Duckweed Harvesting: The Regime That Decides Removal
Duckweed harvesting is the nutrient removal step, not pond maintenance. What the published evidence says about frequency, fraction, cost and biomass fate.
Most of what is written about duckweed harvesting is written by people trying to get rid of it. That framing survives into the technical literature as a habit, and it hides the single fact an operator needs: harvesting is not the cleanup after the treatment. It is the treatment.
Duckweed absorbs nitrogen and phosphorus out of the water and holds them in plant tissue. Nothing has left the pond at that point. The nutrients are still inside the system boundary, floating on it. They leave when the biomass leaves, and only then.
Nothing is removed until the biomass leaves the water
The clearest statement of this comes from a study of the municipal lagoons at Wellsville, Utah, where the duckweed was never cropped. The plants died off, decomposed at the bottom of the lagoon into benthic sludge and released much of the nutrient load back into the water, which the author notes greatly undermines the nutrient removal. The pond looked like a duckweed treatment system all summer. In accounting terms it was a holding tank with a lid on it.
Put a number on the step and it gets sharper still. Across duckweed-covered lagoons, the total nitrogen and phosphorus input recovered in the harvested biomass amounts to 50 plus or minus 20 percent. Harvest is the only mechanism in that figure. Everything else is stripping, denitrification and burial, and burial is a loan rather than a removal.
Even the extension pages that treat the plant as an infestation get the mechanism right by accident. Penn State recommends netting or raking the plants off and disposing of them away from the pond, and notes this is especially attractive because it also removes the nutrients associated with the plant material. Correct physics, opposite intent. The pond owner is trying to end the crop. The operator is trying to run it.
The regime is two numbers: how often, and what fraction
A harvest regime is not a chore list. Modelling work on large-scale duckweed production formalises it as three parameters: harvest frequency, harvest ratio and harvest threshold, fitted with initial mat density and intrinsic growth rate into a regression at R-squared 0.95. Optimised for a site in southwest Florida, the highest annual yield of 70 dry tonnes per hectare came out at a harvest ratio of 0.35 and a harvest frequency of one day.
Daily. Removing about a third. That is the theoretical shape of the answer, and it is worth knowing before anyone quotes you a fortnightly schedule.
Little and often beats a lot and rarely
The comparison has been run directly. Four regimes on swine wastewater, 20 percent twice per week, 40 percent once per week, 60 percent every two weeks and 80 percent every four weeks, and the first won on every axis: highest specific growth rate, 20.0 g fresh biomass per m2 per day, and removal of 83.7 percent of the nitrogen and 89.4 percent of the phosphorus. Harvesting less biomass at a shorter interval favoured both recovery and production.
Flow-through reactor work explains why the long intervals lose. Productivity peaked at day 5 and was reduced to approximately 25 percent of maximum productivity by day 8, which is the crowding penalty arriving. Stretching a schedule from five days to eight does not buy you a bigger harvest. It buys you three days of a population that has stopped working.
There is a floor under how hard you can crop
The other direction has a limit too, and it is not far away. In an aquaponic trial comparing no harvest against 25 and 50 percent of surface area weekly, the unharvested control yielded 17.9 t/ha/yr against 23.4 to 24 t/ha/yr for both harvested treatments, with daily growth rates of 138 percent for the control against 159 and 161 percent. Two things in that result matter. Harvesting increased total production rather than reducing it, which is the counterintuitive part. And 25 and 50 percent were statistically indistinguishable, while the same group’s earlier work on Spirodela found a 75 percent ratio counterproductive because the crop could not replenish itself in a week.
So the usable band is wide in the middle and hard at the edges. Somewhere between a fifth and a half, often. Not four fifths, rarely.
The interval also changes what you harvest
Regime is a composition setting as well as a yield setting. In the reviewed literature, a 20-day harvesting interval yielded the highest protein content at 39.50 percent, against 32.77 to 35.91 percent at 5 to 15 day intervals, while twice-weekly harvesting gave 533 g per m2 fresh weight against 402 for once weekly. More frequent harvesting buys tonnage. Longer intervals buy concentration. You cannot optimise both with one schedule.
One caveat carries more weight than the trade-off itself: composition targets belong to production-grade cultivation. On a remediation pond the only output specification that matters is what left the water, and we return to why below.
What you leave behind is a setting, not a leftover
Every regime above is defined against a residual, and the residual has two jobs: keep the surface shaded so algae cannot establish, and keep the population in the part of its curve where it still grows fast. The standing guidance is that an almost complete cover should remain on the pond surface after plant harvesting, not a cleared pond with a seed population in it.
An outdoor pilot on a treatment lagoon in the Burdekin put concrete numbers under that. A base density of 1 kg per m2 wet equated visually to complete coverage by a single layer of fronds and yielded on average 175 g per m2 per day, with biomass doubling every 5.7 days. The same report says plainly that daily harvesting of incremental growth is theoretically ideal but impractical on labour and cost, and that weekly scheduled harvests were the appropriate compromise. That gap between the model optimum and the staffed reality is the honest centre of this whole topic.
The most instructive failure is quieter. In a Belgian pilot running three outdoor systems on pig manure derived streams for a full season, the mat was deliberately never completely removed so that surface coverage suppressed algae. Because the surface density before and after each harvest was not measured, relative growth rate could not be calculated at all, and the authors attribute part of their week to week variability to exactly that. An unmeasured residual makes a regime unrepeatable, which means the schedule that worked last month cannot be reproduced on purpose.
Methods, in the order operations reach for them
The machine question is a separate decision and we have treated it separately in duckweed harvesting equipment. What belongs here is the order of escalation and the two mechanics that decide whether any of it works.
Hand netting and skimming from the bank. Wind corralling into a downwind corner. Floating grids or booms. Gravity spillways into a draining pan. Skimmer, pump and screen trains. Self-propelled harvesters. Cost and labour rise along that list, and so does the minimum pond size that justifies each step.
The mechanics come from a study built specifically to test skimming. Duckweed does not present as a single layer: plants form clumps, which affects the skimming of the mat, and a skimmer depth of approximately 1.5 cm was the minimum at which clumps moved in consistently. More importantly, tracers placed on the mat and monitored for two weeks under natural wind and skimmer flow showed some movement but no consistent rotation. The mat does not deliver itself to the collection point. Anything that assumes it will is a plan with a hole in it.
That is also the argument for a floating grid rather than an open surface. Dividing the pond stops the crop shifting from one side to the other in wind, which otherwise produces under or over harvesting of certain portions of the pond even when the total harvested mass looks right.
What harvesting actually costs
Four drivers, in rough order of how often they surprise people.
Attendance. Duckweed treatment is not a low-touch asset. Operation and maintenance require a high input of skilled labour, with almost daily attention necessary to maintain optimum growth conditions and treatment efficiencies. The regime evidence pushes toward short intervals, and short intervals are staff.
Water. The crop is 92 percent water and loses up to half of that in the first two days out of the pond. Everything you lift before dewatering is freight you are paying for twice, which is why the Burdekin design let the crop stand in a collecting pan for a day or two before anything transported it.
Perishability. The harvest window closes fast. Storage at 25 C leads to rapid weight loss within 5 days, with 4 C and 10 C extending it only marginally. Drying is what converts a perishable wet mass into something with a shelf life, and the Wellsville biomass was air dried to 5 to 10 percent moisture before it could be stored or ground.
Season. Outdoors, the year is shorter than the calendar. In the Belgian pilot, growth ceased after approximately 100 to 120 days, likely due to rising pH and electrical conductivity, and the systems that did perform produced 8 tonnes of dry biomass per hectare per year at removal rates of 1.2 g N and 0.13 g P per m2 per day. Annualised capital against a four-month operating window is a different number than the brochure one. The rest of that arithmetic is what duckweed cultivation costs.
Where the harvested biomass may go, and where it may not
Harvesting concentrates whatever the water contained. That is the point when the target is nutrients, and it is the problem when the water carried anything else.
Where duckweed has taken up heavy metals or organic toxic compounds, the guidance is unambiguous in both directions: the biomass must be harvested at regular intervals or the toxins settle into the sediment with the decaying plants, and the harvested plants should then be burnt or disposed of in sealed landfill. Regulatory classification can be genuinely unsettled. In the Utah case the harvested biomass was not considered biosolids by the state coordinator, so the biosolids regulations did not apply, which is a local answer and not a general one. The same work is blunt about the feed question: know the source and composition of the wastewater before that option is implemented at all.
Much of the sector does not stop there. The Belgian pilot cited above was explicitly designed to treat manure-derived streams and produce feed biomass from them, and that dual purpose is the norm in this field.
SERAPH does not operate that way, and the policy is published rather than inferred. Biomass grown on wastewater, manure or contaminated water is never sold or blended as food or feed, and the two streams are separated from the first sensor reading. For a harvest regime that means the destination is fixed before the first frond is lifted, and it decides the schedule as much as growth rate does. A remediation pond is cropped to hit a discharge number. A production pond is cropped to hit a composition. The same equipment cannot serve both without becoming the contamination path, which is the practical half of the same rule and the reason a duckweed wastewater treatment train gets laid out the way it does.
You cannot run a regime you cannot see
Every number in this article is a fraction of a standing crop. A 0.35 harvest ratio, a residual of 1 kg per m2, a 25 percent weekly removal. All of them require knowing what is on the water immediately before the harvest, and a net and a scale after the fact tell you what you took, not what was there.
That gap is what SERAPH works on. Our vision system reconstructs a cultivation tray in 3D from a single capture, drawing 170,000 surface points from 5.9 million captured, computes canopy volume and distinguishes duckweed species in a live frame. Growth and stress reading are still in development.
The limit deserves saying as plainly as the capability. SERAPH is at TRL 3, proof of concept, with outdoor field validation still ahead. Not one measurement above is ours. What this article contains is the published record, arranged in the order the decisions actually arrive, and the underlying duckweed growth rate question sets the interval before any of it.
FAQ
How often should duckweed be harvested?
Shorter than most schedules assume. Model optimisation for a Florida site landed on daily harvesting at a 0.35 ratio, flow-through reactor work concluded 5-day intervals, and an outdoor lagoon pilot settled on weekly as a labour compromise rather than an optimum. The pattern across all of them is that small frequent harvests beat large infrequent ones.
How much of the mat should come off at each harvest?
A fraction, never the lot. Regimes removing 20 to 50 percent perform well and 25 versus 50 percent has been measured as statistically indistinguishable, while 75 to 80 percent is counterproductive because the population cannot refill in the interval. A near-complete cover should remain so algae stay shaded out.
Does harvesting actually remove nitrogen and phosphorus?
It is the only step that does. Uptake moves the nutrients into the plant, harvest moves them out of the water, and unharvested duckweed dies, sinks and returns most of its load. Reported recovery in harvested biomass sits around 50 plus or minus 20 percent of input.
How long does harvested duckweed keep?
Days. At 25 C it loses weight rapidly within 5 days, and refrigeration extends that only slightly, so the processing decision has to be made before the harvest rather than after it.
Can biomass harvested from a wastewater or manure pond be used as feed?
Not under SERAPH’s policy, which separates the remediation and production streams from the first sensor reading and never sells or blends remediation biomass as food or feed. Where the water carried metals or toxic compounds, the standard disposal guidance is incineration or sealed landfill.