How conditions are calculated

Three tiers of certainty

Not all layers carry equal weight. Each is labelled so a reading is never mistaken for more than it is.

Authoritative

CMEMS sea-surface temperature and chlorophyll. Peer-reviewed Copernicus Marine products, suitable for assessment and reporting.

Semi-quantitative

Sentinel-2 water indices (NDCI, turbidity, vegetation). Useful spatial detail, but optical and cloud-limited.

Indicative

Biodiversity occurrences from OBIS and GBIF. Presence-only, uneven sampling effort, not an inventory.

From pixels to one number a day

The temperature and chlorophyll products are grids of pixels. For each day, the site value is the spatial mean of every grid cell that falls inside the MPA boundary.

daily value = mean( pixels inside the boundary )

This single daily mean is what every later step works from: the baseline, the anomaly, and the condition flag.

Each MPA against its own history

Conditions are judged relative to the site itself, not a single regional or regulatory threshold. For every MPA we compute a monthly climatology over 2000 to 2020: the long-term mean and the distribution of daily values for each calendar month, built from the CMEMS multi-year reanalysis. A warm July is compared only with other Julys at that site.

Departure from normal

The anomaly is the plain difference between a day's value and that month's baseline mean.

anomaly = daily mean − climatological mean for that month

A positive chlorophyll anomaly means more than usual for the season, a negative one means less. The same applies to temperature.

What counts as a deviation

Each day is graded by where its value falls in the baseline distribution for that month. The bands are two-sided, so an unusually low value is flagged as readily as an unusually high one. This generalises marine-heatwave percentile logic to both temperature and chlorophyll, and to both tails.

Percentiles (p01, p10, p90, p99) are taken from the site's own 2000 to 2020 distribution for that month. A flag means the value is unusual for this place and season, not that a regulatory limit has been crossed.

Two summary figures

Condition is the worse of two flags, one for temperature and one for chlorophyll. Each is taken from the most recent day with data inside the date range you select, which is why the dashboard prints that date next to the flag and under every reading. It describes that day, not the period. Because the worse of the two wins, a red on either variable shows up straight away.

Over the period counts every day in the range by its own flag, again taking the worse of the two variables for that day. Read it alongside the dated flag: a range can end on a calm day and still contain extreme ones, and the day counts are what show that. Each count opens the days behind it, with the anomaly for both variables and which of them carried the flag, so any number here can be checked against the days it came from. Days without satellite coverage are not counted, so the three numbers can add up to less than the length of the range.

Trend is a least-squares straight-line fit through the anomaly series, reported as change per year. It appears only when the selected range covers at least a year, because a slope measured over a shorter span mostly tracks seasonal swings rather than a real direction.

Two regional indicators

Sea-surface salinity and sea ice come from the same regional physics reanalysis as temperature. They are modelled fields that assimilate observations, so they read as authoritative but are model output rather than direct measurement. Both appear only for the regions that carry them.

Salinity is treated exactly like temperature and chlorophyll: graded against its own 2000 to 2020 monthly baseline, with the same two-sided percentile bands and a deseasonalised per-year trend fit on its anomalies.

Sea ice is shown as a value, not a condition flag: its seasonal, mostly-zero distribution makes percentile bands misleading. A per-year trend is shown only when the selected range spans more than two years, since a straight line through one or two seasonal cycles reflects the cycle and the window edges more than any real change.

High-resolution water indices

From Sentinel-2 surface reflectance (10 m), computed only over cloud-free water pixels. These are semi-quantitative proxies, best read as spatial pattern rather than absolute value.

NDCI

(B05 − B04) / (B05 + B04)

Red-edge chlorophyll proxy, sensitive to near-shore productivity.

NDTI

(B04 − B03) / (B04 + B03)

Turbidity and suspended sediment.

Vegetation

NDVI > 0.10 over water

Indicative submerged or floating vegetation. Optically hard to see, so indicative only.

A surface bloom is flagged for a date when at least 12% of the cloud-free water carries a per-pixel NDCI between 0.15 and 0.9. The mean alone stays near zero because blooms are patchy, so the area fraction is what marks a real event. Dates below ~1,500 valid water pixels are treated as too sparse to judge.

Species occurrences

Occurrence records inside the MPA boundary come from OBIS and GBIF. Records are filtered to marine taxa via the World Register of Marine Species, and only openly licensed records (CC0 and CC-BY) are counted. Seabirds count as marine but dominate coastal data, so they are flagged and can be toggled out. This layer is presence-only and reflects uneven survey effort, so it is indicative of what has been recorded, not a complete species inventory.