What is Miscanthus AI?
Aberystwyth University, the University of Lincoln and the University of Southampton have linked up to work on a project together called Miscanthus AI. The aims of the project are to:
Use machine learning models to map the genotype of Miscanthus through to the phenotype
Use adaptive and explainable artificial intelligence (AI) to integrate human reasoning and decision making into intelligent plant breeding
What does this mean for UK agriculture?
By using artificial intelligence to select plants that will perform optimally, plant breeding cycles can be reduced, and the end product produced quicker. Traditionally, highly skilled plant breeders select plants with desirable traits or characteristics and breed them into the next generation. This is not only time-consuming but also relies on what is known as the ‘breeder’s eye’. Miscanthus AI aims to mimic the breeder’s eye.
In the future, AI could predict the plant phenotype before a single seed is sown.
So, how is AI different from genomic selection?
Genetics has come a long way over recent years. Although it can create organisms with the most desired physical characteristics, it doesn’t consider other factors. AI, like human plant breeders, will be able to take into account economic benefits (return on investment, competitive products etc) and environmental-management interactions. These, along with genetic information, will likely produce the most profitable and resilient crops.
Miscanthus AI uses robotics and 3D imaging to collect phenotyping data combining this with extensive genomic data collected on Miscanthus previously.
Why is the project using Miscanthus?
The project is using Miscanthus as a model to test its advances in artificial intelligence within plant breeding. But Miscanthus also has the potential to become a huge economic asset.
Plant biomass makes up 9% of the UK’s renewable energy. Most biomass is imported which affects the environmentally friendly benefit of using renewable sources.
There is an opportunity to expand the UK’s biomass production by using Miscanthus. It is a fast-growing grass, growing up to 4 metres every year, and has a low global warming potential. Miscanthus can produce up to 25 tonnes per hectare, more than other bioenergy crops such as short rotation coppice (SRC) and cereal straw.
It sequesters more carbon than annual crops, requires relatively low input and can be grown on poor quality land. It is important that biomass crops do not compete for space with food crops, so the ability to grow Miscanthus on land unsuitable for crops is a huge benefit.
It is harvested post-senescence therefore still provides habitat and shelter for small animals over the typical harvesting periods and winter months. Late harvesting is also beneficial to farmers growing Miscanthus as contractor rates are likely to be lower out of season.
The downside to using Miscanthus is that is takes 2 years to mature. This has been a barrier for improved breeding as breeding cycles are so long. Using AI to predict phenotype can significantly reduce this.
What does this mean for farmers?
If the Miscanthus AI project is successful, then using artificial intelligence in plant breeding could change the way we farm in the future. Using predictions made by machine learning models will produce the most sustainable, resilient crops. It will enable us to adapt to and keep up with the changing and challenging climate.
Selecting the best crops for your soil type, local climate and target profit will take some of the uncertainty out of farming.