High throughput phenotyping of novel root traits for early stage root bulking in cassava using an Aeroponic imaging platform

The tropical root crop cassava (Manihot esculenta) is a staple food for an estimated 800 million people worldwide: more than a tenth of the world's population. A major obstacle reducing cassava's industrial crop potential is its long and variable growth cycle. Selection for early storage root bulking (ESB) - the formation of edible storage roots via the uptake of starch - would both significantly increase the productivity, efficiency and financial stability of small cassava farmers and open the door to more intensive commercial production methods. Early storage root bulking (ESB) in cassava has become important where increased production on available land is necessary, and in semi-arid regions where ESB cultivars that can be relied upon to mature within one rainy cycle are crucial. ESB can increase both economic stability and returns for smallholder farmers and allow cassava's transition to more intensive commercial production systems that supply factories with raw materials during greater portions of the year. To meet these demands, development of ESB varieties that can be harvested in less than 8 months without loss of yield are urgently required.

Image-based plant phenotyping has attracted rapidly-increasing interest in recent years, as the absence of high-throughput, automatic tools capable of providing accurate, quantitative data on plant structure and function has been widely recognised as the new bottleneck to the production of more efficient crops. However, it usually requires expensive, automated image acquisition hardware, and complex software that provides solutions to challenging image analysis problems to be developed and embedded in reliable tools.

In this project, a partnership between the University of Nottingham and the International Center for Tropical Agriculture (CIAT) in Columbia, we propose to develop low-cost growth and imaging facilities to allow for automated imaging of cassava roots as they grow in an aeroponic environment. We will develop and release for free software to analyze these images automatically, and provide measurements for the cassava roots as they grow. This will enable traits such as the volume of the roots to be measured and monitored. This can be used to record plant response to different nutrient and water levels, and also to provide information regarding how different cultivars of cassava respond to these environments. The complete imaging system will be designed to allow easy replication throughout low and middle-income countries (LMICs).

The system in this project will be used to identify the ESB cultivars needed by commercial breeders and small scale farmers of cassava in LMICs.

The tropical root crop cassava (Manihot esculenta) is a staple food for an estimated 800 million people worldwide: more than a tenth of the world's population. Early storage root bulking (ESB) in cassava has become important where increased production on available land is necessary, and in semi-arid regions where ESB cultivars that can be relied upon to mature within one rainy cycle are crucial. To identify such a trait, we propose the development of a low cost aeroponic phenotying platform, fully instrumented with cameras, to enable higher throughput phenotyping in Low- and Middle-Income Countries (LMICs) - specifically at the International Center for Tropical Agriculture (CIAT) in Columbia. Crucial to this development will be the design of novel 3D reconstruction approaches, using an array of low cost cameras located around the phenotyping setup. These computer vision algorithms will model the cassava root growth in 3D, providing measurements for traits such as volume. Such a system will allow cassava growth to be monitored automatically across a range of cultivars and conditions.

The project will be carried out with the following work packages:

WP1: Instrumentation of an aeroponic growth platform
WP2: Recovery of root volume and growth from multiple images
WP3: Establishment & validation of an integrated cassava root phenotyping platform

The final WP will include identification of root traits for enhancing productivity in nutrient limited conditions, and a germplasm screen for ESB traits.

Impact Summary
Central to this project is the development of the low cost cassava phenotyping system and associate image analysis algorithms. The secondary outputs are reliant on this, and provide biological insight into traits for ESB. Below we consider the stakeholders who will benefit from both outputs.

Plant scientists and breeders (Academic and Industry):
There are two direct impacts which will affect this group. First, the development and publication of the low-cost phenotyping system design itself will allow the project to be replicated locally at research groups throughout Colombia, and in LMICs worldwide where cassava is a key crop. Free availability of the algorithms and software will assist uptake. Second, the biological insight gained into identifying traits of ESB cassava may lead directly to improved plant selection, by both breeders and academic researchers.

Computer scientists and Image analysts (Academic and Industry):
We will be developing cutting-edge 3D reconstruction and deep learning approaches using a low cost imaging setup. Whilst in this project we image cassava, the approaches we develop will translate to other aeroponic growth systems, as well as translating beyond the plant domain entirely. The fundamental algorithms we develop will advance 3D reconstruction and measurement approaches in computer science across many different domains.

Computer science PDRA training:
The PDRA on this project will gain valuable interdisciplinary experience, as well as experience with working with partner institutions in LMICs. This will leave them well placed for employment in UK image analysis or phenotyping groups, for example, both within academia and the commercial sector, ready to engage with partners worldwide. It will also further train them in the increasingly-applicable, cutting edge deep learning techniques, which can be applied in other biology-focused research projects.

Farmers (Industry)
Ultimately providing farmers in LMICs with faster-bulking cassava , and traits that assist growth in low water and nutrient conditions will have a clear impact on the economic potential of this crop.

Consumers (Public)
This technology can help drive more productive cassava growth, ultimately leading to more staple food availability to the 800 million consumers of cassava worldwide. This in turn will likely have an economic impact on consumers, as the cost of crop production could be reduced, and some of this saving could be passed on to consumers.

The 'Maker community':
In recent years a new interest has risen in popularity focusing around building hardware at home with low cost components. Targeting this general-public hobbyist-scientist community with our designs will allow people to become involved in projects using similar hardware and computational setups, to increase interest in this multidisciplinary area, mixing engineering, computing and biology technology. This will have an impact on general public awareness of the work, and benefits of such projects.

Schools:
The project is an ideal example of combining computer science and biology, a career path which school children in our experience have very little experience of. A similar design could be re-created in schools using Raspberry Pis (which many schools now use) to image plant growth over time in cross-discipline biology/computer science lessons. It will also raise awareness of the impact science partnerships can have with other countries (especially LMICs in this situation).