Assessing risk associated with N fertilizer recommendations in the absence of soil tests

Abstract

In many developing countries, fertilizer recommendations must be made in the absence of plant and soil analyses. When the region is variable in terms of soils, weather, and magnitude of response to fertilizer, a recommendation is likely to involve a high degree of risk for the farmer. Quantification of such risk is key to developing appropriate recommendations for the farmer. However, most methodologies generally used in analyzing fertilizer trials do not allow adequate quantification, especially as a continuous function, of the risk associated with a given recommendation. Three years of on-farm trials conducted in the High Valley of Mexico were used to evaluate different methodologies for generating N fertilization recommendations and their associated risk for wheat (Triticum aestivum) production in the absence of soil tests. When the traditional approach, using average yield responses or separate trial results, was used, an economic optimum was identified, but it was not possible to quantify the associated risk. In contrast, however, by using a combination of response surface methodology and simple probability analysis, the risk associated with any given recommendation was developed, even under the highly variable conditions of the study zone. The approach uses a treatment difference matrix (developed using average yield differences between a treatment and the zero N (0N) check) and its associated standard deviation over locations. From the matrix, an equation (being a function of N rate and relative grain:N price ratios) was developed that shows the probability of outperforming the 0N check for the economic optimum rate.