Diminishing fertiliser returns
A Socratic walk-through of diminishing fertiliser returns — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why does doubling the fertiliser on a field add almost nothing to the harvest it yields?
The natural picture is that a crop eats nitrogen roughly as we eat dinner: more food, more growth, until appetite is satisfied and the rest is refused. That picture predicts a straight rise followed by a flat top — and a farmer who doubles the dose while still short of the ceiling should get roughly double the extra grain.
Fields do not behave that way. The curve bends almost from the first kilogram, and at a commercially normal rate the second hundred kilograms buys a small fraction of what the first hundred bought. Something is wrong with the appetite model. What?
Reasoning it through
REASONING #Start by asking what the plant is actually short of. Nitrogen is one input among many: phosphorus, potassium, sulphur, water, rooting depth, sunlight intercepted by the canopy, freedom from disease. Justus von Liebig's insight in the 1840s was that the output is set by whichever of these is scarcest, not by their sum. Relieve nitrogen and you have not made the crop unconstrained — you have promoted something else to be the constraint.
And the constraint that eventually takes over is often the crop itself. A cereal fixes the number of grain sites it can fill weeks before harvest, when it sets tillers and florets. Nitrogen arriving after that decision can make a bigger, greener plant but has nowhere to put grain. The limit stops being supply and becomes sink capacity.
But notice that Liebig's account predicts the wrong shape. A single binding constraint gives a sharp kink — a straight climb, then an abrupt plateau. Real fields give a smooth bend. Why? Because a field is not one barrel. Every square metre has its own binding constraint at its own level, and the constraint shifts through the season with rainfall and root growth. Average many small kinked curves whose kinks sit at different places and the aggregate is a curve. That smoothed version is Mitscherlich's, from around 1909, and it is the shape agronomists actually fit.
Now the half of the answer the appetite model misses entirely. Ask where the unused nitrogen goes. Food left on a plate stays on the plate; nitrogen does not. Nitrate carries a negative charge, so it is not held by the negatively charged surfaces of clay and organic matter, and it travels with drainage water. In wet, airless soil, bacteria strip its oxygen and return it to the air as nitrogen gas and nitrous oxide. Urea or ammonium on a warm surface loses nitrogen as ammonia. Across cereals, only something like a third to a half of applied nitrogen typically ends up in the crop in the season it is applied — and that fraction falls as the rate rises, because the surplus grows while uptake does not.
That is the same curve viewed from the input side: the marginal kilogram is not banked for later, it leaves the field. And past a point the response turns negative — luxury uptake builds leaf and stem, the canopy shades itself, the crop lodges in wind, and dense humid foliage invites disease.
The analogy
THE ANALOGY #Think of widening one road on a congested commute. The first widening helps, because that road was the bottleneck. Widen it again and the queue simply re-forms at the roundabout beyond it; widen that, and the car park at the far end becomes the limit. Nothing about the extra tarmac was wasted in principle — it just stopped being the thing standing in the way, and the traffic you added past that point drains off down side streets and never reaches the destination at all.
road bottlenecks move one at a time, in sequence, which would give the sharp kinked curve. A field is thousands of separate small commutes with their bottlenecks in different places, and that heterogeneity — not any single limit — is what smooths the corner into a bend.
Clarifying the model
THE MODEL #The correction is not "plants can only absorb so much." Uptake capacity is rarely the binding thing. Two separate mechanisms are working at once: the crop's response saturates because something other than nitrogen is limiting it, and the fertiliser that the crop does not take is lost to water and air rather than banked. Either alone would bend the curve; together they bend it early.
Be honest about scope. "Doubling adds almost nothing" is a statement about a field already fertilised near its optimum. On a genuinely nitrogen-starved soil, doubling a small dose can very nearly double the response — the left-hand part of the same curve is steep. The question is really about where on the curve conventional practice already sits.
Two things are genuinely argued over. Agronomists disagree about which functional form is right — quadratic, quadratic-plateau, linear-plateau, or Mitscherlich — and the choice materially changes the recommended rate, because the forms disagree most exactly where farmers operate. And the recovery fractions above are averages over enormous variation in soil, rainfall, timing and placement. The economically optimal rate, meanwhile, is never the yield-maximising rate: it is where the value of the extra grain equals the cost of the extra bag, which always sits to the left of the peak.
A picture of it
THE PICTURE #How to readThe line is cumulative — how much of the achievable response has been bought by the time you reach that rate; each bar is the extra yield delivered by that 25 kg step alone. Read the bars left to right: the first step buys about a quarter of the whole response, the eighth about a thirtieth. Compare the line at 100 and at 200 for the question itself — doubling the rate moves the response from roughly 70 to 91 percent of maximum, so the second hundred kilograms delivers under a third of what the first did. This is the Mitscherlich function plotted, not measured field data, and it shows only the fertiliser-derived portion; a real field yields something on zero nitrogen because the soil supplies its own.
What became clearer
WHAT CLEARED #Fertiliser does not fail because the crop is full. It fails because relieving one constraint promotes another — water, phosphorus, rooting depth, or the grain sites the plant committed to weeks earlier — and because nitrogen the crop does not take is not left waiting, but leaches, gases off, and goes. The bend in the curve is two things at once: a rising share of the dose meeting a limit that is no longer nitrogen, and a rising share of it never being taken up at all.
Where to go next
ONWARD #- Why splitting an application into several smaller doses raises recovery without raising the total.
- How nitrous oxide from surplus fertiliser makes over-application a climate question, not only an economic one.
Key terms
TERMS #| Term | What it means |
|---|---|
| Law of the minimum | Liebig's principle that yield is governed by the scarcest required resource, not by the total supplied. |
| Mitscherlich function | the smoothed diminishing-returns curve, Y = A(1 - e^(-cN)), fitted to nutrient response. |
| Sink capacity | the crop's own limit on where assimilate can be deposited, largely fixed by the number of grain sites set earlier in the season. |
| Recovery efficiency | the fraction of applied nutrient that ends up in the harvested crop in the season of application. |
Every term the collection defines is gathered in the glossary.