Big data gathered underground helps scientists develop crops of the future
Crops of the future will need to be more resilient in the face of climate change. A large research project on plant roots at UCPH deploys big data and custom software to analyze roots in their underground state. While the technology is nothing remarkable, the collaborative nature of the project is reflective of new thinking in the field of agronomy.
600 five-meter-long transparent tubes slant into the ground at the RadiMax root research facility in Taastrup, Denmark. Through them, multispectral cameras capture images of the root systems from various crops.
Thereafter, University of Copenhagen root project researchers scrutinize the subterranean images to identify the more robust crop varieties with deeper roots – these are the ones best equipped to withstand drought.
Root imagery used to be analyzed manually. However, a fruitful collaboration between a plant researcher, an associate professor of computer science and an engineer has led to a custom built software tool that can decode images and translate them into systematic root length measurements for individual images.
“This saves us very, very, very many long hours of work, and makes it possible to conduct root analyses that would otherwise be overly time consuming via manual analysis,” according to Simon Fiil Svane, a PhD student taking part in the root analysis project.
"While there is nothing revolutionary with the technology that we are using for image analysis, it is definitely a revolution in the field of root research, where we have quite simply lacked the tools required to analyze big data."
Software finds genetic variation among roots
The aim of the root research facility is to find plant varieties with the best developed root systems for crops like barley, wheat, rye and ryegrass. These varieties can tap and utilize nutritional substances and water available in deeper layers of soil, which is a great advantage, particularly during years of drought like the current one.
Thus far, researchers have successfully run image analyses on five datasets of 30,000 images each. Software developed by the company Videometer can distinguish differences in pixel reflections from different wavelengths. Each pixel is assigned a probability of whether it belongs to a living root, or soil and dead roots.
Besides being able to save time, the machine never tires.
"After manually pouring over images, day after day, one can get sloppy and change their opinion about what looks like a living root. Our algorithm doesn’t tire or get confused, it decodes uniformly throughout the entire process," says Erik Dam, an associate professor of computer science at the University of Copenhagen’s Data Science Lab.
A challenge for agronomy and biology
Simon Fiil Svane believes that it is generally a challenge to get data analyzed for agronomy and biology research:
"Getting technology to collect raw data is no problem. The problem is that it often ends up on a server," he says, explaining that traditional methods are often sufficient in smaller research projects, but that the RadiMax project needs to analyze large amounts of data for external partners including large companies. This requires the systematization of data analysis.
Associate Professor in computer science Erik Dam, Engineer Jon Nielsen and PhD Simon Fiil Svane have developed a program in MatLab that accentuates linear structures (living roots) while cleaning image noise before root length is calculated.
RadiMax - the project is headed by Professor Kristian Thorup-Kristensen
Read more about root research and UCPH’s Data Science Lab