Re‐visiting potassium‐ and phosphate‐fertilizer responses in field experiments and soil‐test interpretations by means of data mining

Currently, potassium (K)‐ and phosphate (P)‐fertilizer recommendation in Germany is based on standardized soil‐testing procedures, the results of which are interpreted in terms of nutrient availability. Although site‐specific soil and plant properties (e.g., clay and carbon content, pH, crop species) influence the relation between soil nutrient content and fertilizer effectiveness, most of these factors are not accounted for quantitatively when assessing fertilizer demand. Recent re‐evaluations of field observations suggest that even for soil nutrient contents well within the range considered to indicate P or K deficiency, fertilizer applications often resulted in no yield increase. In this study, results from P‐ and K‐fertilization trials (in total about 9000 experimental harvests) conducted during the past decades in Germany and Austria were re‐analyzed using a nonparametric data‐mining procedure which consists of a successive segmentation of the data pool in order to elaborate a modified recommendation scheme. In addition to soil nutrient content, fertilizer‐application rates, nutrient‐use efficiency, and site properties such as pH, clay content, and soil organic matter, have a distinct influence on yield increase compared to an unfertilized control. For K, nutrient‐use efficiency had the largest influence, followed by soil‐test K content, whereas for P, the influence of soil‐test P content was largest, followed by pH and clay content. The results may be used in a novel approach to predict the probability of yield increase for a specified combination of crop species, fertilizer‐application rate, and site‐specific data.