Statistical modeling for on-farm experimentation using precision agricultural technology

CÓRDOBA, M.; BALZARINI, M.; PACCIORETTI, P.; BULLOCK, D.; BRUNO, C.

Conferencia; 12th European Conference on Precision Agriculture; 2019

On-farm, large-scale agronomic field trials were conducted using precision technology to facilitate trial implementation by enabling the automation of treatment assignments. Automation of changing input rates and monitoring associated plot yields provided data for the estimation of yield responses as functions of input treatments and field characteristics, which are useful for developing environmentally and economically optimal crop management prescriptions. The analysis of this type of on-farm agronomic trial data is powered by new developments of methods to estimate spatially restricted predictive models. The spatial variability within the field, which is usually linked to the variability of soil properties and topography, should be taken into account when comparing input rates under different conditions. In these experiments, we fit yield response functions with different nitrogen fertilization (N) and seeding (S) rates, applied with spatial precision during sowing. The performances of several statistical and machine learning methods, used to fit predictive models, were analyzed, using site characteristic variables in the predictor, with and without accounting for spatial autocorrelation.