Online water-quality biosensors

Most online water sensors are physical chemistry: absorbance, scattering, electrochemistry, each tied to a property a molecule or particle carries. A genuine water biosensor does something different. It puts living biology into the sensing path, and for a precise reason: the quantities that matter most in effluent and receiving water — biodegradable organic load, toxicity — are not molecules with a spectral signature. They are summaries over what a living community can do, and only that community can measure them.

Respirometry on a biofilm

The biochemical oxygen demand concept is simple: aerobically degradable organics are food, and eating them consumes oxygen. The standard laboratory test lets a seeded microbial population work for five days, because slow organisms are part of the answer; the number is an operational definition, not a physical constant. An online BOD sensor compresses this into a steady-state measurement. A fixed biofilm grows on a membrane over an oxygen sensor; sample flows past; organics diffuse in; respiration lowers the oxygen flux reaching the sensor. That steady-state deficit is calibrated against the standard method, conventionally using a glucose-and-glutamic-acid reference substrate.

The compression has a price, and the honest instrument states it. A biofilm given minutes measures the readily biodegradable fraction it can metabolize in minutes — related to, but not identical with, the five-day demand, which also counts slowly degraded material. And the biofilm itself is a variable: its composition depends on seed and feed history, a toxic shock can suppress it, and cleaning resets it. The calibration is a treaty with a specific community, not a property of water.

Luminescent bacteria: toxicity as one organism’s opinion

The other reagentless biology reports toxicity rather than load. The marine bacterium Vibrio fischeri emits light as a by-product of electron transport; metabolic poisons dim that light within minutes. An online instrument mixes sample with a bacterial suspension and ratios the light loss against a control, producing a number in minutes to tens of minutes. What it reports is toxicity to one organism under one exposure condition — a fast, cheap, single-species summary. A different species would produce a different number on the same sample, and the ranking of samples can shift between them; that is inherent to a biological summary, not a defect to be engineered away.

The stability problem

The transduction chain is the same as any other sensor’s, but the recognition layer is alive, and this changes the calibration economics. A dye ages slowly and predictably; a biofilm metabolizes, starves, regrows and reorganizes. Its sensitivity drifts with temperature, acclimation and recent history in ways no amount of signal processing removes. The mitigations are procedural: hold the culture at a steady state, check against reference standards on a schedule, replace living cassettes at a defined age rather than at failure.

The honest framing is that such an instrument uses biology as a reagent. The variance of the living element is part of the measurement, exactly as it is in the five-day test — the difference is that the test hides the variance behind a fixed protocol, while the online sensor has to manage it continuously. What makes two readings comparable is not the hardware but the operating discipline that keeps the sensing community the same community between checks.