Methodological requirements for the organization and conduct of field agrochemical experiments

The results of any field trial have practical value only when they can be seamlessly transferred to production fields. A field experiment is unique: it is the only setting where a crop interacts with the entire complex of soil, climatic, and agrotechnical factors. Without field trials, it is impossible to reliably estimate the actual yield increase, the residual effect of fertilizers, or their efficiency in crop rotation. For the experiment not to become a waste of time and resources, methodological rules must be strictly followed.

Rule one: typical conditions and a high background level of agricultural technology

What is typicality, or representativeness, of an experiment? It is the correspondence of experimental conditions to those fields where you plan to implement the technology. The experiment is laid out on the soil types predominant in the farm. In this case, it is essential to take into account the particle-size distribution, humus content, acidity, and nutrient availability.

Climatic conditions are unpredictable, so one successful or drought-stricken year does not provide an objective picture. To level out the influence of weather, field studies using the same design are conducted in cycles. This is the only way to obtain reliable average data for a specific zone.

• Duration of one trial scheme — 3–4 years
• Cultivars for the experiment — regionally adapted and promising
• Agrotechnical background — high

Experiments on a low agrotechnical background have no practical value for production. If the land is neglected, fertilizers will show high efficiency, but these figures will be far from the reality of ordinary fields. It is also important to conduct trials under the conditions of a scientifically based crop rotation, taking into account the predecessors and the actual manure availability of the farm.

Rule two: logical distinction and accuracy of accounting

The principle of the single logical distinction allows for the correct comparison of treatments. This means that on all experimental plots, growing conditions must be identical, except for one factor being studied. In fertilizer experiments, this can be the dosage, form, or application method.

When comparing different forms of fertilizers, their application rates must be identical. To eliminate random errors, the following parameters are strictly leveled in the experimental plot:

• tillage;
• predecessor;
• cultivar;
• sowing dates and maintenance practices.

It is important to avoid formalism, as the principle of single distinction must be strictly logical. For example, the effect of nitrogen and phosphate fertilizers on the yield of winter crops might formally seem to require study under identical conditions. However, the efficiency of these fertilizers is directly determined by the specific timing and method of their application.

The main criterion for the efficiency of any practice in agrochemistry remains the yield and quality of the produce. The yield volume integrates and reflects the effect of all growing conditions on the crop. It is the precise quantitative accounting of the harvest and the assessment of its quality that allow for reliably confirming the benefit of the factors being studied.

A field agrochemical experiment evaluates not only the gross harvest but also key qualitative yield indicators. These include:

• protein content in grain crops;
• starch accumulation in potato tubers;
• sugar concentration in sugar beet;
• fiber length in flax.

However, any quantitative results obtained under field conditions always remain only an approximate expression of the truth. The degree of this correspondence is called the accuracy of the experiment. The smaller the difference between the actual yield on the scales and the real effect of the studied factor, the higher the accuracy and the lower the experimental error.

Anatomy of errors: random, systematic, and gross errors

The reasons for the discrepancy between experimental data and the truth lie in unavoidable errors. To correctly assess the results of the work, an agronomist needs to understand the nature of these errors. In field practice, they are divided into three categories.

Random errors arise due to a multitude of small, unaccounted factors. They occur during the marking of plots, weighing of fertilizers, harvesting, or analyzing its structure. The main cause is the heterogeneity of soil fertility, relief features, as well as the technical impossibility of tilling the soil, applying fertilizers, and distributing seed with absolute uniformity. The influence of random errors is reduced by changing the shape and size of the plots, as well as by increasing the number of experiment repetitions.

Systematic errors constantly distort the result in one direction — either overestimating or underestimating it. Most often, they are associated with malfunctioning or poorly calibrated measuring instruments and scales. Such errors do not compensate for each other when averaging data, so they are fully carried over into the final experimental results.

Gross errors (blunders) arise due to the negligence of personnel or force majeure events. This could be double fertilizer application on one plot, seeding misses during sowing, confusion with bags during weighing, or incorrect records in the logbook.

Mathematical control: how to calculate the precision of an experiment

To confirm the reliability of an experiment, all obtained data are processed using variation statistics methods. The quality of the work performed is assessed via the experiment precision index. It expresses the magnitude of random error as a percentage of the average yield.

Precision requirements depend on the scale and objectives of the research. Long-term stationary experiments always require higher precision than short-term production trials. If you are studying the efficiency of different types of fertilizers and expect a large yield increase, the precision requirement threshold can be lower. When comparing similar forms of fertilizers, where the difference in yield increases is minimal, the experiment must be as precise as possible.

• Number of error types in an experiment — 3
• Precision of stationary experiments — 3–4 %
• Tolerance for production experiments — 8–10 %

In agronomy, it is customary to distinguish between the reliability of a field experiment in essence and the statistical significance of the obtained results. Only adherence to all methodological rules guarantees that the obtained figures truly reflect the performance of fertilizers. This allows avoiding false conclusions when implementing technologies into production.

Any field experiment must answer a specific question: did the studied factor actually work, or was the difference in yield caused by random reasons? To assess the quality of an experiment, one first checks the correctness of its design, accompanying observations, and the accuracy of yield accounting. If the methodology is fully observed and there are no obvious violations, the data is sent for mathematical processing. Statistics help to separate the real effect from random fluctuations in soil fertility or technical errors.

What is the difference between precision and reliability of an experiment

The precision of an experiment and the reliability of its results are related concepts, but they are not identical. Precision shows the magnitude of random errors during accounting. Reliability (significance) proves the mathematical significance of the difference in yields between variants. In practice, these indicators can be combined in different ways.

The difference between precision and reliability is clearly demonstrated by two examples from field trial practice:

• High precision with an unreliable result. When studying the efficiency of various application rates of boron fertilizers, the experiment was conducted very carefully, with minimal error. However, no noticeable differences in yield between the variants were found. The obtained difference turned out to be less than the mean error of the experiment, meaning the result is mathematically unreliable.
• Low precision with a reliable result. In an experiment with another micro-fertilizer, the error was comparatively high. But the yield increase compared to the control without fertilizers far exceeded this error many times over. As a result, despite the lower precision of the experiment, agronomists obtained a mathematically proven reliable increase.