WLD Data

How EU bathing water is classified (Directive 2006/7/EC), explained with real data

Every designated bathing water site in the EU is classified under Directive 2006/7/EC, the Bathing Water Directive. This note explains what that classification actually measures, using 2,773 sites from Greece and Croatia in this catalog's 2025 status dataset as real examples — not a textbook description.

The classification exists so that national authorities, and the public, have a consistent, comparable measure of bathing water quality across the whole EU, built from the same method and the same four classes everywhere the directive applies. Understanding what that measure does and doesn't capture matters as much as the class itself, which is the point of working through real sites from the published dataset rather than a hypothetical example.

A rolling evaluation, not a single test

Directive 2006/7/EC does not classify a site from one sample. A site's class comes from a statistical evaluation of microbiological monitoring results — levels of intestinal enterococci and _E. coli_ — collected across the bathing seasons leading up to the assessment, evaluated together rather than sample by sample. The exact statistical method and the concentration thresholds for each class are set out in Annexes I and II of the directive on EUR-Lex; this note doesn't restate those thresholds, since they belong to the legal text, not to this dataset.

The four classes

The directive defines four classes: excellent, good, sufficient, and poor. Each class is a band of water quality, from the cleanest (excellent) to the class that triggers management action and bather advice (poor). A site keeps its class until the next season's evaluation updates it; this dataset's `classification_2025` column records each site's current class from that process.

Not every site has one of these four classes yet (more on that below), so the shares here are a share of the sites that do: of 2,773 sites, 2,653 sites are classified, and among those, 97.2% are excellent, 2.3% good, 0.3% sufficient, and 0.2% poor.

What "excellent" doesn't mean

An "excellent" classification is a multi-season statistical result, not a same-day water-quality test — always check local signage and authorities before swimming, not this dataset. It also doesn't mean a site is free of other issues the directive doesn't cover, such as litter, currents, or non-microbiological pollution. And because the directive applies the same method everywhere, a classification is only meaningful within that shared framework — it isn't a ranking against bathing water outside the EU or outside this dataset's two countries.

It also isn't permanent. The evaluation behind each class updates every season as new monitoring results come in and the oldest season in the rolling window drops out, so a site's class can move between seasons in either direction as its own monitoring record changes — the history dataset in the companion note on a decade of classifications shows exactly that movement, season by season, for every site in Greece and Croatia.

Sites without a class yet: "new" and "not-classified"

Not every site in the dataset has a class. Two statuses cover sites that do not: `new`, for a site newly added to the monitoring network that has not yet built up enough seasons of data to be classified, and `not-classified`, for a site that was monitored but did not return enough valid samples that season to support a classification.

Drosia in GR ( Ionian Islands) is one real example of a "new" site in this dataset: it is being monitored, but the current evaluation doesn't yet have the seasons of history the directive's method requires.

Chrysi - Gaidouronisi in GR is an example of a "not-classified" site: it has a monitoring history, but this particular season's samples weren't sufficient to support a class.

Together, these two statuses account for 4.1% and 0.2% of all sites in the dataset — small shares, but large enough that a reader comparing "share of excellent sites" between regions or countries should check how many sites in each group are actually classified at all, not just split the remainder across the four real classes. The companion note comparing Greece and Croatia's 2025 numbers applies exactly that caveat.

Sources and further reading