Managing Dissolved Oxygen with Two Sensors in a Closed-Loop Drip Greenhouse

How DO behaves through a drip system

Before placing sensors or writing control logic, it helps to understand where DO goes in a typical system. Water leaves the lagoon injector at the highest DO it will ever have. From that point, it only loses oxygen.

The losses accumulate across four distinct stages:

Lagoon to mix tank — pipe residence time at typical greenhouse flow rates (100–150 L/min through 50mm pipework) is 3–6 minutes for a 25m run. Microbial respiration and chemical oxygen demand (COD and BOD) in irrigation channels and piping are the dominant consumption mechanisms at this stage [1]. Add a tank transit loss of ~0.2 mg/L as the water enters the mix tank and is agitated by the return pump.

Mix tank to header tank — similar pipe loss plus another tank transit. The mix tank can be a significant DO sink if fertiliser chemistry is active (acid dosing, CO₂ interaction with bicarbonates) or if the water is being heated. Combined loss here is typically 0.4–0.7 mg/L. One commercial operation reported supersaturating to 10 mg/L at the reservoir specifically to ensure that 5 mg/L or above would be maintained after nutrients were added and the solution trickled through the greenhouse [2].

Header to furthest dripper — this is the longest and most variable segment. Research on DO dynamics in drip irrigation capillary pipes found that DO loss in aerated water was greater in the last 10 metres of a run than in the first 10 metres, indicating accelerating decay as residence time extends [3]. The same study found that a 2.2 L/h labyrinth emitter caused a 33.2% reduction in DO concentration in aerated water passing through the system [3]. Emitter turbulence at delivery adds a further step loss as pressure drops cause bubble coalescence.

Total decay budget — for a well-maintained system at 20°C with a 135m total pipe run, expect 1.5–2.0 mg/L loss from injector to far dripper. At 24°C this rises due to both higher decay rates (metabolic rate roughly doubles per 10°C increase, Q₁₀ ≈ 2) and a lower saturation ceiling. Long-distance pipeline conveyance isolates water from the atmosphere for extended periods, which reduces DO concentration — a problem documented explicitly in drip systems serving crops over long runs [4]. In black polyethylene pipework exposed to sunlight, temperature-driven DO loss is compounded by solar heat gain in the pipe walls.

This means that to guarantee 6.0 mg/L at the furthest dripper, you need to inject at roughly 8.5–9.0 mg/L at the lagoon. DO above 8 mg/L is generally considered good for greenhouse production, and it is a common problem for DO levels in irrigation feed to fall to hypoxic levels (below 4 mg/L) [1]. Summer conditions frequently push nutrient solution temperatures above 22°C — exactly when plant demand is highest and the saturation ceiling is falling [5].

Oxygenation method and its effect on the decay curve

The two-sensor architecture described here works regardless of what is doing the oxygenation — nanobubble generators, venturi injectors, pure O₂ dissolution, hydrogen peroxide dosing, or any combination. The control logic is the same. What differs between methods is the achievable injection ceiling and how quickly DO decays in the pipe after injection.

Conventional aeration and venturi injection create bubbles in the 50–1000 µm range. These rise to the surface and escape within seconds to minutes. Under pipe pressure they coalesce readily and outgas at every pressure drop — bends, valves, emitters. The wider use of aerated water for irrigation has historically been limited by dis-uniformity in the field and the limited longevity of oxygen in the water [6]. For long pipe runs, conventional aeration struggles to maintain adequate DO at the far end without a mid-system second injection point.

Nanobubble systems produce bubbles predominantly under 200nm. At this scale they resist coalescence and remain in suspension far longer, reducing the effective decay rate constant through the pipe run. Research has shown that nanobubbles prolong the duration of DO in water compared to conventional injection, making them better suited to long distribution runs.

Hydrogen peroxide (H₂O₂) dosing works differently — it delivers oxygen through chemical decomposition at the root zone rather than as dissolved gas in transit. At low concentrations (typically 10–30 mg/L) it releases oxygen as it breaks down in contact with organic matter and root surfaces, providing a localised oxygen source rather than a bulk DO increase in the water column. It is less effective as a bulk pipe-DO management tool and is better understood as a rhizosphere hygiene and supplemental oxygenation treatment.

Pure O₂ injection via pressurised dissolution (Venturi or cone contactors) achieves the highest bulk DO — routinely 20–40 mg/L — but the elevated DO is unstable in open or low-pressure systems and outgasses quickly. It is most effective in pressurised closed loops or immediately before drippers.

In practice, the choice of method determines two parameters the control system needs to know: the maximum achievable DO at the injection point, and the expected decay rate constant through the pipe. Both feed into the setpoint ceiling and the decay budget calculation. The sensor logic itself does not change.

The two-sensor architecture

Sensor 1: lagoon outlet (fast control loop)

Position this sensor immediately downstream of the oxygenation unit, before the first pump or significant pipe run. Its job is purely mechanical: it drives the injector on/off and reacts within seconds.

The on/off thresholds must be set with a hysteresis band wide enough to prevent short-cycling. For a 100m+ system, 0.6–0.8 mg/L hysteresis is appropriate. The reasoning: at typical flow rates, the transport time from injector to furthest dripper is 15–20 minutes. A narrow hysteresis band causes the injector to cycle faster than the water can travel, so the sensor sees the effect of previous cycles before the plant does, creating oscillation.

Injector ON  when lagoon DO < lagoon_setpoint
Injector OFF when lagoon DO > lagoon_setpoint + 0.7 mg/L

Injector ON  when lagoon DO < lagoon_setpoint
Injector OFF when lagoon DO > lagoon_setpoint + 0.7 mg/L

Injector ON  when lagoon DO < lagoon_setpoint
Injector OFF when lagoon DO > lagoon_setpoint + 0.7 mg/L

This loop runs on a short cycle — every 30–60 seconds. It does not touch the setpoint. It only decides whether to inject or not based on the current setpoint value.

Sensor 2: drippers (slow integrating loop)

Position this sensor at a representative dripper — not the nearest one (which will always read high) and not the absolute furthest if that is an outlier. A dripper at roughly 70–80% of the maximum pipe run gives a representative reading of what most of the crop is receiving.

This sensor runs on a 10-minute cycle and uses a rolling average of its readings, not instantaneous values. Instantaneous dripper DO is noisy — it varies with irrigation pulse timing, zone valve sequencing, and sensor response lag. A 10-minute rolling average smooths this to a stable signal.

The slow loop adjusts the lagoon setpoint based on the difference between the dripper average and the dripper target:

dripper_error = dripper_target - dripper_do_10min_average
lagoon_setpoint += dripper_error × 0.4
lagoon_setpoint = clamp(lagoon_setpoint, min_setpoint, sat × 0.90)

dripper_error = dripper_target - dripper_do_10min_average
lagoon_setpoint += dripper_error × 0.4
lagoon_setpoint = clamp(lagoon_setpoint, min_setpoint, sat × 0.90)

dripper_error = dripper_target - dripper_do_10min_average
lagoon_setpoint += dripper_error × 0.4
lagoon_setpoint = clamp(lagoon_setpoint, min_setpoint, sat × 0.90)

The 0.4 gain means a 0.5 mg/L dripper deficit moves the setpoint by 0.2 mg/L per cycle — reaching full compensation in roughly 30–40 minutes. This is intentionally slow. It tracks a warming afternoon or a season change; it does not react to a single noisy reading or a short irrigation pause.

The upper clamp is not a physical ceiling — water can be supersaturated well beyond air-equilibrium saturation, and systems using pure O₂ injection or nanobubble generators routinely achieve 15–25 mg/L. The concern with very high DO in plants is not embolism (a risk specific to fish with closed vascular systems) but diminishing agronomic returns.

The practical upper clamp for the lagoon setpoint is therefore set by two things: the oxygenation unit's rated output, and the point above which further injection provides no measurable return at the dripper. For most greenhouse operations with a 100m+ run, 12–14 mg/L at the lagoon outlet is a reasonable ceiling — high enough to cover the decay budget while remaining within the agronomically useful range at the root zone. The lower clamp should be set above the minimum meaningful injection level — typically 7.5 mg/L — below which there is no longer enough headroom to cover the decay chain.

What the setpoint tells you over time

The lagoon setpoint is not just a control variable — it is a diagnostic signal. Log it continuously alongside water temperature.

On timescales of hours, setpoint variation tracks diurnal temperature. As the greenhouse warms through the morning, the saturation ceiling drops and decay accelerates; the slow loop walks the setpoint up. This is normal and expected. Maintaining nutrient solution temperature between 18–20°C provides a strong foundation for oxygen availability — insulating tanks and managing heat exchange systems are practical starting points [5].

On timescales of weeks, a setpoint that is creeping upward faster than temperature alone explains is almost always biofilm. Biofilm in pipes depletes DO by increasing biological oxygen demand — both the biofilm matrix itself and the organic matter it harbours consume oxygen continuously [1, 7]. Biofilm and organic matter in the water are identified as the main drains on oxygen in the irrigation system, and controlled DO levels can only be reliably delivered to plants when biofilm is removed [7]. Accumulation is gradual — typically over several weeks in systems with recirculated drain water — and the slow loop compensates for it by raising the setpoint. When the setpoint approaches the injector's practical ceiling and the dripper sensor is still below target, the single-injector architecture has hit its limit and a pipe flush is overdue.

The critical diagnostic: setpoint at ceiling, dripper still low

The most important alert the two-sensor system can generate is this:

Lagoon setpoint has reached the oxygenation unit's practical ceiling but dripper DO remains below target.

This state is unambiguous: the injector is running as hard as physically possible and it is not enough. A single sensor at either point alone would simply show "low DO" with no indication of cause. Together, they tell you:

• It is not a dosing fault (the injector is maxed out)

• It is not a setpoint error (the slow loop has already compensated as far as it can)

• The decay between injection and delivery exceeds what one injector can cover

The causes, in order of likelihood:

1. Biofilm — biofilm accumulation in irrigation infrastructure creates conditions for pathogen pressure, nutrient inconsistency, and ongoing DO depletion [8]; schedule a flush and recheck after 24 hours

2. Temperature spike — as solution temperature rises above 22–23°C, the air-equilibrium saturation ceiling drops and decay rates accelerate; any oxygenation system operating near its output limit will struggle to compensate, and decay acceleration erodes the budget faster than expected [1]

3. Flow rate reduction — a blocked filter or partially closed valve increases pipe residence time and therefore decay; check pump pressure and filter differential

4. Oxygenation unit output drop — membranes foul, peroxide stocks run low, venturi air lines block; verify that the unit is delivering its rated DO and inspect or replenish as needed per manufacturer schedule

5. Architecture limit — if the above are resolved and the problem persists, the pipe run is genuinely too long for a single injection point at these operating temperatures; a second injector mid-system is required

Sensor placement practicalities

Lagoon outlet sensor — mount in the outlet pipe, ideally in a flow-through cell rather than a T-fitting. The sensor must be in moving water; a stagnant pocket gives artificially stable readings that lag actual DO by minutes. Optical (luminescent) sensors have fast response and require less maintenance, check our DO sensors here. Calibrate per instructions.

Dripper sensor — this is the harder placement. Options in descending preference:

A dedicated sample dripper running into a small flow-through cell gives a continuous reading but requires a return line back to the drain. This is the most accurate approach and worth the plumbing for a research or high-value operation.

Not recommended: a manual spot-check with a handheld DO meter at multiple drippers, logged daily, can substitute for a fixed sensor in smaller operations. The slow loop setpoint adjustment would then be done manually rather than automatically — but the diagnostic logic is identical.

Temperature co-location — both sensors should log temperature alongside DO. Our DO sensor's already have temp calibration built-in. Without temperature, a raw mg/L reading is ambiguous: 8.0 mg/L at 15°C is 88% saturation; 8.0 mg/L at 25°C is near-maximum. The saturation percentage is more meaningful than the absolute value for setting alarms.

Interaction with irrigation scheduling

One aspect of DO management that most growers overlook: idle pipe time.

When a zone valve closes, the water in the dripper lines from header to emitter stops moving. DO decay continues in that standing water for the entire idle period. When the zone opens again, the first irrigation pulse delivers this stale, low-DO water directly to roots before the fresh, enriched water from the header arrives.

Research on recirculating NFT cultivation trenches documented a clear DO gradient along the channel — while concentration near the inlet was adequate (6.2 mg/L), it dropped to values critical for cucumber at the last plant position downstream (2.9 mg/L) [9]. The same gradient effect applies in drip distribution lines: the furthest emitters from a zone valve receive the oldest, most depleted water in every irrigation cycle.

Practical mitigations, in order of simplicity:

• Short pre-flush pulse — open the zone for 30–60 seconds before the main irrigation event to push stale water through before the full dose. The volume is small relative to total irrigation but the DO benefit at roots is significant.

• Increase irrigation frequency — more frequent, shorter pulses mean less idle time and less stagnation in dripper lines. This also tends to improve substrate moisture uniformity.

• Zone sequencing — in multi-zone systems, stagger valve closing times so no zone sits idle for more than 20 minutes during active irrigation periods.

Summary

Two DO sensors in a closed-loop drip system do fundamentally different jobs. The lagoon sensor is a fast actuator — it fires the injector within seconds of a drop. The dripper sensor is a slow integrator — it adjusts what "enough" means at the lagoon over a timescale of hours to weeks, compensating for temperature, biofilm, and seasonal change.

Neither sensor alone is sufficient. The lagoon sensor without the dripper has no way to know whether its setpoint is correct for current conditions. The dripper sensor alone, controlling injection directly, fights a 17-minute transport delay and oscillates.

Together, they form a control system where the fast loop is never confused by slow drift, and the slow loop is never overwhelmed by fast disturbances. The setpoint that emerges from this system — logged over time — becomes one of the most informative records in the greenhouse: a combined history of temperature, biofilm, and system health that no single measurement can provide.