Using farm experience to improve N management for wheat (LearN)


Cereals & Oilseeds
Project code:
01 September 2013 - 31 March 2017
AHDB Cereals & Oilseeds.
AHDB sector cost:
Project leader:
D.R. Kindred1 , S.M. Clarke2 , R. Sylvester-Bradley1 , D. Hatley2 , S. Roques1 , N. Morris3 , S. Knight4 , D. Langton5 , M.A Blake-Kalff6 1 ADAS Boxworth, Boxworth, Cambridge, CB23 4NN 2 ADAS Gleadthorpe, Meden Vale, Mansfield, NG20 9PD 3 NIAB, Morley Business Centre, Morley, Wymondham, Norfolk, NR18 9DF 4 NIAB, Huntingdon Road, Cambridge, CB3 0LE 5 Agrii, Throws Farm, Stebbing, Gt. Dunmow, Essex, CM6 3AQ 6 Hill Court Farm Research Ltd., Frogsmarsh, Corse Lawn, Gloucestershire, GL19 4PW



About this project


Nitrogen (N) requirements of wheat are notoriously variable and difficult to predict, despite much research to develop recommendation systems. Recent studies also show a significant farm effect on grain protein content. This suggests a farm factor may be causing N requirements to differ, even where the RB209 recommendation would be the same. Some fields and farms may be consistently receiving too much or too little N fertiliser and, therefore, losing substantial profit from wasted fertiliser or lost yield, as well as potentially having an impact on the environment. It is also possible that yields on some farms are substantially constrained by N.

Precision farming technologies can be used to test N rates, through tramline comparisons and analysis of yield maps. The LearN project supported 18 self-selecting, engaged farmers to test whether their standard N rates were about right, too high or too low. They used simple trials on their farms, from 2014 to 2017, to test single replicates of 60kg/ha more and less than their standard rate of fertiliser N in alternate tramlines. On a subset of six heavy soil, high-yielding, long-term arable farms in East Anglia, conventional small-plot N response experiments were conducted on three fields per year, alongside the tramline trials.

The small plot trials showed that despite substantial and consistent differences in yields, proteins and soil N between farms, there was no evidence of consistent differences in N optima between farms. N optima varied as much within farms as between farms (overall range: 88 to 356kg/ha; mean 225kg/ha) and were well predicted by RB209 N recommendations with yield adjustment to 11t/ha. Little relation between soil mineral N or grain yield and N optimum was observed.

LearN farmers showed strong engagement; in principle tramline trials were easy to set-up, manage and harvest. However, many challenges had to be overcome, including: ensuring comparable treatment areas; precise recording of tramline wheelings and treatment boundaries; over-lapping fertiliser application by spinning disc spreaders; a need for farm-specific harvest protocols; transfer, processing and cleaning of yield data; statistical analysis to achieve robust conclusions in the face of residual spatial variation; and appropriate interpretation of multi-site results.

Overall, tramline yields were 11.43, 11.07 and 11.74t/ha for farm-standard, -60kgN/ha and +60kgN/ha respectively, giving differences in margin of -£9.81/ha and -£0.55/ha. Differences in yield between two ‘farm standard’ tramlines were used to indicate errors and infer confidence levels. At £140/t for grain and £0.70/kg for N, a difference of 0.3t/ha was required to pay for 60kg/ha difference in N fertiliser.

Firm conclusions on N management could be made on around half of the 142 tramline experiments. Underlying spatial variation in yield was usually much greater than the N effect. Overall, around 30% of standard N rates were found to be optimal, 26% were super-optimal and 24% were sub-optimal. No farms were found to be consistently applying too much or too little N. Variation in yield was large within and between fields and farms, but N was not the major driver. The causes for the variation in yield and profitability are a major question for the industry, for which, we believe, this approach to scientifically sound farmer-centric agronomic research, with simple field trials and robust statistics, gives good chances of answering.