The Lake Did Not Get 656 Times Dirtier. The Test Finally Looked Smaller.

A net does not measure what passes through it

Many early freshwater surveys pulled a manta net through the surface. The method is useful for larger fragments, but a mesh of 100–300 µm physically cannot retain most particles smaller than its openings. The reported concentration is therefore not “all microplastics.” It is all particles large enough for that sampling and analytical chain to retain and identify.

This is why two credible studies can report numbers separated by orders of magnitude without either being fraudulent. They may be counting different size universes. Our earlier international drinking-water report makes the same caution about comparing countries that use different thresholds.

How the new method works

Hassler and colleagues combined sample digestion and density separation with spectral flow cytometry. Particles pass through an optical measurement zone one by one. Sixty-four detectors spanning 373–812 nm capture a spectral signature, and two machine-learning models decide first whether a particle is plastic and then which trained polymer family it resembles.

The models were trained on 235,955 plastic particles and 258,663 non-plastic particles. They classified PET, polycarbonate, polyurethane and PVC, while polyethylene and polypropylene were grouped because their signatures were not reliably separable. Direct recovery exceeded 95% in most tested natural waters; the full preparation protocol averaged 87% recovery.

The team cross-checked natural and synthetic samples with laser direct infrared spectroscopy. When they aligned the size ranges, the techniques generally landed in the same order of magnitude—important reassurance that the higher counts came from seeing smaller material rather than from an unconstrained algorithm.

What Lake Geneva looked like at smaller scale

Direct measurements in the 5–70 µm band ranged from 32.9 to 687.1 particles per litre in surface waters and from 6.9 to 154.6 per litre through the water column. A seven-sample reproducibility exercise at the lake platform averaged about 121 particles per litre.

The smallest directly measured band dominated: particles from 5–15 µm represented 63.9% at the lake surface, 71.2% in rivers and 90% offshore and at depth. Extending the observed relationship down to 1 µm produced projected averages of 338.6 particles/L for inshore surface water, 680.7/L for deep lake water and 983.4/L for rivers—with wide ranges around each estimate.

Why the headline is 656×—and how not to misuse it

The study’s projection to 1–100 µm exceeded historical manta-trawl lake reports by 226–656 times. That comparison does not mean Lake Geneva became 656 times dirtier. It means the projected size window contains far more particles than the older net retained.

Nor should the upper multiplier be pasted onto every water result. Particle-size distributions differ between rivers, lakes, bottles and treated tap water. The correct takeaway is methodological: always report the lower size cutoff beside a particle count.

What the method still cannot tell us

• It does not directly see below 5 µm. The 1–5 µm portion is a projection, not a counted population.
• It does not yet classify every important polymer. Polystyrene, polyamide, PTFE and rubber were outside the trained set.
• PE and PP remain grouped. The method detects that class but cannot reliably split the two.
• Weathered particles are harder than clean standards. The authors call for more training on naturally altered polymers.
• Environmental risk thresholds remain uncertain. Species-sensitivity comparisons combine different polymers, sizes and organisms.

What this means for readers

It does not change the practical hierarchy in our tap-versus-bottled-water guide. This was natural-water monitoring, not a trial of household treatment. What it changes is how you read a number. “121 particles per litre” is incomplete without “measured from 5–70 µm.” A lower count from a 300 µm net is not automatically cleaner water.

Better instruments often make pollution appear to jump because the invisible part of the distribution becomes countable. That is progress in measurement, not proof of a sudden environmental event.