The answer is blowing in the wind
The microscopic spores produced by disease-causing fungi can become airborne, eventually spreading an outbreak as far as the wind can blow. By harnessing weather data, researchers are using computers to model the likely spatial pattern of dispersal and in the process, better direct control and eradication efforts.
As part of these efforts, CRC for National Plant Biosecurity PhD candidate Steven
Coventry built a wind tunnel at the University of Adelaide. There he studied the airborne behaviour of infectious spores, targeting ascochyta blight, a disease that devastated the chickpea industry in the 1990s in South Australia (SA), Victoria and New South Wales.
Steven said he placed ascochyta-infected stubble at the head of the tunnel to replicate conditions in the paddock during a blight outbreak. “The wind blows on to the lesions that ooze the spores and they are then blown the length of the wind tunnel,” he said.
“Wind speeds were between one and 4.7 metres per second, which is classed as a breeze. We could then follow how far spores travel by using spore traps. This involved using metal rods with double-sided sticky tape, placed at different lengths of the wind tunnel. Spores were dispersed at least 66 centimetres with the wind speeds available.”
The wind tunnel provided Steven with crucial data needed to model spore dispersal using the commercially available modelling program, Mathematica. To check the accuracy of the resulting simulations, he also physically measured spore movement at chickpea field-trial sites in SA.
“We then used information about weather patterns at our field sites to see how closely the model correlates with the field studies,” he said. “These comparisons were used to calibrate the software and we are now at the point where we can pretty much simulate the pattern of disease spread we see in the field. For one chickpea season in SA, the model achieved 90 per cent correlation with what happened in the field.”
Although developed around a chickpea disease, the software can be applied to model dispersal of many other diseases simply by altering information about the pathogen’s biology. So ultimately, the ascochyta software becomes a template to simulate the spread of fungal disease in general.
“In terms of biosecurity, there are a lot of applications,” he said. “It allows us to focus control and eradication measures along the dispersal pathways that spores are most likely to travel. We can help farmers better predict when they need to use fungicidal spray. If we are dealing with preparations for the incursion of an exotic disease then we can run simulations as a precautionary measure.”
In the research and development pipeline are efforts to expand the software’s capacity, particularly to take account of variable levels of genetic disease-resistance in commercially planted crop varieties.
“When you are modelling diseases there are so many things to take into account,” Steven said. “Natural systems are never simple. I looked at the issue of the susceptibility level of plants as part of my project and did find that dispersal patterns are affected. We have other projects within the CRC looking at other aspects of modelling, spread and dispersal, as well as different surveillance tools. That integrated approach among many researchers helps us cope with the complexity.”
These complementary efforts include a CRC project being conducted by David Savage from the University of Western Australia, who is taking a more mathematically driven approach to modelling disease spread.
He works primarily with ‘dispersal kernels’ – a mathematical gizmo (made up of a probability density function) that can describe the distribution of spores or pollen at different times after a release event.
“Traditionally, you set up spore traps and measure the occurrence of spores in the field and fit your model to this data,” David said. “But a paper was published in 2005 that described a ‘dispersal kernel’ that you can derive using only information about the wind speed and direction and the terminal velocity of the modelled particle.”
The 2005 paper, however, dealt only with a one-dimensional kernel, which measures only the distance of the dispersed spores from the origin by assuming uniform wind directions. So David modified it to generate a two-dimensional kernel. It takes into account the distribution of wind directions and gives a better representation of the dispersal.
He then compares his dispersal pattern with the output from mechanistic models such as CSIRO’s The Air Pollution Model (TAPM), which simulates pollutant dispersal using local meteorological, global terrain and land-use data as well as global synoptic analyses. APM is used by 190 national and international users in 25 countries to model air quality.
“When I compare my dispersal patterns with TAPM, they match up quite well,” David said. “The kernel predicts the general shape of the dispersal surface, which gives a good indication of where the spores are more likely to be. But unlike TAPM, the kernel only takes a few minutes to calculate, rather than a few days.”
Developed to model blackleg disease in canola crops, the kernel is constructed by specifying certain parameters, such as the height at which the spores are released. By altering these parameters, the same tool can be used to model other pathogens or even the pollen that cause allergies for many people.
“The spores themselves are considered essentially the same – tiny inert particles,” he said. “The only bit of biology that matters in the air is that blackleg spores are degraded by ultraviolet light. So the longer the spores remain aloft, the less chance it has of being infectious. But another spore I am interested in modelling, which causes rust disease in wheat, contains a pigment that protects them from UV light.”
Once again, the issue of complexity is seeing David attempt to expand the model’s capacity. In the first instance that means integrating the 2-D dispersal kernel into more sophisticated modelling tools that can, for instance, take into account levels of UV light or differences in crop susceptibility to the modelled pathogen. The idea is to produce more informative models without losing computation speed.
One issue in particular has caught his attention. He explains that virulent strains of fungi that overcome the genetic resistance of crops are often present at low frequencies within the population. The introduction of new resistant cultivars acts as a selection pressure on these strains, causing them to become more prevalent. When this happens, the resistant cultivar breaks down and farmers incur substantial disease control costs or may even lose their crops. However, David thinks it is possible to develop models that predict the breakdown’s spatial pattern.
“The idea is to use these predictions to prolong the duration of resistance genes used in commercial cultivars,” he said. “Ultimately, this model, and others that are being developed, can help Australian biosecurity agencies understand how incursions might unfold, and how they can most effectively use the control measures available to them.”
