Under-ice and marginal-ice-zone phytoplankton bloom detection
Phytoplankton blooms initiate beneath sea ice weeks before open water appears, yet ocean-colour satellites cannot see through ice. Detecting them demands a fusion of marginal-ice-zone chlorophyll retrievals, light-transmission modelling constrained by melt-pond and snow data, and validation from biogeochemical Argo floats.
Sensors
- PACE OCI (NASA, launched 2024): Hyperspectral ocean-colour imager covering 340–890 nm at roughly 5 nm spectral resolution and 1 km spatial resolution, with daily global coverage. Designed to resolve phytoplankton community composition, not just total chlorophyll, giving unprecedented discrimination of bloom taxa in ice-free marginal-zone waters.
- MODIS-Aqua: Provides chlorophyll-a retrievals at 1 km resolution with daily polar overpass frequency. The 20-plus-year archive (2002–present) is the primary source for inter-annual bloom phenology comparisons in the Beaufort and Chukchi seas, though cloud and ice contamination can reduce usable observations to fewer than 30 clear days per summer in some sectors.
- AMSR2 on GCOM-W1 (JAXA): Passive microwave radiometer providing daily sea-ice concentration at 3.125–25 km resolution (depending on frequency channel), melt-onset timing, and melt-pond fraction proxies derived from brightness-temperature ratios. Snow depth on sea ice can be estimated from the 18.7/36.5 GHz ratio, though accuracy degrades during melt when liquid water contaminates the signal.
- BGC-Argo float network: Autonomous profiling floats equipped with chlorophyll fluorometers, PAR sensors, and nitrate optodes. Floats operating under ice in the Arctic record subsurface chlorophyll profiles at depth intervals of roughly 1–5 m, providing the only direct sub-ice biological validation. Coverage is sparse: fewer than 50 BGC-Argo floats were operating in Arctic waters as of the mid-2020s.
- MODIS Terra / VIIRS (Suomi-NPP, NOAA-20): VIIRS provides 750 m ocean-colour bands and 375 m imagery for ice-edge delineation, supplementing MODIS-Aqua with a second daily overpass and a longer-term continuity path. Useful for mapping the rapidly shifting ice-edge position that defines the productive marginal zone.
Why blooms start before the ice retreats
The conventional picture of Arctic primary productivity begins with ice melt exposing open water to sunlight. That picture is incomplete. Field campaigns in the Beaufort and Chukchi seas, including the 2011 ICESCAPE expedition, documented massive phytoplankton accumulations at depth beneath sea ice that was still largely intact. The mechanism is melt-pond optics: first-year ice with surface melt ponds transmits photosynthetically active radiation (PAR) at rates an order of magnitude higher than snow-covered ice, sometimes exceeding 10 W m⁻² at the ice base. That is enough to support net phytoplankton growth weeks ahead of the open-water season.
The ecological and biogeochemical consequences are significant. Under-ice blooms consume nutrients before the surface mixed layer is exposed, potentially altering the magnitude and timing of the better-documented open-water bloom that follows. For fisheries managers and carbon-flux modellers, missing the under-ice signal means systematically underestimating annual net primary production in the Arctic Ocean.
What satellites can and cannot see
Ocean-colour radiometry works by measuring the fraction of sunlight backscattered upward through the water column. Sea ice blocks that signal entirely. No amount of spectral sophistication in PACE OCI or MODIS-Aqua recovers chlorophyll from beneath a continuous ice cover. The observable domain is therefore the marginal ice zone (MIZ): the band of fractional ice cover, typically defined as 15–80 % concentration, where open-water pixels exist within or adjacent to the ice pack.
Within the MIZ, retrievals are complicated. Ice-contaminated pixels produce anomalously high reflectance that standard chlorophyll algorithms misinterpret. Adjacency effects from nearby bright ice surfaces scatter light into ocean pixels. Cloud cover at high latitudes is persistent; in some Arctic sectors fewer than 20 usable cloud-free MODIS observations occur per month in summer. PACE's hyperspectral capability improves atmospheric correction and pixel-level ice flagging, but it does not eliminate these problems. Honest MIZ chlorophyll retrievals carry uncertainties that can exceed 50 % in heavily mixed pixels close to the ice edge.
The light-transmission model: inferring what sensors cannot directly measure
Estimating under-ice chlorophyll requires a modelling step. The approach published in the literature couples a sea-ice optical model with a water-column photosynthesis model. AMSR2 passive microwave data constrain ice concentration and melt-onset timing. Melt-pond fraction, which controls ice transmittance, can be estimated from brightness-temperature polarisation ratios at 18.7 and 36.5 GHz, or from high-resolution optical imagery when cloud permits. Snow depth modifies transmittance further: even a few centimetres of snow can reduce PAR transmission by 90 %.
Given estimated PAR at the ice base, a nutrient-limited growth model propagates chlorophyll accumulation through the water column. The output is a spatially gridded, daily estimate of under-ice bloom probability and approximate biomass. The uncertainty is large, perhaps a factor of two to three in absolute chlorophyll concentration, and the model is sensitive to snow-depth inputs that passive microwave retrieves poorly during active melt. BGC-Argo floats are the primary means of constraining and validating these model outputs, but their spatial coverage remains thin.
Reading the BGC-Argo record honestly
BGC-Argo floats are the closest thing to ground truth available for sub-ice biology. A float drifting under the ice records chlorophyll fluorescence, PAR, and sometimes nitrate at each profile, typically every five to ten days. The fluorescence signal itself carries a known bias: non-photochemical quenching suppresses fluorescence near the surface in high-light conditions, leading to underestimates of near-surface chlorophyll. Correction algorithms exist but add uncertainty.
The network is growing but remains sparse relative to the Arctic basin. Floats can become trapped under ice for months, surfacing only when leads allow. Positional uncertainty during ice-covered periods accumulates. Despite these caveats, the float record has confirmed under-ice bloom events in the Beaufort Sea and demonstrated that satellite-derived melt-pond fractions are a credible predictor of bloom initiation timing, lending physical credibility to the light-transmission modelling approach.
Practical products and their honest limits
A working detection system produces three layered outputs. First, a daily MIZ chlorophyll map derived from PACE OCI or MODIS-Aqua, flagged for ice contamination and cloud, with per-pixel uncertainty estimates. Second, a modelled under-ice PAR and bloom-probability field, gridded at the resolution of the AMSR2 ice product (roughly 12.5 km for the relevant channels), updated daily as melt-pond fraction evolves. Third, a phenology summary: the estimated date of bloom initiation, peak biomass, and senescence for each grid cell, compared against the prior-year and climatological baselines from the MODIS archive.
The spatial resolution of the under-ice product is limited by the passive microwave input, not by the ocean-colour sensor. Twelve kilometres is coarse relative to the heterogeneity of melt-pond distribution, which varies at scales of tens to hundreds of metres. Users requiring finer spatial detail need to accept that the under-ice layer is a modelled inference, not a direct observation. Satellize integrates these data streams for clients who need the Arctic biological cycle characterised as a continuous seasonal narrative rather than a set of disconnected sensor outputs. The Tonga crop-estimation programme uses a comparable fusion logic, combining model priors with sparse in-situ validation, which is a pattern that transfers directly to this domain.
For policy and research clients, the most defensible deliverable is a probabilistic bloom-onset calendar, expressed as a distribution of likely initiation dates with confidence intervals, rather than a single deterministic date. That framing is more honest and, in practice, more useful for planning field campaigns or calibrating ecosystem models.
Archive depth and what it enables
MODIS-Aqua has operated since 2002, providing more than two decades of MIZ chlorophyll observations. That archive is long enough to detect trends in bloom phenology: studies have documented earlier bloom initiation in parts of the Chukchi Sea correlating with earlier melt onset. AMSR2 data extend back to 2012; its predecessor AMSR-E (2002–2011) provides a longer passive microwave record with some intercalibration uncertainty between instruments.
PACE OCI, launched in February 2024, adds hyperspectral capability that no prior ocean-colour mission provided. Its ability to resolve phytoplankton functional types from spectral absorption features means that future analyses will distinguish diatom-dominated blooms from flagellate assemblages, information that matters for carbon-export efficiency and food-web structure. That capability is new, and the community is still developing validated retrieval algorithms for polar conditions. Treat PACE-derived community composition in the MIZ as a research product for now, not an operational one.
Typical figures
| MIZ chlorophyll spatial resolution | 1 km (MODIS-Aqua, PACE OCI); 750 m (VIIRS) |
| Under-ice PAR / bloom-probability grid resolution | ~12.5 km, constrained by AMSR2 passive microwave input |
| Revisit (ocean-colour sensors) | Daily at polar latitudes; usable cloud-free observations often fewer than 20 per month per sector |
| Passive microwave revisit (AMSR2) | Daily global coverage; twice-daily at high latitudes |
| PACE OCI spectral range | 340–890 nm at ~5 nm resolution; polarimetric HARP2 and SPEXone instruments co-manifest |
| AMSR2 frequency channels used | 18.7 GHz and 36.5 GHz for melt-pond fraction proxy; 6.9–89 GHz full range for ice concentration |
| BGC-Argo profile interval | Typically every 5–10 days; depth resolution 1–5 m; Arctic fleet sparse (<50 floats mid-2020s) |
| MODIS-Aqua archive depth | 2002–present; AMSR2 from 2012; AMSR-E predecessor 2002–2011 |
| Minimum detectable chlorophyll (open water) | ~0.05 mg m⁻³ in clear conditions; uncertainty exceeds 50 % in heavily ice-contaminated MIZ pixels |
| Under-ice chlorophyll estimate uncertainty | Factor of 2–3 in absolute concentration; bloom-onset timing uncertainty typically ±1–2 weeks |
Analytics Satellize can run
| Daily MIZ chlorophyll map with ice-contamination flags | Standard ocean-colour chlorophyll algorithms (OC3M / OCI) applied to MODIS-Aqua or PACE OCI; per-pixel ice-fraction masking using AMSR2 concentration; uncertainty propagation from atmospheric correction residuals | GeoTIFF layer, daily, with per-pixel quality and uncertainty band; ingested into client GIS or delivered via API |
| Under-ice PAR transmission field | Sea-ice optical model (e.g. Grenfell and Maykut transmittance parameterisation) driven by AMSR2 melt-pond fraction proxy and snow-depth estimates; solar elevation computed from date and latitude | Gridded NetCDF at 12.5 km, daily, covering user-defined Arctic domain |
| Under-ice bloom probability and approximate biomass | Nutrient-limited phytoplankton growth model constrained by PAR transmission field; initialised from climatological nutrient profiles; validated against available BGC-Argo float profiles within domain | Probabilistic raster layer (bloom probability 0–1) plus estimated chlorophyll-a concentration range; updated daily during melt season |
| Bloom phenology calendar | Time-series change-point detection on combined MIZ chlorophyll and modelled under-ice signal; compared against MODIS-Aqua climatological baseline (2002–present) | Per-grid-cell report of estimated bloom initiation, peak, and senescence dates with confidence intervals; annual PDF summary |
| Inter-annual trend analysis | Mann-Kendall trend test on bloom-onset dates and peak chlorophyll from MODIS-Aqua archive; correlation with melt-onset dates from AMSR2/AMSR-E passive microwave record | Decadal trend maps and statistical summary tables; suitable for inclusion in environmental impact assessments or research publications |
| BGC-Argo validation summary | Collocation of float profiles within model domain; comparison of modelled vs. observed subsurface chlorophyll; non-photochemical quenching correction applied to fluorescence profiles | Validation report with bias and RMSE statistics; flagged where model diverges from float observations, with explanatory notes |
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.