Abstract
The use of in-situ sensors capable of real-time monitoring of soil nitrogen (N) may facilitate improvements in agricultural N-use efficiency (NUE) through better fertiliser management. The optimal design of such sensor networks, consisting of clusters of sensors each attached to a data logger, depends upon the spatial variation of soil N and the relative cost of the data loggers and sensors. The primary objective of this study was to demonstrate how in-situ networks of N sensors could be optimally designed to enable the cost-efficient monitoring of soil N within a grassland field (1.9ha). In the summer of 2014, two nested sampling campaigns (June & July) were undertaken to assess spatial variation in soil amino acids, ammonium (NH4+) and nitrate (NO3−) at a range of scales that represented the within (less than 2m) and between (greater than 2m) data logger/sensor cluster variability. Variance at short range (less than 2m) was found to be dominant for all N forms. Variation at larger scales (greater than 2m) was not as large but was still considered an important spatial component for all N forms, especially NO3−. The variance components derived from the nested sampling were used to inform the efficient design of theoretical in-situ networks of NH4+ and NO3− sensors based on the costs of a commercially available data logger and ion-selective electrodes (ISEs). Based on the spatial variance observed in the June nested sampling, and given a budget of £5000, the NO3− field mean could be estimated with a 95% confidence interval width of 1.70μgNg−1 using 2 randomly positioned data loggers each with 5 sensors. Further investigation into “aggregate-scale” (less than 1cm) spatial variance revealed further large variation at the sub 1-cm scale for all N forms. Sensors, for which the measurement represents an integration over a sensor-soil contact area of diameter less than 1cm, would be subject to this aggregate-scale variability. As such, local replication at scales less than 1cm would be needed to maintain the precision of the resulting field mean estimation. Adoption of in-situ sensor networks will depend upon the development of suitable low‐cost sensors, demonstration of the cost-benefit and the construction of a decision support system that utilises the generated data to improve the NUE of fertiliser N management.
