Skip to content
menu icon

Field trial sampling protocol targets 50 per cent saving

Ohio State University Professor of Statistics Omer Ozturk says structured sampling approaches for research can cut the costs of field experiments.
Photo: Katherine Hollaway

Systems for accurate and efficient analysis of field trials under international spotlight.

Key points

  • The University of Adelaide's Biometry Hub recently hosted a visit by US Professor of Statistics Omer Ozturk
  • Professor Ozturk is an expert in ranked-set sampling - a technique that could save up to 50 per cent on sampling costs in field experiments

During a recent visit to Adelaide, Professor Omer Ozturk - a leading biometrician from the US - challenged Australian researchers to cut field-trial costs in half using an innovative sampling procedure.

Every research dollar invested by GRDC is targeted at improving the success of the Australian grains industry.

But field research is an expensive and time-consuming business and biological systems are notoriously demanding to research, particularly in the paddock where variables are more difficult to control.

To maximise the outcomes from this research, GRDC invests in cutting-edge statistical analysis through the Statistics for the Australian Grains Industry (SAGI) program.

International skills

While SAGI's research is already world-class, the team continues to draw on the skills of international colleagues to find fresh ways to get more from Australia's field-research program.

The SAGI South team at the University of Adelaide recently welcomed ranked-set sampling expert Professor Ozturk, from the Ohio State University, to help develop new ways to manage the complexity of sampling - particularly in the areas of:

  • disease monitoring;
  • weed assessment; and
  • soil-health data collection.

Historically, field-sampling design assumes the random sampling procedure - based on the theory that random selection will yield a sample that is representative of the population.

But there are drawbacks to this technique. For instance, the random sample may or may not be representative of the whole population. This challenge is normally overcome by increasing the sample size, but this adds to the overall cost and degree of difficulty.

In contrast, the ranked-set sampling technique - one of several structured approaches - is particularly useful when data collection is costly, destructive or time-consuming.

Structured approaches to sampling involve using additional information to direct attention towards measurements that are more representative of the population. Sometimes this information is collected by taking simple measurements, but it can often be done by eye.

Big savings

"It is essentially making an informed decision about which samples need to be taken," Professor Ozturk said during his six-month sabbatical.

"Done well, the technique can save up to 50 per cent on sampling costs."

For instance, sensor technology could be used to rapidly measure the relative greenness of plots to subdivide them into groups by colour. From here, a selection of plots from each subgroup is identified for more detailed and destructive sampling.

This approach was first proposed by an Australian - CSIRO's George McIntyre - in 1952 as a way to improve yield estimates without substantially increasing the number of plots from which detailed, expensive and tedious measurements needed to be collected.

SAGI South's team leader, Dr Olena Kravchuk, first discussed the approach with Professor Ozturk at a biometry conference in Michigan and was delighted when he offered to work with the statisticians and data scientists at Adelaide's Biometry Hub to extend the applicability of structured sampling procedures to field surveys in agriculture.

Done well, the technique can save up to 50 per cent on sampling costs. - Ohio State University Professor of Statistics Omer Ozturk

The Biometry Hub hosted an overlapping visit from Dr Blair Robertson, from the University of Canterbury in New Zealand, to consider extension of the theory to applications of spatially balanced sampling in field work.

The collaboration established during this visit will grow for many years, with plans for the development of an open-source R-package for ranked-set sampling.

Biometry Hub biometricians Peter Kasprzak and Sam Rogers will be closely involved, building the local capacity in these methods for Australian growers.

GRDC Research Codes UA00157, DAN00172, DAS00167, DAV00150, UA00164

More information: Dr Olena Kravchuk, 08 8313 7252,

back to top