We would like to cultivate a culture of discussion

...with our agriconBLOG

agriconBLOG

01.03.2019 - Manuel Ermann

Downed - when trials go wrong

At universities and other research institutions in the field of crop science, exact experiments are predominantly carried out on small plots of land. Explicit care is taken to ensure that disturbing factors are excluded. The homogeneity of a plot is the most important starting condition for this type of field experiment.

However, our environment is anything but homogeneous: it is diverse and full of influencing factors. The evaluation of real cultivation systems and strategies therefore focuses on the naturally occurring heterogeneity. In order to achieve this, on-farm research experiments were developed several decades ago. With these experiments, the special evaluation of system comparisons on the basis of practical proposals becomes possible. The complete field plot is used as the experimental area. In addition, the farmers working on the land are involved in the OFR trials. In this way, randomised long plots can be cultivated using standard farming techniques. All external influencing factors such as soil,
terrain and initial stands are additionally recorded and analysed geostatistically.

The fact that the execution of OFR experiments can also go wrong if the above-mentioned principles are not observed (we will list them again later) is illustrated by research work carried out by the Schleswig-Holstein Chamber of Agriculture at Gut Helmstorf a good ten years ago. The sensor fertilization should be compared with the usual fertilization as well as the single/dual fertilization. The project managers came to the following conclusion: For the three fertilisation strategies, the variant "customary in the farm" on average [...] proved to be the most economical. Compared to single/dual fertilization, N-sensor fertilization did not bring any economic advantages.

Unfortunately, this result is purely coincidental and therefore has no claim to general validity - neither in Schleswig-Holstein, nor in Germany or Europe. In this blog post we explain to you which mistakes the Chamber of Agriculture of Schleswig-Holstein made during its OFR experiments. It is important to us that you, as a reader, understand the stumbling blocks that can be overcome in the practice-oriented evaluation of cultivation systems and strategies.

The history of OFR experiments

In 2011, the task force on Agricultural Experimentation of the German Region of the Biometric Society will develop a guideline for the planning and evaluation of OFR experiments.

This was done in cooperation and close coordination with

the research group Experimental Science of the Gesellschaft für Pflanzenbauwissenschaften (Society for Crop Science)
the task force Biometry and Experimental Methodology of the German Phytomedical Society
the Biometrics and Bioinformatics task force of the Society for Plant Breeding
the Technical Committee "Experimental Systems in Plant Production" of the DLG

The experts involved defined six important principles which must be adhered to in OFR experiments:

At least triple repetition of the test elements
Random distribution of plots (randomization)
Digital, spatial recording of the test facility, the application of the equipment, the yield and all disturbance factors
The geostatistical evaluation must include the spatial dependence of the examined parameters/measured values (semi-varioprogram model).
Establishing and testing a regression model
Display of results with statistical measures such as standard error and significance (p-values)

In a press release issued in May 2013, the DLG also clarified once again: "Agricultural field trials [...] require well-founded statistical test planning. [...] If methodological minimum requirements such as repetition, equal treatment (ceteris-paribus principle) and randomisation are not observed, the results obtained can only have a demonstration character and are not suitable for publication. Furthermore, the Higher Regional Court of Düsseldorf (OLG Düsseldorf, AZ I-20 U 141/08) emphasized in a judgment that a stripwise arrangement without repetition rarely makes methodical sense and "prohibits the use of such data in the context of comparative advertising". [...] Without repetition, attempts are worthless."

The experiment at Gut Helmstorf was consistently faulty

Let us switch back to the Chamber of Agriculture of Schleswig-Holstein and Gut Helmstorf. As already mentioned at the beginning of this article, none of the clearly defined six basic rules were observed in the OFR project on N fertilization.

The following comparison shows the first mistake: The picture on the left shows a test shot with a strip line that has proven itself in practice and fully meets the requirements of OFR tests. The picture on the right illustrates the test rig of the "Hansberg" strip from the OFR project at Gut Helmstorf. The subdivision of the field into three parts carried out here is inadmissible in the experimental system and therefore does not meet the minimum requirements:

Equal treatment (ceteris-paribus principle)
Random distribution of plots (randomisation)
reiterativeness

Figure 2 shows that in the project at Gut Helmstorf the simple subdivision of the test shots into three parts was chosen. Unfortunately, this means that all the results achieved there are beyond any meaningful statistical consideration.

The disadvantage of individual test elements is highly problematic

It is to be assumed that individual test elements were disadvantaged from the outset by the impermissible choice of test facility. As an example for the next fault, the blow "Hansberg" is used again. The picture on the right shows the yield potential map as shown in public lectures of the Schleswig-Holstein Chamber of Agriculture. The seven-year average yield is shown in concrete terms.

If the data is processed, the error committed becomes clearly visible: on the left the division of the test elements, on the right the yield potential map: the brighter the sub-areas are, the lower is the yield potential. It becomes immediately apparent that these bright partial areas are primarily located in the YARA N-Sensor® plot.

A calculation of the average yield potential for each variant confirms the optical impression:

once or twice fertilizer application	sensor fertilization	usual
135 %	83 %	118 %

The sub-area of the sensor fertilisation shows only 70 per cent of the yield potential of the sub-area "normal" and 61 per cent of the sub-area "single/dual fertilisation". Nevertheless, in the Helmstorfer Gut trials with the YARA N-Sensor® almost the same yields were achieved as in the comparison variants.

Résumé

We have been able to show by way of example that the OFR experiments on N fertilization at Gut Helmstorf do not meet the minimum methodological requirements (repetition, equal treatment, randomisation) for experiments of this type. Therefore, the results obtained at that time cannot be used for the general qualification of fertilizer strategies.

Unfortunately, these "independent" test results were taken as an opportunity to deny the benefits of the YARA N-Sensor® and the N-fertilization supported by it. From our point of view, this created uncertainty and confusion for many farmers and wasted great potential for economically and ecologically more successful nitrogen fertilization. Over the past 20 years, the potential of the YARA N-Sensor® has been demonstrated in over 250 ORF trials in various regions
of Europe:

3-6% more yield,
up to 12% N savings,
up to 30 kg N/ha N balance improvement,
12-20% increase in performance in threshing,
0.2-0.5% higher and more homogeneous quality,
complete prevention of nitrogen-induced bearing and
on average 100 €/ha additional surplus

.. annual!

source materials:

Biometric society: Guideline with biometric principles for planning and evaluation of so-called "On Farm Experiments", 02.02.2016: http://www.biometrische-gesellschaft.de/arbeitsgruppen/landwirtschaftliches-versuchswesen/on-farm-experimente/einfuehrung.html

Statement DLG Committee for Experiments, May 2013: Landwirtschaftliche Feldversuche: Auf methodisch korrekte Planung, Durchführung und Auswertung achten, Mai 2013

Dr. Ulfried Obenauf: On Farm Research, N-Sensor in practical use, lecture DLG-Workshop, 21./22.11.2012 Groß-Umstadt

Dr. Ulfried Obenauf, Imke Borchardt, Christoph Lubkowitz (Schleswig-Holstein Chamber of Agriculture), Carsten Kock (Helmstorf Estate Management), topagrar 02/2013 Precision Farming: N-Düngung im Praxistest