Paper at the European Society of Agronomy in 2016:

The challenge facing agriculture of producing more whilst impacting less is real and immediate. For success it is crucial that researchers, farms and the supply chains engage effectively. Knowledge exchange must become a multi-way process including increased interaction between the researcher and the farmer. Thus far, any extrapolation between small (science) and large (industry) scales has entailed large untestable ‘leaps of faith’; the two communities have worked at different scales, with different concepts for the analysis of crop performance and different standards of proof. We contend that what is needed is a shared interest in the challenges and constraints faced in the field.
ADAS recently established an ‘Agron􀇀mics’ initiative to develop common concepts, metrics, targets and techniques that could enable joint analysis of crop performance – in terms of both productivity and sustainability – by farmers and researchers. Traditional crop research employs experimental designs that minimise effects of
uncontrolled environmental variables so that responses (e.g. in yield) to controllable inputs (seed, tillage, fertilisers, pesticides) can be tested. However, the area of these small plot trials commonly restricts their relevance to one site and, while their small scale may optimise internal precision and accuracy of the test, it limits wider relevance of the results.
We propose that, over and above the scientific challenges at lab scale, there are big opportunities for science in investigating the multiple variables and emergent properties affecting translation between small and large scales; not least amongst these are the interactions between agronomic innovations – new germplasm, chemistry or machinery – and soil. Research is needed to understand such interactions but current dependence on small-plot trials generally proves inadequate for this because ‘sites’, even if there are many, confound many factors with
soil, especially climate and farming system. However, technologies for on-farm automation (‘precision farming’) now provide opportunities for quantitative phenotyping at field and farm scales, and also (critically) they provide new understanding of spatially variable factors, particularly soil. 
We have identified five key challenges necessary to support an ‘Agron􀇀mics’ approach: (i) acceptance by farmers and scientists of common concepts for explanation of crop performance, such as ‘resource capture’; (ii) motivated and co-ordinated networks of farms that embrace regional and landscape dimensions; (ii) more precise and accurate farm machinery; (iii) new spatially-referenced statistical techniques for modelling and testing on-farm data at intra-field scale; and (v) facilitating software.