Rothamsted Research has been awarded £4m by BBSRC to upgrade its capabilities in high performance computing (HPC). The funds will help turbo-charge the institute’s ongoing digital transformation to place it at the forefront of digital agriculture research in the UK.

“The new enhanced HPC infrastructure will enable us to build a solid foundation of diverse AI (artificial intelligence) and ML (machine learning) capabilities. This will enable faster and more cost-effective data analysis as well as increased throughput.” said Professor Chris Baker, Rothamsted’s Director of Intelligent Data Ecosystems.

The new investment will help Rothamsted build a stronger foundation for data-driven science and leverage new advances in AI. It will also help the institute to become more sustainable by optimizing energy efficiency and reducing lifecycle operational costs.

“The research needed to engineer the UK’s transition to more sustainable farming systems is already generating millions of new data points from our long-term monitoring programmes and lab studies,” said Baker. “Translating this data into accurate predictions is the journey we are on. AI is a crucial part of this and the new HPC will provide the requisite compute power. We urgently need to be able to offer the farming community reliable and timely solutions to the sustainability challenges they face. We won’t get there without our enhanced HPC capability.”           

The HPC upgrade will bring significant advancements to the Rothamsted Digital Twins pilot studies at the North Wyke Farm Platform (NWFP) in Devon. Focussing primarily on the impacts of livestock on grassland, the platform is Britain’s most instrumented and measured farm – everything in and out of the system is monitored. As a result, it has created vast open datasets of over 80 million measurements which play a crucial role in identifying sustainable land management strategies for improved food production, nutrient transfer and mitigations to climate change.

“A good example of a digital twin is the weather forecast we see on TV every day. It’s based on a simulation using billions of data points recorded over time and space, using advanced models that compute multiple scenarios. A farm digital twin can generate predictions of crop yields or greenhouse gas emissions using data from the farm, weather stations, and satellite images and allow us to ask what-if questions. This will help us advise farmers more reliably of the viability of a given crop in a changing climate. Using high performance computing we will be in a position to provide better estimates of the uncertainty in these predictions.” said Baker. 

The improved memory and processing capabilities of the HPC will also boost studies in soil metagenomics – probing the complexity of the soil microbiome to find insights to drive new ways of cultivating the land. There is huge potential to enhance the microbiome’s ecological role in land use sustainability, and research is opening up promising avenues for new fertilizer treatments, disease control and improvements to soil resilience.

Currently, Rothamsted holds 257 soil metagenomes, with an average of 343 million reads per sample, resulting in a staggering 88.2 billion sequences. This vast dataset is equivalent to approximately 8,000 human genomes, highlighting the immense complexity of these metagenomes, which harbour around 10,000 to 50,000 microbial species per gram of soil.

“The goal is to generate metagenome-assembled genomes (MAGs) from unculturable bacterial populations obtained from long-term field experiments and various other treatments. When complete our new Rothamsted Microbiome Atlas will provide invaluable insights into the resilience of complex soil systems,” said Baker.

The new system is scheduled to be up and running by March 2025.