Answer: average net speed magnitude is about 0.0694 m/s, not 12 m/s.
Problem 5: A component displacement uncertainty is 1 km over 48 h. Answer: component velocity standard deviation is about 0.00579 m/s, assuming that is how the product defines the uncertainty.
Problem 6: A floe moves 8 km east then 6 km north. Answer: path length is 14 km; net displacement is 10 km. The endpoints do not reveal the intermediate route.
Problem 7: A vector is 6 km east and 8 km north. Answer: the net direction is about 36.9° east of north, pointing toward the destination, not from it.
Problem 8: A chart says ice concentration 85%. Answer: it means fraction of area covered by ice under the product’s convention, not drift speed or thickness.
Problem 9: Why must a two-day drift vector not be described as current instantaneous speed? Answer: it is derived from net motion between observations over a long window.
Problem 10: A missing vector occurs in melt season. Answer: report missing tracking support or other QC reasons; absence of an arrow is not proof of stationary ice.
Problem 11: A product fills summer gaps with model output. Answer: label those cells as filled or model-assisted, not as independently tracked by satellite.
Problem 12: A projected x vector points right on the screen. Answer: confirm the grid mapping before converting to earth-relative eastward motion.
Problem 13: Two ice products report 12 km/48 h and 8 km/24 h. Answer: average net speeds are roughly 0.0694 m/s and 0.0926 m/s respectively, before method and coverage comparisons.
Problem 14: A 48-hour drift of 0 km has a winding path. Answer: net displacement can be zero despite nonzero travelled distance; do not infer no motion without intermediate observations.
Problem 15: A standard deviation is quoted for each displacement component. Answer: do not relabel it a guaranteed radial error or a future-track confidence circle without a propagation model.
Problem 16: Two neighbouring displacement vectors differ sharply near a filled-data boundary. Answer: inspect provenance before interpreting an extreme strain rate.
Problem 17: An ice-motion arrow points northeast but wind direction says southwest. Answer: these may agree if the wind direction means air arriving from southwest and moving northeast.
Problem 18: A displacement of 3 km occurs in 24 h. Answer: average speed magnitude is 3,000 / 86,400 ≈ 0.0347 m/s.
Problem 19: A monthly average vector is near zero. Answer: opposing daily vectors may cancel; mean vector magnitude is not identical to average daily speed.
Problem 20: A file gives a tracking displacement but lacks the elapsed interval. Answer: do not invent the interval. Preserve the displacement and flag the missing time basis.
Frequently asked questions
Does a drift arrow show how much ice there is?
No. Drift represents pattern displacement over a time interval; concentration describes fractional ice coverage.
Is drift direction the direction ice moves toward?
Usually yes, though the product’s coordinate convention must be verified. Meteorological wind direction is often expressed oppositely.
What does a 48-hour motion vector mean?
It indicates estimated net displacement between two observation states, and only an interval-mean velocity when divided by elapsed time.
Are all gridded drift vectors directly tracked?
Not necessarily. Some products interpolate or use model-assisted gap filling, which should be reported separately.
Why is summer tracking difficult?
Melting and changes to surface signatures make matching harder, often raising uncertainty and increasing interpolated coverage.
Can this map guide a vessel through ice?
A low-resolution historical motion product cannot substitute for current operational ice charts and authoritative safety guidance.
What should a translated caption include?
Product release, sensor basis, start and end times, displacement or velocity units, direction convention, validity flags and uncertainty definition.
Method and source documentation
OSI SAF global near-real-time low-resolution sea-ice drift OSI-405-d: https://osi-saf.eumetsat.int/products/osi-405-d
EUMETSAT user portal multimission sea-ice drift: https://user.eumetsat.int/catalogue/EO%3AEUM%3ADAT%3AMULT%3AGBL-LR-SIDR
OSI SAF global sea-ice drift climate data record OSI-455: https://osi-saf.eumetsat.int/products/osi-455
Google Search Central people-first content guidance: https://developers.google.com/search/docs/fundamentals/creating-helpful-content
Related eduKate explanations
How to Translate Easily series index: https://edukatesg.com/how-to-translate-easily-series-index/
Passive microwave sea-ice concentration and contamination: https://edukatesg.com/2026/10/30/how-to-translate-easily-passive-microwave-sea-ice-concentration-weather-land-spillover/
ASCAT scatterometer ocean wind vectors and flags: https://edukatesg.com/2026/10/30/how-to-translate-easily-ascat-scatterometer-ocean-winds-backscatter-ambiguity-rain/
How to read ocean drifter tracks and drogue status: https://edukatesg.com/2026/10/28/how-to-translate-easily-ocean-drifter-tracks-drogue-windage-current-velocity/