How TYRN SEA works

What blooms and mucilage are

An algal bloom is a rapid, abundant growth of phytoplankton, the microscopic grass of the sea. On the Calabrian coast it typically appears as green water — often the non-toxic microalga Pyramimonas. It is a natural, seasonal phenomenon.

The mucilage is a gelatinous aggregate that forms after a bloom, when the sea stays warm, stratified and poorly mixed for a long time: in those conditions the organic matter produced by the algae does not disperse and thickens at the surface.

Where the data comes from

TYRN SEA does not install sensors at sea: it uses data from the Copernicus Marine Service, the European Union's ocean observation programme. For the Mediterranean, the Med-MFS (physics) and MedBFM (biogeochemistry) models integrate satellite, buoy and ship data and produce a daily forecast up to 10 days ahead.

TYRN SEA relies mainly on two physical quantities, which the model forecasts reliably:

• Sea temperature (SST) — the engine of the coastal bloom: marine heat waves trigger it.
• Mixed-layer depth — measures how stratified and calm the sea is: a shallow layer means a still sea, prone to stagnation.

The model-forecast chlorophyll, used in the first version of the index, is no longer used: at a 4 km resolution it does not capture the coastal green-water events. The section on the index computation below explains this in detail.

How the index is computed, step by step

The index was redesigned after a check against real data. The first version relied on the model's forecast chlorophyll; comparison with ARPACAL data showed that the coastal green-water events are too local to appear in the model chlorophyll (4 km resolution). The current version therefore relies on the physical preconditions that generate the bloom — which the Copernicus model captures well.

Step 1 — The forecast data

Several times a day the system automatically downloads the 5-day forecast of sea temperature and mixed-layer depth, for each of the 6 coastal zones. Copernicus publishes up to 10 days but we stop at 5: beyond the third or fourth day the model skill drops rapidly for coastal dynamics, and showing peaks at +6/+7 days with low reliability would be misleading.

Not every forecast day carries the same weight: the model is most reliable in the first 1-3 days and its accuracy gradually declines towards the seventh. The more distant days should be read as a trend, not as a precise value — see «How much you can trust it» below.

What we publish (overview)

The 5-day forecast is the heart of the system, but it is flanked by other data layers, each with its own spatial and temporal scale:

• 5-day forecast — 6 coastal zones, bloom and mucilage indices 0-100, physical Copernicus Med model at about 4 km
• Live — SST, anomaly, MHW, stratification, wind and waves today; for the Amantea-Lamezia zone also Sentinel-3 chlorophyll at 1 km and Sentinel-2 NDCI at 20 m
• Seasonal outlook — 3-month forecast of 500 hPa pressure (ECMWF SEAS5), subtropical ridge ↔ Atlantic flow: qualitative indication of seasonal MHW risk
• Archive — reanalysis 1999-2025 of the three indicators on the 6 zones (maps and timeseries)
• Bathing — official ARPACAL analyses + pollution bans, synced daily
• Citizen reports — photo + GPS + type of phenomenon: the only reliable source for shoreline blooms (5-7 m, sub-pixel for satellites). Public CC-BY 4.0 API for scientific reuse

Note on the climatologies: on the forecast / live / archive pages we use 1999-2025 as the reference period (extended to today to represent the actual recent climate, not the 30-year past). The Studies pages, doing academic climate analysis, use the WMO 1991-2020 standard. The seasonal outlook uses 1993-2016, the ECMWF SEAS5 model standard. Each climatology is consistent with the dataset it is computed from.

Step 2 — Comparison with the seasonal normal

Each forecast value is compared with the climatology: the statistics of that exact day of the year over 27 years of data (1999-2025). This identifies what is genuinely anomalous for the season.

Step 3 — Thermal window

The coastal bloom requires warm water. A thermal factor is 0 below ~21 °C and rises to 1 in the warm window (24 °C and above). This is why in spring, with a cold sea, the index is near zero.

Step 4 — Stratification and calm sea

We measure the absolute value of the wind-mixed layer (MLD, Mixed Layer Depth): MLD ≤ 5 m = max stratification (still sea, vertical exchange interrupted); MLD ≥ 30 m = well mixed. The physical principle is well documented as a precondition for coastal blooms [7]: stratification isolates the surface layer, accumulates residual nutrients from river input, and limits the dispersion of phytoplanktonic cells. A shallow MLD typically also indicates weak wind.

Step 5 — Marine heat wave

International operational definition by Hobday et al. 2016 [5]: a marine heatwave (MHW) is a period of at least 5 consecutive days in which the temperature exceeds the 90th climatological percentile (computed over our reference period 1999-2025). The Hobday scale assigns 4 intensity categories: Moderate, Strong, Severe, Extreme. For the Mediterranean coasts, Pisano et al. 2025 [7] quantify an SST increase twice the global mean: this makes MHW increasingly frequent and prolonged. Internal validation on real data confirmed that the major green-water events (July 2025: 12 million cells/L at Lamezia) coincide with intense heat waves.

Scientific caveat — Li et al. 2024 [3] have shown that in deeper Mediterranean-NW waters winter heat waves can suppress the spring bloom by up to -70% because they induce early stratification and block nutrient renewal from deep waters. Our model, focused on the Calabrian coastal waters (typical depth < 50 m), addresses a different regime, with nutrients arriving here mainly from surface fluvial inputs; nonetheless we keep this caveat in mind for long-term winter risk interpretations.

Step 6 — Bloom precondition index

The bloom precondition index (0-100) combines the three factors: thermal window × stratification × (heat-wave intensity). It is high when warm water, a still sea and a heat wave coincide. The multiplicative principle of physical co-stressors reflects the HAB literature consensus [4] [1]: none of the factors alone is sufficient, but their coincidence multiplies the effects.

Step 7 — Mucilage precondition index

The mucilage is the bloom that persists under strong stagnation: it is the bloom precondition index multiplied by stratification. It stays low if the sea, though warm, is not still long enough. The SST-mucilage outbreak frequency correlation in the Mediterranean was established on 60 years of data by Danovaro et al. 2009 [8] (Spearman rs=0.50; p<0.003).

Methodological note — what we tested and discarded

For methodological transparency, we tested the addition of a Redfield N:P factor (surface nutrient nitrogen-to-phosphorus ratio from Copernicus biogeochemical reanalysis) as a potential additional predictor of mucilage. The 1 June 2026 verification over the 5 TYRN SEA zones plus the Amantea-Lamezia focus showed N:P ratios between 1.3 and 9.3, well below the classic 16:1 Redfield ratio. Result consistent with the oligotrophic and P-limited structure of the Mediterranean (Krom et al. 2010 [9]): in the Mediterranean basin the absolute N:P ratio does not discriminate between zones with low and high mucilage preconditions, and applying it as a multiplicative factor flattens the signal rather than enhancing it. We therefore excluded it from the operational model. More promising for the future is monitoring the seasonal anomaly of nutrients against their climatology (a proxy of post-rain river runoff), which we defer to a forthcoming iteration.

Version 1.6 Beta. The model is under continuous refinement and validation: forecasts are compared with satellite observations and with official ARPACAL data (microscopy, water chemistry). Thresholds will be tuned on real data at the end of the season.

Validation against real-world data

The v3 index is not merely theoretical: it was tested in hindcast — reconstructed from the 1999-2025 Copernicus reanalysis — against 10 bloom events documented under the microscope by ARPACAL along the Calabrian coast in 2024 and 2025.

• July 2025, fully captured. The five green-water events of July 2025, including the Pyramimonas bloom of 12 million cells per litre at Lamezia, all fall within the most extreme 1% of all summer days in the 1999-2025 record. The index flags them at its maximum value.
• An anomalous summer. In July 2025 the mean index for the southern Tyrrhenian coast was 5.5 times the July climatological norm, during a full marine heatwave. The bloom was therefore foreseeable in advance from the physical preconditions alone.
• A known limitation. Late-summer blooms of Gymnodinium (a dinoflagellate) develop even without a marine heatwave: the index places them at the «moderate» level rather than «high». This is a different ecological mechanism from the green-water bloom the index is calibrated on — a limitation we state openly.

Validation last updated on 20 May 2026, based on official ARPACAL microscopy data.

Direct satellite validation — Lamezia 06/07/2025 case

Beyond the precondition-index validation, we also checked whether the satellite signal used by the Sentinel-2 pipeline (developed in May 2026 for the «Amantea-Lamezia» section of the Live page) really captures real-world blooms on our coast.

On 6 July 2025 ARPACAL sampled the seawater in front of the La Marinella beach club at Lamezia Terme (38.9249°N - 16.2152°E) following a report of «anomalous greenish discoloration». The lab report (no. 25CZ1528B/01-1) quantified a bloom of Pyramimonas sp. of 12,089,100 cells per litre, equivalent to about 6–12 mg/m³ of chlorophyll.

We pulled the closest Sentinel-2 Earth-observation scene (5 July 2025, cloud cover 9.5%) and compared the NDCI over the same pixel with a pre-bloom scene from 15 June 2025. At the exact sampling point (250 m radius, ~150 S2 pixels):

	Pre-bloom 15/06/2025	At bloom 05/07/2025
NDCI median	+0,0033	+0,0071 (+115%)
NDCI p95	+0,0107	+0,0163
Valid water pixels	354	313

At first glance, a pre-bloom scene (15/06/2025) and the event scene (05/07/2025) already show a slight increase in green-yellow pixels near the point. But to avoid rushing to conclusions from a two-date comparison, we built a timeseries of nine Sentinel-2 scenes over the same pixel from April to end of July 2025.

The timeseries result is scientifically sober: the 05/07/2025 NDCI (day of the ARPACAL bloom) is +0.0067, completely within the seasonal range –0.012 in April to +0.013 at end of June. The seasonal median peaked on 30/06/2025 (+0.0133), one week before the documented event. The single-pixel NDCI does not discriminate the 12 M cell/L Pyramimonas bloom from the seasonal background noise.

Why it still makes sense to use it. Single-pixel NDCI is not enough, but it remains one of the highest-resolution signals available (20 m, vs 4.6 km of the Copernicus Med model). We do not use it as a threshold, but as one of several signals to be combined:

• Temporal comparison over the same pixel: the fact that the bloom is not an outlier indicates seasonal noise is present, but a 50% change over 7 days is still a useful signal to flag in an alert.
• Spatial pattern: a coherent plume attached to the coast is more informative than a point value.
• Combination with other sources: SST, stratification, wind, NRT Sentinel-3 at 1 km, citizen reports. Sentinel-2 NDCI contributes to confirm/exclude a suspicion, not to raise it alone.

In other words: publishing a single NDCI value over a beach as «bloom probability» would be scientifically incorrect. We use it as one of three quadrants in the «Amantea-Lamezia» section of Live, with an explicit note that interpretation must be cross-checked with the other indicators.

Scientific reuse of citizen reports

Reports submitted by the public (photo + GPS + type of phenomenon + description) are a structural component of the TYRN SEA scientific pipeline, not stand-alone observations. They are used to:

• Validate the model: each report is a ground-truth point to compare against the bloom precondition index computed by the model for the same date and zone.
• Correlate with Copernicus conditions: we cross-check SST, MLD, Sentinel-2/3 NDCI from the days preceding each report to build an empirical statistic of conditions that actually precede Calabrian blooms.
• Cover the sub-pixel coastal scale the satellites cannot see (5–7 m shoreline strips, see below): here citizen reports are the only available source.
• Cross-check against ARPACAL: when a report matches an official ARPACAL sampling, we have double independent evidence and can tune thresholds on real data.

Reuse infrastructure

• Public JSON API: /tyrn_sea/api/segnalazioni.json.php (read-only, no email/IP, params from/to/bbox/tipo/limit, open CORS)
• Weekly automatic sync via GitHub Actions (sync_segnalazioni.yml) to CSV + GeoJSON + aggregate stats (per TYRN SEA zone, type, month, day) committed to the repo.
• Data licence: CC-BY 4.0 (attribution: «TYRN SEA citizen reports - studiopitagora.biz/tyrn_sea»)
• Privacy: email stays private (never exposed by the API), IP is stored only as SHA-256 hash for rate limiting.

Want to use the data for a study? The API is public, a simple HTTP request is enough. For complete datasets (with photos) or academic collaboration write to studio.pitagora @ gmail.com.

Final confirmation: bloom-pixel vs open-sea-pixel comparison

To close the loop we compared the NDCI timeseries over the ARPACAL pixel at La Marinella with a second point off Gizzeria (38.9720°N - 16.1455°E, ~5 km to the northwest, open sea, never reported for green water):

On 05/07/2025, day of the 12 M cell/L Pyramimonas bloom in front of La Marinella, the NDCI at the ARPACAL pixel was +0.0067. The same day, over the open-sea pixel 5 km away where no green water has ever been reported, the NDCI was +0.0098: higher. The satellite signal does not distinguish the real bloom.

The physical explanation is simple: in Calabria, coastal blooms of Pyramimonas and the like appear as a 5–7 metre wide strip along the shoreline, no more. A Sentinel-2 pixel is 20×20 metres, and the pixels closest to the coast are often masked by the classifier's land contamination. The phenomenon is simply sub-pixel: no currently available free satellite can see it. A citizen-science channel (user reports) or a shore-based sensor is required to cover this type of event.

How to read the two indices

Both indices range from 0 to 100. The higher the value, the more favourable the forecast conditions.

Index	Level	Meaning
0–24	Low	Conditions within the seasonal norm
25–49	Moderate	Preconditions partly present
50–74	High	Conditions favourable to a bloom
75–100	Very high	Heat, stagnation and heat wave together

The «Live» page

Alongside the forecast, the site has a Live page that shows the current physical state of the sea — not a forecast, but the most recent datum observed and analysed by the Copernicus model. The pages refresh automatically several times a day: the exact time of the last update is shown at the top of each page. The Copernicus datum has, in any case, about a 24-hour latency, the standard for marine satellite data. The values are available for all 6 coastal zones.

The page reports four indicators:

• Sea temperature (SST) — today's temperature, shown next to the reference seasonal norm.
• Anomaly — how far the temperature deviates from the mean of that day of the year over the 1999-2025 period (Copernicus Med Sea Physics reanalysis). Positive values (sea warmer than normal) in red or orange, negative ones in blue.
• Marine heatwave — flags whether a heatwave is under way (at least 5 consecutive days above the 90th climatological percentile), with its category I-IV (Hobday scale) and duration; it also shows how many of the last 30 days were above the norm.
• Stratification — how still and layered the sea is: 0 = well mixed, 100 = strongly stratified. It is the same physical quantity used in the bloom precondition index.

A chart finally shows the course of the temperature over the last 45 days, compared with the seasonal norm and with the heatwave threshold.

In short: the Forecast page tells you how favourable the coming days will be to bloom and mucilage; the Live page tells you the state measured today of the sea. The two complement each other, because today's physics is the engine of tomorrow's conditions.

How much you can trust it

The forecast is most reliable in the first 1–3 days and gradually less precise towards the seventh. It is a trend indicator, not a certainty.

The Copernicus data TYRN SEA downloads every day comes from the near-real-time analysis and forecast system: by its very nature it is preliminary data. Copernicus later reprocesses it — cross-checking it against all the satellite and in-situ observations gathered in the meantime — and publishes, several months later, a consolidated, more accurate version (the reanalysis). The most recent values shown on the site, especially on the «Live» page, may therefore be slightly revised afterwards. The Archive and the reference climatologies, by contrast, use the already-consolidated reanalysis (1999–2025), which is not subject to revision.

• What it forecasts. The favourable conditions for a bloom, not the single patch: the index describes the marine basin, not the single beach.
• Depth. The data refer to the first 5–10 metres of water.
• Validation. The index is calibrated on the real events documented by ARPACAL and by satellite.

TYRN SEA is an informational tool: for bathing-water safety the official ARPACAL checks are authoritative.

How the app was built

TYRN SEA is an independent project. Part of the software development — the computation code, the web pages and the update automations — was carried out with the support of Claude Code, Anthropic's AI coding assistant. The scientific design, the choice of data sources and the interpretation of the results remain the responsibility of the author.