A bloom-covered reservoir, a fish kill, and two proposed fixes

Dissolved oxygen is a balance between photosynthesis, community respiration and reaeration from the atmosphere, and each term moved the wrong way.

• Cloud and calm cut light (less photosynthesis) and eliminated wind-driven reaeration and mixing. The 11 mg/L at 16:00 was local supersaturation in a thin surface layer that could not diffuse downwards without turbulence.
• Bloom biomass respires heavily all night. A dense Microcystis population plus bacteria decomposing dead cells can strip the surface layer to near zero by dawn; fish suffocate in the early hours, which is why kills are found at first light.
• The anoxic hypolimnion has been accumulating ammonia (0.9 mg/L unionised NH3 is directly toxic at that concentration), sulphide and reduced iron. Any nocturnal or wind-driven partial mixing, or a convective overturn, injects this cocktail into the surface.
• Scums can clog gills mechanically and locally deoxygenate; some strains release microcystins.

So the kill reflects a nocturnal oxygen minimum plus toxics from the hypolimnion, not a shortage of oxygen at the time the measurement was taken.

Aerators. Ten fountains at the inlet add oxygen locally but cannot destratify or oxygenate 600 ha; they treat a symptom. Worse, the lake's phosphorus problem is partly internal: anoxic sediment releases phosphate, so full-depth oxygenation of the sediment interface would help, while inlet-only fountains will not. Cost is better spent on the outfall.

Mass balance on tilapia. Water-column P ≈ 0.19 g/m³ × 3 m × 6 × 10⁶ m² ≈ 3.4 tonnes, before sediment release. Fifty thousand tilapia at, say, 1 kg each contain roughly 0.3% P wet weight ≈ 150 kg P - about 4% of the standing stock, and only if every fish is harvested and none die or are eaten in the lake. Ecologically, tilapia is non-native, excretes nutrients, and can suppress large zooplankton that graze algae, favouring cyanobacteria.

Verdict. The durable fix is diverting or treating the sewage, cutting external P, then managing internal loading (sediment oxidation, possibly an aluminium-based inactivating agent). Biomanipulation with native planktivorous fish can be a supplement only after external loading is controlled; stocking a non-native cichlid is not a nutrient-removal strategy.

The risk. Microcystins are largely intracellular while the bloom is intact, so routine cell counts and even filtered samples can look reassuring. Chlorine attacks the cells first; lysis releases the toxin into the dissolved phase. Chlorine does oxidise microcystin-LR, but algal cells and extracellular organic matter exert an enormous chlorine demand, so the free residual is consumed before the toxin is destroyed. Chlorinating intact cells can therefore increase dissolved toxin at the tap, and the taste-and-odor compounds released make the water undrinkable before it is unsafe.

Insist on

1. Cell removal before disinfection: coagulation with a polyaluminium or ferric salt, flocculation, sedimentation or dissolved-air flotation, then filtration - this strips intact cells and with them the toxin payload.
2. No pre-chlorination of raw water during a bloom.
3. Powdered activated carbon for any dissolved toxin, dosed on jar-test evidence.
4. Monitoring of dissolved microcystin by ELISA with LC-MS confirmation, against the WHO provisional drinking-water value of 1 µg/L, plus ammonia and dissolved oxygen at intake depth.
5. Abstraction from below the scum layer only if that depth is oxygenated; otherwise treat and blend, and be prepared to stop abstraction.