Virridy Home | Lume — Water Quality Sensing Water for Carbon
Produce Safety

Produce Water Sentinel

Fresh produce keeps making people sick, and the industry cannot tell whether anything it does about it works. Lume closes that gap with continuous fecal-indicator monitoring at the two points where water touches the crop — the irrigation source before harvest, and process water before it contacts product — giving growers, processors, and the brands that stand behind them a shared, auditable record of the water that reaches food.

A new application of a proven sensor. The performance figures below are from validated Lume deployments in drinking, recreational, and coastal water — see Research.
Why Now

The 2026 cyclospora outbreak is a symptom, not the story

By mid-July 2026, public health officials had reported more than 8,600 cyclosporiasis cases, a record, against roughly 3,000 to 3,600 in each of the four preceding years. FDA initially attributed a large share to iceberg lettuce supplied to Taco Bell by Taylor Farms, driving a recall across 27 states, then announced the identification was a mistake. Five states have reported 1,644 illnesses among people with that exposure, with 94 hospitalizations.

The parasite reaches produce the same way most produce pathogens do: it is shed in human feces and arrives on the crop through contaminated irrigation water. Cyclospora is the headline, but it is a small share of a much larger problem. Contaminated salad greens alone are associated with an estimated 2.3 million illnesses and $5.2 billion in annual health care costs, and produce-borne pathogens account for roughly a quarter of the estimated 1,300 annual deaths from the most commonly recognized foodborne agents.

8,600+US cyclosporiasis cases as of 17 Jul 2026, a record year
2.3MAnnual illnesses attributed to contaminated salad greens
$5.2BAnnual US health care cost of those illnesses
48MAnnual US foodborne illness episodes, few ever traced to a food
The Gap

Nobody can measure whether an intervention worked

The recall machinery is downstream of the real failure. As Timothy Lytton put it in the New York Times this month, growers can switch from canal to well water, and processors can gas lettuce with ozone, and neither can demonstrate that it reduced illness. Surveillance captures only a small fraction of foodborne illness, doctors rarely test for a pathogen that resolves in days, and patients cannot recall what they ate two weeks ago. Of 48 million annual episodes, only a tiny fraction are ever linked to a food category and a smaller fraction to a specific product.

That is an outcome-measurement problem, and no water sensor solves it. What a sensor can fix is the exposure side. If you have a continuous record of what was in the water before and after an intervention, the intervention finally has something to be evaluated against. Grab sampling cannot do this: it is a snapshot on a fixed schedule, structurally blind to the transient events, a storm, an upstream discharge, a failed lift station, that carry the contamination.

What is measured today

Periodic grab samples sent to a lab, with a multi-day turnaround, on a calendar that has no relationship to when contamination actually occurs.

What that misses

Excursions. Fecal loading in surface irrigation sources is event-driven and short-lived. A weekly sample has a low probability of intersecting the event that matters.

What continuous data adds

A defensible baseline, detection of the excursion while the water is still in the ditch, and a before-and-after record that makes a capital intervention auditable.

Who Carries the Risk

The brand takes the heat for water it never sees

When contaminated produce reaches a consumer, the name in the headline and on the recall notice is rarely the farm. It is the restaurant that served the salad and the retailer whose store brand is on the bag. In the 2026 cyclospora outbreak a national quick-service brand was publicly tied to the illnesses on the strength of an identification the FDA later retracted — the reputational damage lands first and is corrected later, if at all.

Those brands carry the liability and the consumer trust, yet they have no direct visibility into the irrigation and process water that decided whether the crop was safe. Their only instrument today is a supplier audit built on periodic grab samples. Continuous monitoring changes what a brand can require and verify: instead of trusting that a supplier followed a procedure, it can hold a continuous record of the water that actually touched the product — the difference between “we met the audit” and “here is the evidence.”

Restaurant & foodservice brands

Bear the outbreak headline and the recall. Continuous supplier-water data turns a food-safety specification from a paper requirement into a measured one.

Retailers & store brands

Own the brand of record on private-label produce, and the recall that follows it. A shared water record de-risks the categories most exposed to fecal contamination.

Growers & processors

Where the water is, and where the measurement happens. An auditable record becomes an asset in buyer negotiations, not just a compliance cost.

The Measurement

A direct measurement, not an inference

Tryptophan-like fluorescence is not a proxy. Illuminated at 275 nm, tryptophan re-emits at 340 nm, and TLF is a direct optical measurement of free tryptophan, tryptophan-containing peptides, and exposed tryptophan residues in the water column. That dissolved-protein pool is itself a near-direct measure of fresh microbial biomass: free tryptophan in surface water has a half-life of days, so the signal reflects recent contamination rather than legacy organic matter. Lume is calibrated in CFU/100 mL because CFU is the regulatory unit, not because culture is the more fundamental quantity.

Fecal sources dominate the signal

Enteric bacteria are rich in tryptophan-bearing membrane, pili and flagellar proteins, and lyse rapidly on leaving the gut. Fecal plumes deliver this material at 106–108 cells per mL, orders of magnitude above the slow turnover of sparse, oligotrophic native bacteria. E. coli additionally expresses tryptophanase, the metabolic trait behind the century-old Indole Test, and both tryptophan and E. coli degrade on similar environmental timescales.

Agricultural organics are optically separable

The dominant non-fecal organics in farm and surface water, humic and fulvic acids from soil and plant decay, fluoresce in a different region of the excitation-emission matrix (roughly 340/440 nm) and contribute little to the tryptophan channel at 275/340 nm. Built-in turbidity and temperature channels correct the remaining confounders that defeat static-threshold TLF instruments.

One scoping note that is a matter of fact rather than caution: TLF measures fecal and microbial loading, not a named organism. It does not speciate cyclospora. That is the correct scope, because cyclospora reaches produce through fecally contaminated irrigation water, which is exactly what the instrument measures.

The Product

Two control points, one instrument

Water touches the crop twice, and both contacts are measurable with the same sensor and the same calibration. Because Lume resolves down to drinking water concentrations, the plant can be instrumented on either side of chlorine injection: the intake gives source quality and incoming load, the post-dose line gives the water actually contacting product. Measuring both turns disinfection from an assumed step into a verified one, which is the difference between a dosing record and evidence.

Control point 1
Pre-harvest irrigation source

Canal, reservoir, well or surface intake. Event-driven fecal loading from upstream human and animal sources. Continuous TLF establishes a source baseline and flags excursions before the water reaches the field.

Lume in situ
Field
Growing and harvest

Contamination acquired here travels with the crop. No water sensor observes this stage, which is why the source measurement upstream is the one that carries information.

Control point 2
Post-harvest process water intake

Plant intake, reconditioned or recycled wash water returning to the flume, final potable rinse, hydrocooling and ice-making supply. Measured pre-dose, post-dose, or both.

Lume in line

Control point 2 carries most of the regulatory weight, and being able to read across the disinfection step is what makes it commercially interesting rather than a compliance box. A plant that can show incoming load, post-dose residual quality, and the effect of a water change has an operating instrument, not just an audit artifact. The value is highest where intake quality varies: plants on their own well or surface source, and plants reconditioning wash water back into the flume, which is increasingly common under water cost and drought pressure.

Validated Performance

Already proven at drinking water concentrations

The measurement question a produce buyer will ask is whether the instrument resolves clean water, because a treated intake and a post-chlorination rinse sit far below the concentrations found in a river. That case is already validated. Lume was tested against 216 paired Compartment Bag Test observations in Rwanda and Kenya, classifying at the ≥10 CFU/100 mL drinking water threshold with 85% balanced accuracy (76% sensitivity, 93% specificity), approaching the reference method’s own inter-method reproducibility ceiling. Detection extends to below 0.1 ppb tryptophan in deionized water.

< 0.1 ppbMinimum detection limit, tryptophan in deionized water
85%Balanced accuracy at the ≥10 CFU/100 mL drinking water threshold (n=216 paired CBT)
> 94%Categorical accuracy, site-specific calibrated contamination risk
7%Mean absolute percentage error in log concentration space vs. culture
60 sTime to reading, against 24–48 h for culture-based methods
76%Categorical accuracy in initial ocean deployment from six training samples

Two consequences for this application. Chlorinated process water is within range, so the intake can be measured before and after dosing rather than only at the source. And because the instrument performs at drinking-water concentrations, it operates in the range where the post-harvest standard lives — giving continuous assurance that disinfection is holding and an immediate flag when it is not. Lume complements the confirmatory laboratory test that certifies conformity; it does not replace it.

Regulatory Hook

The two control points sit under two different standards

Pre-harvest agricultural water (FSMA Produce Safety Rule)The 2024 revision replaced fixed numeric microbial criteria with a pre-harvest agricultural water assessment. An assessment-based rule rewards operators who hold continuous evidence about their source, rather than a handful of grab results.
Post-harvest water (Produce Safety Rule, subpart E)Water contacting produce after harvest is held to no detectable generic E. coli. That is a bright line, and continuous monitoring against a bright line is a far stronger compliance story than monitoring against a risk narrative.
A note on scopeVirridy’s EPA Alternate Test Procedure work sits under the Clean Water Act (40 CFR 136.4) — ambient and wastewater. Produce safety is an FDA domain, so that approval does not carry over here; the produce case rests on the field validation above, not on the ATP.
Weakening federal capacityUSDA food safety research funding has been disrupted, CDC parasitic disease staffing cut, and the National Advisory Committee on Microbiological Criteria for Foods eliminated. Private continuous measurement becomes more load-bearing as public capacity contracts.
How to Engage

A pilot, not a purchase order

Lume is validated in drinking, recreational, and coastal water; produce is a new application of the same instrument and the same calibration. The right first step is a season of continuous monitoring on a working irrigation source or process-water line, run alongside whatever grab sampling already happens there, so the produce sector builds its own dataset rather than borrowing ours from rivers.

Growers, districts & processors

Instrument a source or process-water line for a season. You get an early-warning record of your own water and a defensible baseline; we get a produce-specific validation. Deployment is a short step from the monitoring we already run for utilities and regulators.

Brands, retailers & foodservice

Put continuous water data behind your supplier food-safety program. We will stand up a pilot on a supplier’s water and share the record with your food-safety team, so you can see what continuous evidence adds to what your audits already require.

Associations, auditors & researchers

Help define what continuous water data has to prove before it can count in a pre-harvest assessment or an audit. We would rather build to that specification from the start than deliver another dashboard that sits beside the record instead of counting toward it.

Water-instrumentation channels

The same sensor and calibration cover agricultural and process water. For partners already selling water instrumentation into food and beverage accounts, produce is a natural extension of an existing channel.

What we bringA continuous fecal-indicator sensor validated against culture from drinking-water concentrations upward, running today for utilities and regulators, at both points where water touches the crop.
What we are looking forA working site for a season, access to the grab sampling already happening there to compare against, and partners who want the produce sector to have continuous water evidence of its own.
Positioning

The defensible claim

Lume provides a continuous, auditable record of microbial loading in the water that touches food, at both points where that water is still controllable, at concentrations from drinking water upward and on either side of disinfection. It does not replace laboratory microbiology, and it does not speciate a parasite. It makes a prevention investment measurable, and measurability is precisely what the fresh produce sector has never had.

This is the same accountability argument Virridy makes in surface water and in carbon, applied to a sector that is under acute public pressure and has no continuous measurement layer of its own.