Vision No. 14 June 2009

Simple legumes, complex analyses

Since starting work at VSNi it’s no surprise to find so many applications of our software, be it GenStat or ASReml, are working within some aspect of agricultural research; projects that range from development of new plants to new herbicides or pesticides. But this isn’t just a race to find the best, most productive type of grain, or the most effective pesticide, in so many cases this is linked to the welfare of the local ecology. More and more agricultural producers can now see a benefit of using the earth’s natural resources to assist in crop production.

Researchers at the New South Wales Department of Primary Industry in Australia highlight the complexities of farming using one of the more traditional methods of crop production. Southern Australian farms have long used annual legumes in the pasture phase of crop rotation methods. These act as natural disease breaks and restore soil structure and nitrogen in the soil, as well as providing valuable, high quality feed for livestock. Fantastic! In the panic of today’s world on pesticide poisoning this seems an ideal solution, and has certainly been a good one – here we are using the earth’s natural resources and traditional crop rotation methods to help increase production, and as a bonus create livestock feed.. however it is not that straightforward.

The problem with annual pastures and crops, is that they are often shallow rooted and therefore use less rainfall than the deeper rooted perennial species. This has led to a significant increase in groundwater recharge with higher water tables and dry land salinisation in parts of Australia where the agricultural production has meant a change from perennial deep-rooted species to annual crops and pastures. Simple answer! Surely we just swap the annual pastures and crops for perennial ones… but which one? Australia is a big country with a diverse range of climates. How then do you determine which perennial legumes and herbs are best suited to Australian conditions, with so much variation between regions?

A team of researchers headed by Guangdi Li took on the challenge. The purpose of their study was to evaluate a wide range of herbaceous perennial species as potential new species for incorporation into the farming system of Southern Australia. Guangdi Li’s team needed to identify species with an ability to adapt to a broad range of environments, and species which are particularly suited to specific soil conditions, environments and climatic patterns. They evaluated 91 perennial legumes and herbs (entries) at 10 sites across Southern Australia. There were originally 17 sites but 7 were abandoned because of weeds or poor establishment leaving 4 in Western Australia, 1 in South Australia and 5 sites in New South Wales. Only a subset of the 91 was sown at each site, local conditions not being conducive to the successful establishment of many of the varieties. In fact only a handful of the entries were sown at all the sites. Some of the sites were chosen because they posed environmental constraints to plant growth and hence could be used to identify alternate species that may be better adapted to these more demanding environments. Those constraints were saline or waterlogged soils, although drier than usual conditions minimised the effects of waterlogging. Over the three years of the experiment there were a number of occasions that herbage mass and plant frequency was assessed across the ten sites, creating 67 and 21 ‘environments’. Each environment provides information about the performance of the varieties, but of course successive samplings from the same site are not independent of each other, and a high correlation was expected.

ASReml was chosen to analyse this data because it can handle the estimation of the genetic correlations between pairs of environments. The analysis involves a mixed model in which the variance-covariance matrix of the entries in different environments is modelled using a factor analytic (FA) structure. The FA model facilitates the accurate prediction of entry means for individual environments using best linear unbiased prediction. Selection across a range of environments, be it the saline or waterlogged sites or even for those samplings taken in Summer for instance, is easily achieved by using the Predict facility in ASReml. Entries that showed potential in saline or waterlogged conditions can be targeted for inclusion in future breeding programs.

The complexities of this study very much reflect the complexities faced by the agricultural producers. In farming there are often no straightforward, obvious answers to the questions posed and the problems faced. This is why research and studies such as Guandi Li’s are so important; they give evidence based on sound science and statistical analysis techniques which show the importance of specific types of plants; in this instance Lucerne, which performed well over a broad range of environments. This in itself suggests projects are needed to fully exploit this plant so that its limiting factors (susceptibility to acid conditions and heavy grazing) can be overcome. Additionally Guangdi Li and his team showed that there is a range of deep-rooted perennial legumes and herbs which could be used in the Southern Australian farming system, rather than relying on a narrow range. The impact of the results of this and future trials are huge, it is yet another step closer to even better planned farming policies and methods, and hence to a true understanding of how we best feed ourselves and the world.

There is no doubt that sound scientific research is needed on which to base future crop, farming methods or policy decisions; research that is built using proven scientific tools. VSNi are a key provider of these tools; we know this because many agricultural scientists rely on us to continue providing high quality, relevant statistical analysis software built by statisticians who understand the complexities of agricultural research and data.

For more details on ASReml, its capability and why over 41% of papers presented at the last meeting of the Australian Association for the Advancement of Breeding and Genetics cited ASReml look at the VSNi website.

Technical tip – New Guide for Analysing Climatic Data in GenStat

We are delighted to announce a new guide which is available for download from the VSNi website. The new guide titled Analysing Climatic Data using GenStat, covers an introduction to GenStat, data input and basic data manipulation, simple statistical inference, including the use of boxplots and comparison of means and basic regression. After mastering the basics, the guide explains the summary and presentation of climatic data and then more detailed explanation of statistical methods including linear, non-linear, multiple regression and multivariate analysis. Finally, the guide assists users with GenStat programming for those who want to move from menus to more complex modelling and analysis.

Our thanks to the Statistical Services Centre at the University of Reading for this guide.

Latest training courses

New courses coming over the summer include:

  • Introduction to ANOVA, 9th-10 July, Apsley, Hertfordshire, UK
  • Introduction and Intermediate ASReml, 12-13 August, Apsley, Hertfordshire, UK
  • REML and multi-level models in GenStat, 15 September, Apsley, Hertfordshire, UK

GenStat users who are looking for assistance with Analysis of Variance should book onto the Introduction to ANOVA on 9th to 10th July, in Apsley, Hertfordshire, UK.

The aim of the course is to introduce ANOVA techniques and develop the fundamental knowledge and skills to use them correctly and effectively. The basic principles of experimental design will also be included to aid the effective planning of experiments and investigations.

By the end of the course users will be able to:

  • Select an experimental design appropriate to the requirement of your investigations.
  • Use randomization to avoid bias in the allocation of units to treatments, to ensure that results are reliable and unaffected by any systematic patterns in the units.
  • Determine how many replicates are required for your designs.
  • Use blocking to increase the accuracy of an experiment by forming the basic units (e.g. plots or subjects) into groups with similar properties.
  • Analyse simple to sophisticated designs, explaining ideas such as balance, and to recognise the GenStat features available for the analysis of unbalanced designs.
  • Compare several types of treatment in the same experiment.
  • Interpret experimental results and produce relevant tables, graphs and figures for publication in reports and papers.

Dr Arthur Gilmour is running Introduction and Intermediate ASReml Courses on 12-13 August again in Apsley, Hertfordshire. Participants will gain an insight into the concepts underlying ASReml and hands on experiences relevant to their own applications. Participants are welcome to prepare examples they would like analysed during the workshop and forward them before the end of July to support, so that the courses can be tailored to include specific user examples.

Day one will cover an Introduction to ASReml and simple univariate models. On day two participants will cover variance structures and be able to bring their own examples for discussion.

On 15th September we’ll be running a course on REML and multi-level models in GenStat. To take advantage of the early bird discount book before 31st July 2009. The course is designed to introduce the user to REML in GenStat, and provide the underlying knowledge and confidence to use it correctly and effectively. It highlights how the REML menus guide the user through very complicated analysis and explains the REML commands so that users can program any non-standard analyses.

The course covers 4 main sessions:

  • Analysis of variance
  • Meta analysis
  • Spatial analysis
  • Repeated measurements

For more details and to book email support or go to the training pages on our website.

Out and about with VSNi

We are delighted to be sponsoring two events in Italy – the ICORS 09 (International Conference on Robust Statsitics) in Parma from 14-19 June, and then later the EUCARPIA conference in Bergamo on 21-24 June.

This year we are looking forward to visiting old and meeting new friends when our CEO Stewart Andrews speaks on GenStat Discovery and the importance of agricultural research in Africa at the ISI 2009, in Durban (16-22nd August). Stewart will also lead a session on data analysis software for the developing world; more details to follow.

August is looking a busy time as we are also speaking and sponsoring the 14th Australasian Plant Breeding & 11th SABRAO Conference from 10-14th August at Cairns Convention Centre in Queensland.

From 2-4th September we’ll be in Dundee at the EUCARPIA 2009 XIV meeting of the Biometrics in Plant Breeding Section.

If you would like to meet directly with a VSNi attendee at these events please email Carey Biggs to make the arrangements.

We’re always updating the list of events we can support and sponsor – so please send us details of any events you are organising or involved in, and as we decide on more events for the future we’ll list them on our website.


Comments are disallowed for this post.

Comments are closed.