The Labless Sensor · field trials 2027

Water-quality estimates on a timed schedule, without reagents or a trip to the lab

Farms, utilities and river programs mostly learn what is in their water from grab samples analyzed days later. With our R&D partner Laguna Innovation (Israel), we are co-developing a low-cost sensor that estimates key indicators every 15 to 30 minutes and says how far to trust each reading. In 2027 we test it against accredited lab results at real sites.

Illustration of a river monitoring site: a small enclosure and solar panel on a post on the bank, with a cable running to a sensor in a pipe in the stream
Planned field installation: sensor in the stream, enclosure and solar panel on the bank.

What it estimates

Three physically different sensing channels feed an on-device model. No liquid reagents, a few watts of power, and readings on a fixed, unattended schedule.

NitrateFertilizer runoff, drinking-water and reuse limits
Salinity and conductivityIrrigation water, drainage, reuse
Organic loadWastewater strength and treatment performance
Heavy-metal signaturesScreening that flags when to send a lab sample

Every estimate comes with an uncertainty range, and the sensor flags readings it should not be trusted on, such as a fouled sensor or water unlike anything it was validated on, instead of reporting a confident wrong number. Microbial signals are early warning only, not a substitute for lab microbiology. There is no turbidity channel yet.

Why field trials

The models that turn the sensor's signals into estimates have to be trained and checked on real water. We place one sensor node beside an existing monitoring station or sampling point for a season or more. The node takes readings on its own schedule, typically every 15 to 30 minutes. Each time the host takes its routine lab sample, we line that result up with the node's reading from the same time. Over months, those pairs show where the estimates hold up and where they miss, and they train the models that make the estimates.

Where the project stands. The node is at the software and bench stage. The first field unit is planned for February 2027, with installs from spring 2027. Its readings are estimates, not a substitute for certified lab or regulatory testing, and the models need field training and validation before anyone should rely on them.

How a site is set up

The node is self-contained. It does not connect to the host's instruments, data logger or data stream.

What gets installed

ItemWhat it isWhere it goes
Sensor headA flow-through block about 10 × 3 × 3 inches (25 × 8 × 8 cm) carrying three sensing channelsIn the water beside the existing probe, in its own deployment pipe, or on a tap or bypass line at a treatment site
EnclosureA small weatherproof box with the edge computer that runs the model at the site, a cellular modem and the battery or power supplyOn its own post or bracket, wherever the host prefers
PowerA few watts. A small solar panel and battery, or spare station power if the host offers itOn the same post
Data linkThe node's own cellular modem. Nothing connects to the host's logger or telemetryIn the enclosure
CablingOne sensor cable from the water to the enclosure, plus a power lead, in protective conduitRouted and secured only where the host allows
Design of the sensor head: a black machined block with a clear top showing the flow channel, optical cell, electrode strip, impedance pins and temperature probe
Sensor head design. The first field unit is being built now; final hardware may differ.

Three sensing channels

  • Deep-UV absorbance and fluorescence: nitrate and organic load
  • Stripping voltammetry: heavy-metal signatures
  • Impedance spectroscopy: conductivity and salinity

An on-device model combines the channels. Every estimate carries an uncertainty range, and the node flags readings it should not be trusted on, such as a fouled window or water unlike anything it was validated on.

Fluorescence gives early warning of microbial load only; it is not a substitute for laboratory microbiology. The node does not yet have a turbidity channel.

How the readings are checked

Every routine sample the host already takes becomes a reference point. Where a site has no lab program, samples go to an accredited laboratory.

Sensor estimateCompared with
NitrateLab nitrate + nitrite as N (EPA 353.2)
Organic loadCOD (EPA 410.4) and TOC/DOC (SM 5310); BOD5 at treatment and wastewater sites
Conductivity / salinitySpecific conductance (SM 2510) and TDS
Heavy-metal signaturesDissolved metals by ICP-MS (EPA 200.8)

Each sample is matched to the sensor reading from the same time, with field duplicates and blanks for quality control. The host gets the full comparison, including where the estimates miss.

A trial, step by step

  1. Before installSite walk and agreement. We confirm the spot, power, access and any paperwork the host needs, such as a written agreement or an insurance certificate.
  2. Spring 2027Install. One visit: post, enclosure, cable and sensor head. The host's equipment is not touched.
  3. Every 4–6 weeksService visit. Clean the optical window and electrodes, check each channel against standards, take a reading before and after cleaning, and pull data. Wastewater sites start at every 1–2 weeks.
  4. Each lab samplePair the data. The host's routine result is matched with the node's reading from the same time.
  5. WinterFreeze plan. We protect the sensor head from ice or pull the unit until spring.
  6. End of trialReport. The host receives the full comparison. The node stays on a data plan or we remove it at no cost.
Illustration of a technician at a stream edge cleaning the window of a compact sensor, with sample bottles in a field case
What a routine service visit will look like, every 4 to 6 weeks.

What a host provides, and what we cover

The host

  • Room for the sensor head near the water and a post for the enclosure
  • Permission to install, and help with any landowner or right-of-way access
  • Routine lab results for the site, with sample times, where they exist
  • Any requirements of its own: agreement, insurance certificate, notice before visits

Tolman Solutions

  • The node, install and all maintenance, at our cost
  • Reference lab testing where the site has none
  • Readings kept outside the host's QA program and data systems
  • The full comparison report; the host's site data stays the host's

Host a trial

We are looking for a small number of monitoring stations, farms, wineries and treatment sites for 2027. If you run one and want to see how a low-cost sensor compares with your own lab results, get in touch.