Does precision agriculture pay? Profitability potential of variable rate application of fertilizers: Econometrics and geostatistics approaches

Precision agriculture (PA) in crop production refers to the management approach in which inputs are applied according to the site-specific demand and timing in order to improve efficiency of inputs and farm profitability. This study studied the profitability potential of a variable rate application (VRA) management strategy for phosphorus and nitrogen fertilizer compared with the conventional uniform application (UMN) following the University of Minnesota Extension. The four objectives were to (1) show that there is both temporal and spatial variation in corn response to phosphorus and nitrogen fertilizer, (2) estimate the potential gain from optimally managing this temporal and spatial variation, (3) determine the degree to which unpredictable temporal variation diminishes the value of a VRA strategy, and (4) explore whether soil test and landscape information can be used to predict temporal and spatial variation. The results of the analysis of corn yield data obtained in years 1997, 1999, and 2000 generated in southern Minnesota showed that there were temporal and spatial variations in corn response to fertilizer inputs; the estimated site-specific optimal phosphorus rates varied between 0.0 and 112.3 kg ha-1, while the estimated site-specific optimal nitrogen rates varied between 0.0 and 202.2 kg ha-1. The potential gain from managing this variability was estimated to be {dollar}63.58ha -1, {dollar}64.34 ha-1, and {dollar}117.42 ha-1 in 1997, 1999, and 2001. These gains easily cover, {dollar}16.10 ha-1, the extra cost of switching from UMN to VRA. These gains were decomposed into two parts: the contribution by managing temporal variation only and the contribution by managing spatial variation only. Temporal variation in the site-specific crop response was estimated to reduce the potential gain from a VRA strategy by about a third. Landscape variables were better predictors of variation in crop response than soil test variables. The reconstructed site-specific crop response functions (SS-CRFs) using the soil test variables resulted in a deviation of size 3.73 +/- 0.85% from the original SS-CRFs in terms of estimated return on fertilizer inputs, where the reconstructed SS-CRFs using both the soil test and landscape variables resulted in a deviation of the size of 5.50 +/- 1.17%.