Computer power starts to build risk-management muscle
GroundCover™ Issue: 88 | 01 Sep 2010
By Dr Gio Braidotti
Environmental stresses pose such multi-faceted problems for agriculture that a new breed of scientist is emerging to help meet crop production challenges. Neither pre-breeder nor agronomist, the new specialist is drawing on the best of previously distinct disciplines – genetics, agronomy, ecology … and mathematics.
One such pioneer is Professor Graeme Hammer, the GRDC-sponsored leader of the Agricultural Production Systems Research Unit at the University of Queensland.
He says the idea is to weave together into one integrated ‘knowledge bank’ information about land management, soil, climate and plant behaviour and link it to software that can predict how these are likely to interact under a variety of typical seasonal conditions.
The approach is dealing with a long-standing scourge of stress-related research – the complex way that genetic variation, management practices and environments interact to thwart predictability and resilience within experiments and cropping systems.
Given its overarching nature, the new discipline has been dubbed the genotype (G) by environment (E) by management (M) approach … or G-by-E-by-M (GxExM).
Professor Hammer uses computers to capture vast amounts of data on climate, soils and cultivars in parallel with field experiments to understand the principles that affect crop productivity. These are then embodied in mathematical equations that allow the scientists to simulate plant breeding and farm management programs inside a computer.
In developing applications for this modelling technology, Professor Hammer collaborates closely with pre-breeders working on water stress, especially those taking paddock-driven approaches to genetic gain. For sorghum, this includes Dr David Jordan and Dr Andrew Borrell from Agri-Science Queensland; and for wheat, Dr Scott Chapman and Dr Fernanda Dreccer from CSIRO Plant Industry, along with Dr Jack Christopher and Dr Karine Chenu from Agri-Science Queensland.
There is also a decade-long collaboration with Pioneer HiBred International, whose scientists are adapting the Australian GxExM simulation technology to its US maize-breeding program through an Australian Research Council (ARC) linkage grant.
Although the ultimate aim is greater water productivity, Professor Hammer is not exclusively targeting genetic tolerance.
“The focus is more on an ecological level,” he says. “We want to exploit knowledge about the soil type, the rainfall sequence during the season, where the moisture is in the soil profile, when the crop is going in, how it is being managed, and how the plant is responding. The goal is to identify ways to use water most productively in various circumstances and to give farmers options for managing risk.”
To work at this level, Professor Hammer starts by profiling the environment where growers are experiencing difficulties. Rather than collate information about rainfall or temperature into an ‘environmental index’, however, he prefers a biological index based on the plant’s ‘perception’ of its growing environment.
In practice that means feeding 100 years of weather data into a computer, adding information about soils and translating the lot into measures of the plant’s perceived stress levels at each point in its lifecycle.
Field tests check simulations
Seasonal variation can then be classified according to crop productivity. At this point, the computers are sidelined while the scientists run detailed field experiments to measure the impact of variation in nitrogen, temperature, light, water, crop management, and genetics on plant growth and development.
“We try to measure enough factors so that we can figure out the processes that regulate how plants develop root systems and canopies, how they capture light and water, how they allocate nitrogen,” Professor Hammer says. “This is in contrast to just asking how much plants yield in different situations.”
As the data piles up well beyond the ability of humans to spot connections, Professor Hammer captures these dynamic interactions in mathematical algorithms, or computer code. The result is software that behaves like a plant right down to a reliance on environmental conditions.
A similar approach allows the impact of management practices to be included into the software. This makes it possible to model the impact of on-farm decisions such as planting time, row width, sowing density and soil condition. By collecting information on lines with known genes or pedigrees, then genetics too can be included in the modelling.
The pay-off is the ability to scan across entire farming systems for ways to make better use of available water. The simulations monitor impacts not just on yield; they also track the way roots penetrate the soil to access water and nitrogen, the time required to flower, leaf area development and biomass accumulation.
“This is a deliberate attempt to study factors that influence yield but that do not reside exclusively within a genome or a management system but arise as interactions between them,” Professor Hammer says.
“We then test those predictions in all sorts of ways to make sure they are robust. So it’s a gradually evolving spiral of knowledge that constantly bounces back between field work and simulations.”
The predictive powers of this tool are helping breeders and agronomists to explore in days vast numbers of combinations that would not be possible via conventional field research and on-farm experimentation. The technology is especially useful for identifying:
- plant traits and management options that best serve in different environments;
- trade-offs between high yield (in good seasons) and resilience (in bad years) that can help farmers manage risks and costs; and
- opportunities to extend cropping zones into new environments.
Screening focus increased
Breeders are among the earliest adopters of GxExM modelling technology – especially with regards to screening sorghum, bread wheat and maize populations for performance under stress. Typically breeders can identify high-performing lines but they rarely clarify the reason for the advantage.
“The GxExM models help us understand which traits are providing the adaptation advantage, even if on the surface the traits do not appear related to drought tolerance,” Professor Hammer says. “We can then use our software to explore how those traits are likely to behave in different seasons and growing regions, identify simple ways to select them during breeding, and support discovery of associated genomic regions.”
Pre-breeders are further using the technology in studies of sorghum’s ‘stay green’ water productivity trait and in reducing tiller numbers in northern adapted wheat to better conserve soil-stored moisture.
A rudimentary ability to model root architecture is seeing the technology extend to underground traits, where Professor Hammer has found that variation in the angle of a seedling’s first few roots can serve as a water-productivity marker in sorghum and bread wheat in certain environments.
“The root models are still a bit crude because we don’t have good data yet, but we can certainly model how root systems penetrate into the soil profile and access water and nitrogen,” he says. The software is also finding applications in genome-based approaches to crop improvement. In this mode, valuable traits are associated with discrete regions of the genome (known as QTLs or quantitative trait loci).
“Scientists trained in the molecular world of genomics do not necessarily have a good grasp of the plant as it exists within its environment,” Professor Hammer says. “But the modelling technology provides a way to connect the clever bits of genome data to something that can predict what it means in the paddock.”
With the technology now capable of simulating entire breeding programs – both conventional and molecular – scientists, farmers and agronomists are in a position to preview thousands of combinations before selecting the most promising combinations of traits, environments and farming systems. As far as Professor Hammer is concerned that means that there is no such thing as ‘good’ and ‘bad’ germplasm, just germplasm that is useful in different contexts. And it has always been like that, he says, but it is especially true for complex traits.
“You always get trade-offs in different situations. There are probably some things that are broadly useful. But this notion of specific adaptation using the GxExM concept is where you can make significant advances.”
GRDC Research Code UQ00042
More information: Professor Graeme Hammer,
Computer solutions to climate risks
Professor Hammer describes the effects of climate variability as a real issue for Australian grain growers, more so than in other parts of the world because of the El Niño-Southern Oscillation (ENSO) cycle.
“Before I started working with breeders on water stress, I spent time on the climate variability question,” Professor Hammer says. “Modelling technology allowed us to test various adaptation strategies and generate the information needed by farmers – the risk associated with different investment strategies.”
The same software platform is now being used to understand the impact on farmers of various climate change scenarios. The trick, he says, is to take the 100-year weather record – that already contains variability associated with the ENSO cycle – and realistically adjust it to incorporate climate trends, especially changes in temperature and rainfall.
“To project impacts of climate on a farming system into the future, we need weather trajectories that contain both the trends and the information on year-to-year variation,” Professor Hammer says.
At CSIRO Sustainable Ecosystems and AgriScience Queensland, Steven Crimp and Peter deVoil are well on their way to producing climate models for use in crop modelling.
Rather that take the simple option of just adjusting the existing weather record by changes in average temperature and rainfall, a more realistic (and more complex) option is being pursued. Mr Crimp says that by using Bayesian statistics, they can get away from simple-minded use of averages and instead use real-world climate trends to infer probabilities.
“The statistical approach we use looks at historical trends in the weather data and which of those are robust enough to actually project forward,” Mr Crimp says. “We then also include information produced by mainstream climate models so we can make sense of how variability and extremes could change in the future. At this stage this approach is probably the most robust way to produce the daily future projections needed to study impacts on crop production.”
Mr Crimp also works with agronomists and farmers, directly exploring management options that could work in a climate-changed world.
In a pilot project in NSW, regional climate projections were generated and farmers were asked to propose suitable adaptive practices. The effects from these adaptations were then modelled using the Agricultural Production Systems Simulator (APSIM) – the same basic modelling engine used in Professor Hammer’s GxExM simulations.
“We could then try to identify management interventions that provide resilience to extreme conditions and to small changes in climate,” Mr Crimp says. “One of the findings so far is that adaptation is very regionalised. So what works well in one location may not work at all in another.”
An example of regionalised adaptation was the finding that on the eastern seaboard, where fertiliser is applied mostly at sowing, modifying the application of inputs to seasonal conditions can be an effective adaptation to the climate variability that is occurring now and to the kind projected in a climate-changed future.
“While production levels fell slightly using this risk management option, the overall gross margins were much better for farmers because of the lower use of nitrogen in poor years,” Mr Crimp says.
With funding from the GRDC, a national version of the NSW study is being rolled out to better understand the adaptive capacity of crop production systems across Australia and to assess management intervention for their effectiveness at reducing negative effects.
“I think the greatest challenge we are facing with climate change research is providing information in the context of all the other issues that farmers and the primary industries face in this country, including commodity prices and input costs,” Mr Crimp says. “So for me, the issue of climate change is embedded in risk management – it is just an additional consideration, another layer of the risk portfolio.”
GRDC Research Code CSA00022
More information: Steven Crimp, email; www.grdc.com.au/CSA00022
GRDC Project Code UQ00042
Region National, North
Was this page helpful?