The foundation of all methods for determining/predicting nutritional status described
below is based
on experiments with plants, typically fertilization trials. These experiments can be conducted in an
almost unlimited variety of ways, although there is an optimal experimental design for each
research question. To what extent compromises can or must be made in the experimental design
and execution cannot be discussed in detail here. However, let me quote a colleague's saying:
"What you cannot do correctly (or appropriately), you should better leave undone."
Fundamentals:
Experiments with plants serve to derive "laws" or generalizable insights.
In the applied field of plant
nutrition, this often involves establishing dose-response relationships and/or time series analyses.
Dose-Response Relationships (dose-response curves)
The effect of increasing one or more factors on the expression of one or more traits
is investigated.
Besides adhering to the rules of statistics (if statistical inferences are desired – which
is not
mandatory), it is important to:
- Choose the range of the varied factor
so that the range of interest is covered (preliminary
experiment).
- Select experimental conditions so
that expected differences become measurable (climate
conditions).
- Choose the number of gradations so
that a meaningful dose-response curve is generated
(minimum 4 levels).
- Time Series Analyses: The temporal
change of one or more traits is investigated. The
principles mentioned for dose-response relationships apply accordingly.
Execution:
The proper execution of experiments is significantly influenced by available experience
(one's own
and others'). A few rules stemming from personal experience:
- Ask only questions that have not
been asked before, or not in that specific way (don't reinvent
the wheel or its essential parts).
- Formulate the research question simply
and clearly (Everyone, including non-specialists, must
understand what you want to know – and why).
- If possible, vary no more than 2
factors in one experiment (the number of comparisons and
interactions becomes unmanageable).
- Do not bend yourself, statistics,
or common measurement methods to circumvent technical
deficiencies or inadequate equipment (If you have no PAR sensor for photosynthesis
measurements, don't use a lux meter as a substitute).
- Measure that, and only that, which
is necessary to answer the research question (Data
hoarding is superfluous and costs money and time in setup).
- Report the results to colleagues
and discuss them openly.
Analysis:
In principle, the procedure for analyzing the experiment must be established before
it is conducted.
Only this ensures that all necessary variants are present and all variables are measured that are
required to answer the research question. Beyond standardized procedures, here are a few
personal remarks:
- Satistics relieves you of the decision
of whether something is significantly different from
something else or not – it does not relieve you of the decision of whether this difference
is
important or not.
- Statistical/mathematical curve fittings
only make sense if there is a scientific rationale for their
application. Often, the hand, eye, and brain make equally good curve fits, but unfortunately(?)
without equations and coefficients of determination.
- Large datasets are much easier to
comprehend as graphical representations than as tables.
However, more than 5 curves are difficult, and 5 intersecting curves are practically impossible
to clearly represent using symbols, line styles, and line weights (neither in black-and-white nor
in color).
Possibilities/Limitations
Pot/Container Experiments with Plants: Due to the small substrate volume
and the typically
movable containers, they offer the possibility for:
- A high number of treatments (experimental
units)
- With a high number of replications
- With limited space requirements
- In climate chambers, greenhouses,
and wire houses.
However, due to the small volume of the root zone, they are only a limited representation
of growth
conditions under field conditions. This significantly restricts their transferability.
Field Experiments on Natural Soils are characterized by:
- A (large) spatial variability of
soil conditions, which must be accounted for in the experimental
design.
- Unpredictable weather patterns, which
can only be imperfectly compensated for by irrigation or
sheltering.
- Increased requirements for time,
labor, and space.
