How to do a DIY experiment on your block

to measure. It’s the ones that get turned into profit such as the stuff you harvest: lambs, grapes, apples, lettuces, wheat grains. Disappointingly, this is often the measurement that gets missed by scientists.

However, it is also important to measure ‘intermediate’ parameters, such as growth during the whole season, plant nutrient levels etc, as these are important for helping you to understand what is going on.

There is a mantra in science that ‘correlation does not imply causation’. That is, if you only measure yield, you don't know why the yield increased so you only have a correlation which is weak science. If you measure other parameters, these can point to how the increase was caused, giving you stronger science.

What time frame do they cover?

For products such as biostimulants that have an immediate and relatively short term effect, trial duration is typically one crop cycle, based on the assumption that there is little or no residual effect; if you stop using the product, then the effect stops after a week to a few months.

However, it is rare for effects to be truly short term so, if resources allow, the experiment should be run for three to five years to see what the long-term effects are.

For products such as biofertilisers or anything that impacts on soil processes, duration should be as long as possible because soil processes and performance change very slowly. It really can take decades for the long-term effects to be fully shown. Truly long-term soil experiments around the world have now been running for over a century and data from these shows that it takes up to 50 years for soil to truly reach a new equilibrium. When the first 10-30 years data from these experiments are analysed they often give quite different results compared with analysis after 50 years. If scientists are being really hard core about such trials, they will throw out the data from the first five years, have a look to see if there are any trends in the next five years, and then consider data after the first decade as starting to become reliable. If you are running or looking at data from experiments that affect the soil, a trial should really be kept running for five years at a minimum, but ideally a decade.

How to do your own experiments

Agricultural experiments are among the simplest, and the value of DIY experiments is that they are done on your crop or pasture so the results are 100% meaningful for your operation. All you need to do is follow a few simple rules.

Limit what you’re going to test

Treatments are the different products you want to test. More is not always merrier as the amount of work increases considerably.

Decide on a regime

It is important to decide the application regime from the start: is the product to be applied once at the start of the trial, or sprayed on weekly? The application regime should match what would be done in the real crop.

Have a ‘control’

You need to have something to compare, a control area, where nothing is applied to the crop, and/or you use your current practice, eg your current fertilisers. The control needs to be replicated and randomised just the same as the treatments.

Decide on a duration

Getting the experimental duration right is really important. For biostimulants that is typically one crop cycle but ideally three or more, while for biofertilisers the duration should be as long as possible, ideally five years but a decade is much better.

Repeat, repeat, repeat

Like the farmer spraying several strips of seaweed fertiliser on his peas from page 24, you need to have several applications of the treatments. Traditionally the minimum is four replicates but in a perfect world six to eight is best.

Be as random as you can

It is impossible to emphasise how important proper randomisation is. The pea farmer sprayed alternating strips up his field, but it would have been better if he’d flipped a coin at the end of each row – head for spray, tails for a control – and kept going until he had enough replicates (spray strips) of the seaweed

and unsprayed (control). Randomisation helps take chance out of the experiment, so you know you didn’t accidentally add all the treatment you are testing on an area that by chance had higher or lower fertility anyway. The standard layout for field trials is the randomised complete block (RCB). Figure 1 shows a RCB experiment layout with four treatments (a, b, c, d) and four replicates. The key to blocking is that each of the four treatments (or however many there are) is found in every one of the blocks, creating a complete block. Plots need to be big enough so that the natural variation found in agriculture is minimised, so bigger is better. Don't make the mistake the pea grower did of taking his harvester’s performance as a measure of pea volumes. It is essential to measure the final product, the thing you sell to make money. That's easy in agriculture or horticulture but harder with livestock (very large plots and lots of stock are required), so with animals, the surrogate measure of pasture growth and laboratory analysis is mostly used. Statistics are typically the most confusing part. Fortunately the ANOVA test (analysis of variance) found in most spreadsheets is typically used. However, if you are not comfortable with statistics get help from someone who is. While the basics of an experiment, as outlined, are really pretty straightforward, there are niceties in the details that take experience to get right so talking to a scientist can help you. If the product you are applying costs $200/ ha to use but only increases income by $100 you are $100 out of pocket (profit has reduced $100), although there may be some other benefit, like an increase in soil organic matter over the longer term which results in bigger yields in future. But the ultimate measurement of an experiment is not yield, it is profit, so it is critical that gross margins for all the treatments are calculated to test for the level of profit or loss.