Sea-ice extent and concentration mapping
Passive microwave radiometry has tracked sea-ice extent and fractional concentration daily since 1979, but its 25 km resolution floor hides the leads and polynyas that matter most. Sentinel-1 SAR closes that gap at the cost of coverage and latency.
Sensors
- AMSR2 (JAXA GCOM-W1): Passive microwave radiometer operating at 6.9–89 GHz. Delivers daily global sea-ice concentration at 25 km grid cells (6.25 km experimental products at 89 GHz). Primary workhorse for extent and concentration time-series since 2012.
- SSMIS (DMSP F-17/F-18): Special Sensor Microwave Imager/Sounder. 19–91 GHz passive microwave; 25 km standard products. Provides continuity with the SSM/I record back to 1987, enabling the longest consistent concentration archive in operational use.
- Sentinel-1 A/B SAR (ESA): C-band synthetic aperture radar at 5.405 GHz. Extra-Wide Swath mode delivers 400 km swath at 40 m resolution; IW mode reaches 5–20 m. Resolves individual leads, ridges and small polynyas invisible to passive microwave. Polar revisit roughly 1–3 days depending on latitude and operational tasking.
- MODIS (Terra/Aqua): Visible and thermal infrared at 250 m–1 km. Useful for ice-edge mapping and surface temperature in cloud-free conditions. Cloud cover, which is pervasive in polar regions, frequently renders MODIS unusable for days at a time.
What brightness temperature actually measures
Sea ice and open water emit microwave radiation at very different efficiencies. First-year ice has an emissivity near 0.95 at 19 GHz; open water sits around 0.45. That contrast, expressed as brightness temperature, is what passive microwave radiometers record. The difference is large enough to be detectable through cloud and polar darkness, which is why these sensors became the backbone of the sea-ice climate record.
Concentration, rather than a simple ice-or-water binary, is derived from the brightness temperature of a mixed pixel. A 25 km grid cell containing 60% ice and 40% open water produces a brightness temperature somewhere between the two end-members. The algorithms that invert that signal into a concentration fraction are the subject of decades of refinement and, still, genuine scientific dispute.
NASA Team and Bootstrap: what each algorithm assumes
The NASA Team algorithm, developed in the 1980s and still widely used on SSMIS and AMSR2 data, uses polarisation ratios and spectral gradient ratios at 19 and 37 GHz. It performs well in winter but is known to underestimate concentration during summer melt, when liquid water on the ice surface depresses emissivity and mimics open water. Published comparisons put the underestimate at several percentage points in melt-pond-rich conditions.
The Bootstrap algorithm takes a different approach: it fits observed brightness temperatures to a tie-point space defined empirically from known ice and water signatures, then interpolates. It tends to produce higher concentration estimates than NASA Team during summer and is generally considered more stable across seasons, though it carries its own sensitivity to tie-point drift as sensor characteristics age. Neither algorithm is unambiguously correct. Operational centres including NSIDC and EUMETSAT publish both, and analysts comparing interannual trends should use a single consistent algorithm throughout.
AMSR2's 89 GHz channel allows experimental 6.25 km products that resolve finer structure, but atmospheric water vapour absorbs strongly at that frequency, introducing noise that requires correction. The 25 km standard product remains the reference for climate-quality records.
The 25 km problem: leads, polynyas and what gets missed
A lead, the narrow fracture of open water between ice floes, can be a few metres to a few kilometres wide. A polynya, a persistent open-water area maintained by winds or upwelling, may span tens of kilometres. Both are oceanographically significant: they are the primary sites of ocean-atmosphere heat exchange in winter, and they control the formation of dense bottom water that drives global thermohaline circulation. At 25 km resolution, a lead narrower than roughly 5–10 km is simply absorbed into the surrounding pixel and reported as high-concentration ice.
Sentinel-1 SAR resolves this. At 40 m in Extra-Wide Swath mode, a 500 m lead is visible as a distinct dark return against the brighter backscatter of surrounding ice. The physics is different from passive microwave: SAR measures surface roughness and dielectric properties rather than thermal emission, so it discriminates ice types, detects new ice formation and maps deformation features. The cost is coverage. A single Sentinel-1 pass covers a 400 km swath; achieving daily coverage of the full Arctic requires combining multiple passes and accepting gaps, particularly at lower polar latitudes.
MODIS adds a third layer when skies cooperate. At 250 m in visible bands, it can map ice-edge position and surface albedo with a clarity that neither microwave sensor can match. In practice, cloud cover over the Arctic and Antarctic exceeds 80% on a climatological basis, making MODIS a useful supplement rather than a reliable operational source.
The archive: why 1979 matters and where the seams are
The passive microwave record begins with the SMMR instrument on Nimbus-7 in 1978, transitions to SSM/I on DMSP in 1987, and continues through SSMIS and AMSR2 today. NSIDC's Sea Ice Index, which is the most widely cited extent product, is built on this chain. The record is long enough to detect multi-decadal trends in Arctic minimum extent, which has declined at roughly 13% per decade since 1979 based on published NSIDC analyses.
The seams between sensors matter. Each instrument has slightly different channel frequencies, calibration characteristics and orbital geometries. Intercalibration offsets between SMMR, SSM/I and SSMIS are documented but not trivial. Any analysis claiming to detect trends at the few-percent level needs to account for these discontinuities explicitly. The EUMETSAT Ocean and Sea Ice Satellite Application Facility (OSI-SAF) publishes reprocessed, intercalibrated products specifically to address this problem.
Combining sensors without overcomplicating the product
The practical workflow for most operational users is a hierarchy. Passive microwave provides the daily, all-weather, all-darkness baseline concentration map at 25 km. Sentinel-1 SAR is tasked over areas of interest, such as shipping corridors, research sites or anomalous concentration signals, to provide fine-scale context. MODIS or VIIRS imagery is ingested opportunistically on clear days to validate ice-edge position.
Fusion is not straightforward. The sensors measure different physical quantities, and a pixel classified as 70% concentration by AMSR2 does not map cleanly onto SAR backscatter classes. The honest approach is to treat each layer as answering a different question rather than producing a single merged concentration number. Satellize applies this multi-sensor hierarchy in its analytics workflows, running open constellation data and adding commercial SAR tasking where a client's area of interest demands finer resolution.
Latency varies significantly by sensor. AMSR2 near-real-time products are typically available within a few hours of acquisition. Sentinel-1 standard products appear on the Copernicus Data Space within 1–3 hours of downlink. MODIS daily composites are available same-day through NASA Earthdata. For operational decisions, that is fast enough. For climate research, the preference is for reprocessed products that may lag by weeks or months but carry better calibration.
What the numbers cannot tell you
Concentration is an area fraction. It says nothing about ice thickness, age, strength or navigability. A pixel reporting 90% concentration could be a metre of solid multi-year ice or a thin skim of new nilas that a vessel could push through. That distinction requires altimetry or SAR texture analysis, which are covered on the sibling pages for sea-ice thickness and polar shipping-route ice condition monitoring.
Melt ponds are a persistent source of error in summer concentration products. When the ice surface is covered with liquid water, passive microwave algorithms read lower emissivity and report lower concentration than is physically present. Published studies have documented underestimates of 10–20 percentage points in heavily ponded Arctic conditions during July and August. This is not a solved problem. Users interpreting summer minimum extent figures should treat them as lower bounds rather than precise measurements.
Typical figures
| Standard spatial resolution (passive microwave) | 25 km grid (AMSR2, SSMIS); 6.25 km experimental at 89 GHz (AMSR2) |
| Spatial resolution (Sentinel-1 SAR) | 40 m (Extra-Wide Swath); 5–20 m (Interferometric Wide Swath) |
| Revisit (passive microwave) | Daily global coverage; AMSR2 and SSMIS provide two passes per day at most latitudes |
| Revisit (Sentinel-1 at polar latitudes) | 1–3 days depending on latitude and operational tasking plan |
| Frequency / spectral bands | AMSR2: 6.9–89 GHz (multi-channel); SSMIS: 19–91 GHz; Sentinel-1: C-band 5.405 GHz |
| Concentration accuracy (winter) | Approximately ±5% for NASA Team and Bootstrap algorithms under cold, dry conditions |
| Concentration accuracy (summer, melt ponds) | Underestimates of 10–20 percentage points documented in heavily ponded conditions |
| Archive depth (passive microwave) | SMMR from 1978; SSM/I from 1987; continuous intercalibrated record via NSIDC and OSI-SAF |
| Near-real-time latency | AMSR2 NRT: within ~3 hours; Sentinel-1: 1–3 hours post-downlink on Copernicus Data Space |
| Delivery formats | NetCDF (NSIDC, OSI-SAF), GeoTIFF, HDF5, shapefiles for ice-edge vectors |
Analytics Satellize can run
| Daily sea-ice extent and area time-series | NASA Team or Bootstrap algorithm applied to AMSR2/SSMIS brightness temperatures; 15% concentration threshold for extent definition following NSIDC convention | Automated daily GIS layer and CSV time-series with anomaly flagging against 1981–2010 climatological baseline |
| Fractional concentration grid | Bootstrap algorithm on AMSR2 dual-polarisation channels at 18.7 and 36.5 GHz | NetCDF or GeoTIFF concentration map, updated daily, covering user-defined Arctic or Antarctic sub-region |
| Ice-edge position vector | Threshold contour extraction at 15% concentration from passive microwave product, validated against MODIS visible imagery on clear days | Daily shapefile or GeoJSON ice-edge polyline with confidence flag indicating cloud contamination of optical validation |
| Lead and polynya detection | Sentinel-1 SAR backscatter segmentation; low-backscatter regions within high-concentration passive microwave pixels classified as open-water features | Polygon layer of detected leads and polynyas with area estimates, delivered within 4 hours of SAR acquisition |
| Interannual trend and anomaly report | Linear regression and anomaly standardisation against the intercalibrated OSI-SAF or NSIDC Sea Ice Index passive microwave archive | Quarterly PDF report with time-series plots, ranked anomaly tables and algorithm-consistency notes |
| Multi-sensor ice classification (type discrimination) | SAR texture and backscatter intensity combined with passive microwave spectral gradient ratio to separate first-year, multi-year and new ice classes | Weekly classified raster over a defined area of interest, with class legend and known confusion matrix from published validation studies |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.