Some tools for parsimonious modelling and interpretation of within-field variation of soil and crop systems

Parsimony is a guiding principle in scientific investigation which goes back to the medieval schools. It is proposed in this paper that, in different guises, the principle of parsimony (using no more complex a model or representation of reality than absolutely necessary) is essential for agricultural research at the scale of the field or larger regions. Two examples of the principle are given, illustrated with examples from research on precision agriculture:1. atio-temporal variability of a sequence of yield maps is considerable, and often defeats simple approaches to interpretation and analysis. Reducing the variability to a small number of basic temporal patterns with spatial expression allows useful information to be extracted. The normalised fuzzy partition entropy is a parsimonious criterion appropriate for identifying the number of distinct patterns in the data. It is shown in a case study how this can aid interpretation of a complex data set. of the joint effect of different factors on crop yield can take various forms. A simple assumption is that the different factors are additive, the most complex models describe interactions. Justus von Liebig’s ‘Law of the minimum’ is a model of little complexity and may often be a parsimonious and powerful tool for modelling crop responses. A comparison of these models on some crop response data, using the Akaike information criterion to measure the parsimony of the different models, is given.