Computer-generated maps can mislead
Computer-generated maps can mislead
BEWARE of pretty computer-generated maps which claim to accurately represent soil nutrient status. They may not be what they seem and could waste money.
Too often, such maps rely on filling in missing information between widely-spaced sampling points, which can lead to big errors, ADAS crop consultant David Parish claims.
"Nutrient mapping usually involves dividing the field up on a 100m grid and taking a soil sample at each point. Then a computer generates a map showing the changes in nutrient status across a field. "
Basing the map on one reading a hectare is unsatisfactory. It is not an average result of that hectare, nor even a reliable one of the point, says Mr Parish.
"Some recent research has suggested you need at least nine sub samples and preferably 16 over a 4x4m grid to be confident of the reliability of a single point result." However, point values often take five samples over 5sq m.
Predicting the change between points is also open to error – research in the 1970s suggests half the variation in nutrient status in a field can occur within 1sq m. Furthermore, computer-generated prediction of change is based on a statistical technique untested for soil nutrient purposes. Adjusting the weighting of the computer fit can produce dramatically different results, adds Mr Parish.
He suggests farmers should apply their farm knowledge to enhance the reliability of maps.
He questions the economics of nutrient mapping, particularly for straw crops.
"There is only likely to be a yield response to applied fertiliser if the soil index is 0 or the lower part of index 1."
• Mark Glyde, of Hants-based SOYL, maintains that the 100m grid is sufficiently accurate for a given soil type. "Its a lot more than farmers were doing before.
"Ideally we would like to sample more frequently, perhaps every 50m. But the cost works out at a prohibitive £30/ha."
However, sampling agents do ask farmers where changes in soil type occur, and for details of problem areas, says Mr Glyde. Targeted intensive sampling may then be carried out.
SOYL collects 16 sub-samples over a 10m area to make one representative one (see story opposite). Accuracy is maintained between the points, he adds. He disagrees that soil nutrient status changes as fast as Mr Parish suggests.
"In all tests, we have never seen drastic changes, except between soil types." SOYL uses a technique which, in tests, has given an accurate representation of nutrient status changes, he maintains.
Economically, he points out that Mr Parish bases his argument on correcting low soil nutrient values. But that ignores the fact that savings can be made where P and K indices are high.
Acomputer is used to predict change in nutrient values between sampling points. But adjusting the computer weighting (above) produces dramatically different results, says ADASs David Parish.