Economic optimisation of soil testing on commercial farms: space, time and variables

A holistic evaluation of agricultural systems requires mechanistic understanding of physical, chemical and biological interactions both aboveground and belowground, yet obtaining this information on commercial farms is a challenging task. In order to support practical decision making by commercial producers, it is therefore necessary to identify system-wide performance indicators that are observable presently and cost-effectively. Data acquired through commercial soil testing satisfy these conditions; however, the relationship between the density of information — thus the cost of testing — and the value of information as a guideline for on-farm managerial changes is not well-understood.

Using high-resolution soil data from the North Wyke Farm Platform in the UK as a case exemplar, this solicited talk discusses theoretical and computational frameworks to quantify the value of an information package defined by soil testing strategies. A bootstrapping experiment revealed that the information value is often a concave function of the spatial sampling frequency, indicating that “half-hearted” soil data are unlikely to be able to inform optimal farm management. On the other hand, a high degree of serial correlation as well as atemporal inter-variable correlation resulted in some measurements identified as being redundant, as the incremental value of additional information was often found to be small and occasionally negative. Given the time and budgetary constraints, therefore, it is suggested that more effort should be spent on snapshot spatial sampling of a small number of variables, rather than continuous spot sampling of a large number of variables.