The paper asks whether low-cost printed sensors can track soil microbial decomposition continuously and without digging the soil up. Each sensor is a printed resistor made from an ink of PHBV, a biodegradable polymer, mixed with carbon flake. As soil microbes degrade the PHBV it swells and pushes the carbon particles apart, so the resistance rises; the rate of that rise is used as a proxy for microbial activity. The sensors were tested in a polytunnel mesocosm experiment with 36 intact cores: species-rich grassland soil versus winter wheat soil, each under drought, flood or control watering, w

European Journal of Soil Science · 2026

Slide the soil water and watch microbes eat a printed circuit — while its resistance climbs.

Active — moist soil

R/R₀ 1.00

resistance ÷ starting value

Day 0 · 00:00

reading #1

Soil microbes decide how much carbon a soil keeps — but watching them work has meant digging the soil up. What if the soil could report, every 30 minutes, how hard they are eating?

Dark platelets: carbon flakes in the ink

Amber sparks: current hopping flake to flake

Pale green rods and threads: bacteria and fungi

White dots: oxygen in air-filled pores

Illustrative model of the mechanism described in the paper: microbes degrade the PHBV binder, it swells, carbon flakes separate and resistance rises. Rates are not to scale. The number on screen is computed from the drawn flake network: each contact conducts less as the PHBV around it is eaten.

01

The trouble with a shovel

When a leaf or a root dies, soil microbes — bacteria and fungi too small to see — break it down. That breaking down is decomposition . It decides how nutrients are released, how well soil holds water and how fertile it stays; soil organic matter, the carbon-based part, is central to soil health. Yet monitoring decomposition in real time, the authors write, “has proved elusive”.

Dig it up

Enzymes, nutrients and texture need a soil sample. Taking it disturbs and destroys the soil you wanted to watch — and disturbing soil artificially boosts microbial activity. Most indicators are taken at a single point in time.

Bury a tea bag

Weighing how much buried material has disappeared does measure decomposition. But mass loss is slow: it can’t show changes over hours or days.

Catch the CO₂

Chambers measure the CO₂ leaving the surface — but about half of it comes from plant roots and their fungal partners, and it swings with time of day, air temperature and plant cover. Stations are expensive; typical setups have only 6–12 sampling points.

Three digs

Somewhere on this timeline is the real decomposition curve from the grassland drought cores. You have three soil cores to sacrifice. Tap the timeline to dig.

Three cores, three moments. Where would you dig?

In this game a ‘dig’ simply reads the sensor curve at one moment. A real dig measures different things — and destroys the core.

Who breathed this CO₂?

The chamber only sees bubbles. About half of soil-surface CO₂ comes from roots and their fungal partners, so a chamber can’t isolate the decomposers.

02

A recipe microbes can eat

The team’s answer is an ink. They blended carbon flake , which conducts electricity, into PHBV — poly(hydroxybutyrate-co-valerate). PHBV resembles natural organic matter that bacteria make from sugars and lipids; many bacteria use it as an energy store, and it is widely used as a biodegradable plastic. A wide range of soil bacteria and fungi degrade it, with or without oxygen.

Current crosses the printed strip by hopping between touching flakes. Its resistance — how hard it is for current to get through — depends on how well those flakes touch. As microbes degrade the PHBV in moist soil, it swells and pushes the flakes apart, so resistance rises over time. How fast it rises reflects how active the microbes are.

Figure 1 (centre, right) from Fry et al. 2026, Eur. J. Soil Sci., CC BY; cropped.

Printed. Three “dog-bone” strips are stencil-printed onto exposed pads on a printed circuit board.

Protected. A 3D-printed carriage (about 138 mm long) shields the connections; a 3D-printed spade cuts a slot in the soil to slide it in.

Logged. Custom low-cost loggers use the voltage-divider principle — 6 channels each — and save a reading every 30 minutes to an SD card.

Why three strips, side by side?

The strips are wired in parallel : current can take any of the three. That controls for random variability or damage to a single trace. Tap a strip to scratch it.

Combined resistance: R ÷ 3 — all three strips carry current.

The team also watched each sensor’s resistance while inserting it, to make sure no trace was damaged going in.

Making sensors comparable

No two printed traces start identical. Here are three made-up sensors with different starting resistances (1.1, 1.9 and 2.8 kΩ) but the same proportional growth, plus day–night wiggles.

Raw resistance: three curves, three starting points.

In the paper, readings above 10 kΩ were removed, each sensor was divided by its own starting resistance R₀ (its average on 28 April, 00:00–12:00), and daily averages removed day–night effects.

03

Is it really the microbes?

A strip whose resistance changes could be reacting to heat, salt or water rather than to hungry microbes. The team’s earlier studies checked. Changes in soil temperature, moisture and salinity caused only temporary, reversible changes. Live soil and biologically active liquids raised resistance permanently. Identical sensors made with non-biodegradable binders showed no response, and PHBV sensors in sterile media showed no response either.

Which will rise and stay up?

Predictions are optional — tick the jars you expect to climb for good, or just watch: the bench runs ten simulated days by itself.

The bench starts on its own in a few seconds.

These checks come from the team’s earlier studies (Atreya et al. 2023; Sharpe et al. 2026), as summarized in this paper. Curves are schematic. In an earlier field trial (Sharpe et al. 2026), 44 sensors ran for 50 days in a UK grassland restoration experiment, and sensor response was positively related to microbial biomass carbon.

04

Thirty-six soil cores, one drought, one flood

Intact cores of soil, 15 cm deep and 10 cm across, came from two farms near Penrith in north-west England with the same sandy loam. One was species-rich grassland — probably never fertilised, with sedge, ribwort plantain, ryegrass, bird’s-foot trefoil and sweet vernal grass. The other grew a winter wheat monoculture. In a polytunnel at Lancaster University’s Hazelrigg station: 2 soils × 3 treatments × 6 replicates = 36 cores.

Drought: left to dry.

Flood: standing in rainwater level with the soil surface.

Control: watered three times a week.

Treatments began on 11 April 2024 and ran 32 days. On 13 May the stress ended: drought cores got 1 L of rainwater and flooded cores were drained. Then every core received 500 mL twice a week until harvest on 18 June. Sensors logged every 30 minutes for about seven weeks — roughly 2,400 readings each.

Grassland · Drought — R/R₀ 1.01 · soil water 2%

The drought ends on 13 May. What will the drought sensors do?

Will flooded wheat behave like flooded grassland?

Measured. Treatment means redrawn from Fig. 2a (values digitized; approximate). Pot water levels from Fig. 4a (approximate). Pot drawings are schematic; each sensor’s glow follows the day-to-day slope of its group’s measured curve.

05

What the sensors heard

To compare groups, the team measured each sensor’s slope — how fast R/R₀ was rising — in a treatment window (1–13 May) and a recovery window (16 May–17 June), then tested whether the two differed.

Median sensor slope, by window

Treatment window (1–13 May): drought sensors sat slightly below zero; flooded wheat was already climbing fastest.

Axis: median sensor slope (as reported in Fig. 2c). Read the comparisons, not the absolute units. p-values compare the two windows within each group.

Drought → a rewetting surge

In both soils, sensor decomposition increased sharply once the soil was rewetted (grassland p = 0.0041; wheat p = 0.0023). The mean R/R₀ dipped slightly below 1 during the drought, then climbed to about 4 by mid-June. This follows a well-known pattern, the Birch effect (Birch 1958): microbial activity jumps when drought stress is removed — thought to be fed by nutrients released when some cells burst on rewetting, plus previously protected organic matter.

Caveat: the drought soil was too dry for the spade, so it was dug out and repacked around the sensor. That may have affected contact and influenced the result.

Flood → a slower comeback in grassland

Grassland sensors recovered after flooding, but significantly more slowly than after drought (p = 0.055 between windows). Under water, oxygen runs out at the sensor, which slows or stops decomposers that need it. The authors suggest recovery is slower because water drains out of soil more slowly than dry soil rewets.

Flooded wheat didn’t slow at all

In wheat soil, flooding did not slow sensor decomposition. There was no significant difference between windows (p = 0.41), and rates were higher throughout than in any other treatment — controls included. Mean R/R₀ reached about 5–6. Why might that be?

06

Water isn’t the whole story

Here is the same microscope, locked to waterlogged soil: the pores are full of water, oxygen is scarce and the microbes that need it stall. The authors suggest that actively growing wheat may have kept an oxygenated rhizosphere — the thin zone of soil around roots. That could happen through greater evapotranspiration (plants pulling water out of the soil), and possibly through aerenchyma : air channels that wheat roots can form when waterlogged.

Oxygen-starved — waterlogged

R/R₀ 1.00

Day 0 · 00:00

Illustrating the authors’ hypothesis — not yet tested. More work is needed: future experiments should measure soil water dynamically and image roots for aerenchyma.

Clues that point this way

Flooded wheat soil was less wet (about 40% water) than flooded grassland soil (about 50%).

Flooded soils showed more activity of some enzymes at harvest.

The takeaway: microbes respond to oxygen as well as water — and plants can change the soil right around a sensor.

07

What a single dig would have said

On 18 June — five weeks after the stress ended — the cores were harvested and analysed the destructive way. The snapshot mostly separated the two soils: wheat soil had higher pH and higher CBH and GLC enzymes; grassland had more root biomass, soil carbon and nitrogen, and more of the POX enzyme. But how did each measure respond to the watering treatments?

Did it detect drought or flood?

Soil carbon and soil nitrogenNo watering effect detected

Microbial biomass carbon and nitrogenNo watering effect detected

Plant biomass, above and below groundLower after drought and after flood

Enzymes PER, POX, NAGHigher after flooding

Weekly CO₂ fluxNo treatment-specific signal; unrelated to sensor slope

Printed sensorsWhen and how strongly activity changed — every 30 minutes

At harvest, soil carbon, nitrogen and microbial biomass showed no watering effect — “indicating either successful recovery or no effect”.

08

A field full of listening posts

Because each sensor is cheap and small, the authors argue, large numbers could be deployed across soil types and treatments — revealing how decomposition responds across space to changes in the environment and in management.

Tap the field to add sensors

24 sensors listening. Predict: where will the brightest flush appear?

An illustration of the possibility the authors describe — not data. Sensors glow by the same illustrative rules as the microscope above.

What comes next

The authors conclude that low-cost printed sensors can track decomposition autonomously, at high resolution in time and space — something that “has, until now, not been possible without repeated sampling”.

Other degradable materials, to monitor specific microbial or enzyme processes.

Genetic analysis of the microbes living on the sensor surface.

Long-term evaluation in the field.

So: can the soil report, every 30 minutes, how hard its microbes are eating? In 36 cores through a drought, a flood and a recovery, these printed strips did — no shovel required until the very end.

Source

The paper

Ellen L. Fry, Taylor J. Sharpe, Madhur Atreya, Gregory L. Whiting, John N. Quinton (2026). Printed PHBV-Based Sensors as a Real-Time Proxy for Soil Microbial Decomposer Activity During Drought and Flood Recovery. European Journal of Soil Science 77: e70358. https://doi.org/10.1111/ejss.70358

University of Manchester; Lancaster University; University of Colorado Boulder. The article is open access under the Creative Commons Attribution License (CC BY).

Figures 1–4 reproduced from the article under the Creative Commons Attribution License; some figures cropped; data curves redrawn from digitized figure values. All interactive microscopes, the bench, the parallel-strip and normalization demos, the CO₂ card and the field map are illustrative models, not data.

The authors’ public analysis notebook, named in the paper’s Data Availability statement: https://rpubs.com/TaylorSharpe/PolyDecomp