Analysis of yield response variability to nitrogen fertilization in experiments performed in the Argentine Pampas

Abstract

Nitrogen (N) fertilization has become a common practice in corn and wheat crops in the Argentine Pampas during the past decade. In this region, great environmental variability determines erratic responses to fertilization. The quantity of data necessary for defining yield response models to N has not been investigated, and the relative yield transformation, combined with the total nutrient approach, has been widespread when analyzing fertilizer response results. The objectives were to determine the minimum data set necessary for fitting average yield functions suitable for fertilizer recommendation at regional scale and to investigate the consequences of using relative yield on N response functions when the total nutrient approach is used. Published results from two extensive fertilization networks, one with corn and the other with wheat, were used. Data were aggregated at different levels, because one single experiment to the entire network results, and yield response functions to N were fitted. Yield models tended to stability when a set of around 100 or more data points, generated in experiments performed across different sites and years, were used for fitting models with both crops. This amount of data was generated by performing 20 experiments in the corn network and 35 in the wheat network. Relative yield transformation allowed us to obtain models with lower dispersion than yield, but in the case of corn a biased model was generated that leads to underestimating fertilizer requirements. In wheat, similar fertilizer recommendations were produced from yield or relative yield functions. Response variability to fertilization must be addressed in the experimental area by increasing the amount of data used, rather than by applying the relative yield transformation.