Speaker
Description
Farming is vital business. Agricultural experiments have long been carried out on crops including sugar, wheat, potatoes and grass. The second oldest grassland experiment in the UK has been in continuous action at Newcastle University’s Cockle Park farm in Northumberland since 1897. (The oldest dates from 1856 at Rothamsted Research Lab.)
Over the years data on grass (hay) yield, fertiliser treatments, soil structure, grass composition and the weather have been meticulously recorded in handwritten notes, spreadsheets and pdf files. Data was analysed in 1960 by Pawson and in 1980 by Coleman. Further analysis has been piecemeal and hampered by the disparate sources of data.
Using Artificial Intelligence (AI) to convert a photo of handwritten data or a pdf file into a comprehensive spreadsheet has been transformative. It has motivated work on collating this valuable data into a definitive resource available for analysis by soil scientists, climatologists and statisticians.
The full set of yield data from 1897 to 2025 has now been reanalysed replicating the basic analysis carried out in 1980 and extending it with newer methods of assessing correlations between plots and detecting carry-over effects and cycles in the yields. The wider project aim is to collate the data into a definitive, settled dataset that could be made available more widely via an open-source website.
The analytical results for hay yields are presented in this paper and clearly show dramatic changes over time, and relationships between the 14 experimental plots receiving different nutrient regimes. This is the upside of AI.
The downside of AI is that it does not solve the challenges in collating disparate data into a sound resource. Administrative decisions have to be made and recorded. AI is instrumental but still needs significant input from personnel with sound domain knowledge. Data is extremely valuable, the cost in time, land use and labour is enormous for 128 years’ worth of data. AI unleashes the opportunity to maximise the information from experiments and improve quality and efficiency in this agricultural business. We discuss these issues in the paper.
| Classification | Both methodology and application |
|---|---|
| Keywords | agricultural business, time series, treatment effects |