# Precision agriculture

Why within-field variability is the premise of precision agriculture, how RTK positioning turns a map into an action — and why a management zone has to be stable over years before it is worth mapping at all.

The field is not uniform, and the machine applies inputs as if it were. Everything else follows from closing that gap.

Source: https://en.bioecon.ru/docs/agri-food/agri-tech-precision/precision-agriculture-agritech/
Updated: 2026-09-06



Precision agriculture rests on one observation: a field is not a unit. Yield, soil texture, organic matter, water-holding capacity and pest pressure vary across metres, while the seeder, the sprayer and the fertiliser spreader have historically applied the same rate across the whole of it. The average rate is right almost nowhere — over-applied on the weak ground that cannot use it, under-applied where the crop could have paid for more. The value on offer is not a new input; it is matching an existing input to a spatial pattern that was always there.

## Positioning is not agronomy

The enabling technology is satellite positioning corrected in real time. A bare GNSS receiver is accurate to metres; RTK correction, which compares the rover's carrier-phase measurement against a base station of known coordinates, resolves the ambiguity down to a couple of centimetres, and repeats that year on year so a pass in the third season can follow the wheel track of the first. This is what makes controlled-traffic farming, auto-steer and inter-row mechanical weeding possible at all.

But centimetres of positioning accuracy do not imply centimetres of agronomic response. The machine can be told exactly where it is; whether nitrogen applied at that point produces a different result from nitrogen applied five metres away is a separate, biological question. Roots forage laterally, water moves downslope, nutrients diffuse. The scale at which a crop actually responds is coarser than the scale at which the equipment can act, and confusing the two is the field's most common error.

## The stability problem

A variable-rate prescription assumes a management zone: a patch of field that behaves consistently enough to deserve its own rate. The zone must be stable to be worth mapping. Some drivers of variability are stable — soil texture, depth to a restricting layer, topography and the drainage that follows from it. Others are not: a wet year rewards the light ground and punishes the hollows, a dry year inverts the ranking exactly. Multi-year yield maps from the same field routinely show the pattern re-sorting between seasons, which is why zones built from stable physical properties — electrical conductivity surveys, elevation-derived wetness indices, soil sampling — tend to outperform zones built from a single season's yield map.

This is also why economic results in the literature are heterogeneous rather than uniformly positive. Where variability is large and stable, matching rate to zone saves input or lifts yield. Where the field is fairly uniform, or where variation is weather-driven and reshuffles annually, the prescription is noise dressed as precision, and the equipment cost is not recovered.

## What the rest of the cluster inherits

Every page here is a specialisation of the same logic: sensing produces a proxy, the proxy is turned into a spatial or temporal decision, and the honest question is whether the decision beats the uniform default. [Drone and satellite analytics](../drones-satellite-agro-analytics/) supply the observation layer, [autonomous machinery](../agricultural-robotics-autonomous-tractors/) the actuation layer, and [soil carbon MRV](../soil-carbon-mrv/) shows what happens when the quantity to be measured changes more slowly than the measurement error.