Below are is the list of topics currently available under the Digital Agriculture PhD Scholarship fund.
Novel phenomics tools to screen for drought tolerance in wheat
Supervisors: Bob Furbank (ANU) and Fernanda Dreccer (CSIRO).
Drought is a major environmental factor reducing wheat yields across Australia. Drought tolerance is a complex multigenic trait and difficult to assess rapidly in the glasshouse or the field. This project seeks to develop new methods based on the hyperspectral reflectance, the spectrum light reflected from the wheat leaves and canopy, by using machine learning to build predictive algorithms for selection of drought tolerant wheat in the field. These tools can be used both to find superior germplasm and to understand the physiological basis of drought tolerance mechanisms at key stages of crop development.
Digital soils and landscape
Supervisors: Luigi Renzullo (ANU) and Brendan Malone (CSIRO).
The interaction between static soil properties and dynamic climate and moisture data layers are foundational to agricultural productivity and sustainable landscape management. This PhD will explore ways in which to combine, fuse, assimilate and outscale existing and new sensing technologies across a range of scales. The project will leverage soil, drone, satellite and proximal digital data to build a powerful inference utility that can be applied for a diverse array of observation, prediction and forecasting contexts related to agricultural production systems. This could include soil moisture sensing, prediction and forecasting systems. Work would entail data layer integration of static environmental datasets and dynamic time-series datasets sourced from climate and remote and proximal sensing platforms. The project would require the use and development of terrain analytics, digital soil mapping and related geospatial analysis work pipelines.
Nutrient balance and Carbon trajectories
Supervisors: Craig Strong (ANU) and John Kirkegaard (CSIRO).
Crop nutrition in agriculture is a complex interplay of crop physiology and timing and availability of nutrients, with both demand and supply affected by climate, soil, and management. Its study requires consideration of a multi-layered system in time and space. We suspect that insufficient crop nutrition (nitrogen in particular) is the biggest cause of the current “yield gap” in dryland agriculture (difference between actual and potential yield). At the same time, we know that soil “nutrient mining” through inadequate nutrition drives the loss of carbon (organic matter) in soil. Fine tuning and de-risking our nutrient management to maintain a nutrient balance at the systems level is key to dealing with these two critical but linked national goals. More deeply understanding the nutrient balance will have impacts not only towards crop productivity, but also carbon accounting and environmental impacts on soils, water, and landscapes.
Multivariate and machine learning statistical tools for linking crop multi-omics data
Supervisors: Bob Furbank (ANU) and Shannon Dillon (CSIRO).
Advances in crop genomics and phenomics mean that it is becoming cheaper and easier to gather large sets of data from the genome, transcriptome, metabolome, proteome and physiological phenome of diverse germplasm. The power of these datasets to understand the genetic architecture of important crop traits is currently limited by analytical tools to drill through these diverse data layers and turn data into knowledge. This project seeks to build such multivariate statistical and machine learning tools using wheat datasets as a use case.
Detection and diagnosis of crop pathogens
Supervisors: Benjamin Schwessinger (ANU) and Melania Figueroa (CSIRO).
Crop pathogens cause significant yield losses in a wide variety of crops around the globe. The emergence and proliferation of plant diseases has been facilitated by increased human activities associated with agriculture. This includes changes to plant communities and native ecosystems as well as the establishment of dense monocultures. The tracking and monitoring of plant pathogens is a major challenge to ensure food production and quality. Real-time detection and rapid pathogen identification are essential components to safeguard agricultural and natural (eco) systems. This leads to increased biosecurity requirements, which is key to Australian agricultural exports. Furthermore, solutions for more sustainable agriculture demand precision crop protection strategies that effectively reduce the use of pesticides and fungicides to decrease ecological impacts. Therefore, accurate and reliable in-field pathogen surveillance and diagnosis is essential for tailored and rapid responses and disease prevention. Recent breakthroughs in sensor-based methods through sequencing and phenotyping technologies combined with advances in Machine Learning and Artificial Intelligence offer opportunities to develop robust and mobile pathogen detection techniques as well as disease forecasting models. However, more research involving data acquisition, integration, analysis, and prediction is needed to capitalise on these opportunities. Through this CSIRO-ANU Digital Agriculture Supplementary Scholarship, a PhD student will pursue the development of point-of care diagnostic methods for plant pathogen detection in major crops while working closely with experts in disease and bioinformatics across CSIRO, Data61 and ANU.
Agricultural data fusion – earth observation
Supervisors: Luigi Renzullo (ANU) and Roger Lawes (CSIRO).
This project will investigate ways to utilise rapid advances in earth observation from a range of satellite data sources to solve science challenges in agricultural and landscape decision support. Increasingly, deep learning algorithms are being deployed in soil health, production estimates under climate risks, water management and natural capital accounting.
This work will advance data science for improved management of modern data streams and their availability for application of novel computer learning approaches.
Large scale unsupervised sequence learning
Supervisors: Lexing Xie (ANU) and Christian Walder (CSIRO).
Recent advances in deep learning models for natural language processing have resulted in large scale approaches for learning from sequence data with few labels. Identifying suitable approaches for genomic and proteomic sequence data would allow novel scalable methods to be used for improved bioprediction.
Sensing the next step-change in water productivity in dryland agriculture
Supervisors: Craig Strong (ANU) and John Kirkegaard (CSIRO).
Better knowledge of soil water content at paddock and farm scale underpins improvements to almost all decisions facing agriculture in an increasingly variable climate. If, when and what crops to sow, how much fertiliser or other inputs to use, whether to cut crops for hay or grow on for grain in drought and even yield predictions for marketing purposes are all decisions that can be influenced by a more accurate measurement of the water available to plants in the soil. Historically, soil water availability was estimated from rainfall data using simple calculators and local knowledge of the soil, but new soil water sensing technology is making it increasingly possible to measure soil water in real time. Wireless soil water sensors that can monitor soil water in real time to 2m depth are now a reality, but the value of that information in the context of variable seasonal outlooks and processes is uncertain. In terms of de-risking management and input decisions, knowing how much water is in the soil and available to plants together with seasonal forecasts could be transformational. De-risking management decisions is the key to improve the productivity and water-use efficiency of our dryland systems and reduce the environmental impacts on soils, water, and landscapes.
The PhD project will use farming systems analysis and modelling to better understand the benefits of real-time knowledge of plant-available water in on-farm decision making at the paddock and farm scale to improve the water productivity of dryland farms. The project will develop an in-depth understanding of soil water use in cropping systems, and the impact of management decisions on water-use efficiency drawing on data from CSIRO’s Boorowa Agricultural Research Station (BARS) and other long-term cropping trials for verification and to augment the simulated data.
Plant Architecture Mapping in the Lab and in the Field
Supervisors: Tim Brown (ANU) and Matt Adcock (CSIRO).
In digital agriculture, we now have the ability to scan plants in the lab and field at millimeter resolution. However, mapping these data to real plant architecture is still a largely unsolved problem. Pose estimation, deformable models, GANs and Neural Radiance Fields have yet to be applied widely to tackle this challenge. This project will explore ways to extend and adapt techniques that have been effective in controlled conditions to the challenges of plant architecture acquisition out in the field.