Evaluation of small site-specific N fertilization trials using uniformly shaped response curves

The spatial variability of yields and optimal N rates is well documented for several crops and the benefit of studying site-specific N response seems self-evident. Unfortunately, trials established to examine site specific N response are mostly quite large and expensive, since those trials usually cover nearly the whole field. Selecting only small subareas of a field to host an N fertilization trial will reduce costs and therefore possibly allow more trials on different locations. On the other hand, the evaluation of small trials can become difficult, resulting from a low number of observations compared to the number of parameters that need to be estimated. Aim of this study was to develop a framework that allows evaluating yield response from small site-specific trials by reducing the number of parameters of the response curves. The framework was tested on a data-base of a two-year site-specific N fertilization trial in northern Germany. Three different types of N response curves were evaluated individually for each subarea: Linear plus plateau (LP), quadratic (Q) and quadratic plus plateau (QP). In a second step an alternative parameterization that includes a shape parameter was introduced. For each type of response curve, this shape parameter was set to a common value for all subareas within a field and year, thus reducing the number of parameters that need to be estimated. Results show that the reduction of parameters did not reduce the goodness of fit when describing yield response, since the yield response of each subarea still depends on different maximum yields and different N amounts to attain their maximum yield. It turned out that the models using a common shape parameter allowed to evaluate even subareas with separately indetermined parameter values. Furthermore, agronomically meaningful correlations of the estimated response curves to the empirically observed characteristics of their respective subareas were improved. (C) 2016 Elsevier B.V. All rights reserved.