In the last blog post on yield maps, I pointed out that future yields cannot be planned and that it is much more important to measure the needs of an input correctly. Yield potential maps and yield targets are mostly used to measure the optimal nitrogen fertilisation. However, as I will explain in a moment, this approach is misguided. The attempt is made to calculate the required N fertiliser by balancing plant growth, N consumption and supplied N fertiliser, as in a static black box. However, nature is not that simple.
The (non-) correlation between optimal yield and optimal N fertilisation
The level of optimal N fertilisation is defined at the point where the cost of the last kilogram of nitrogen is still financially covered by the yield increase. If one wants to approach this question in a scientifically correct way, N-increase trials are usually carried out on plot trials. The result is shown by the following example of a 20-year trial series from Saxony, representative of many similar trials in Germany and worldwide.
The crop grown on the same field was always winter wheat. In 13 out of 20 years, the fertilisation optima were between 25 and 100 kg N/ha above or below the mean value. If one estimates this deviation cautiously on the basis of a mean production function for nitrogen, the yield or nitrogen advantage for differentiated N fertilisation is about 80 €/ha and year. This is an order of magnitude worth considering. Not included in the calculation are also the positive effects with regard to storage prevention, quality improvement, harvesting and disease risk. The following graph shows the table columns "optimal N application" and "optimal yield". Neither the target yield (in the sense of optimal yield) can be predicted with certainty, nor can one draw any direct conclusions about the optimal N fertilisation. So two things come together here that rule out this path.
Only 14% (R²=0.1438) of the amount of optimal N fertilisation is explained by the optimal yield. This is now also the second reason why yield potential maps do not work for fertilisation. Once again as a reminder:
