When planning a plant experiment, one of the most important decisions is made before any data are collected: what should we measure? The value of an experiment depends not only on how much data are collected, but also on selecting the types of measurements that are most informative for addressing the research question.
Mathematical models connect plant biology with data. By representing biological processes as equations, a model allows us to investigate plant behaviour and compare predictions with experimental observations. Mathematical analysis can also help identify which measurements would be most informative before an experiment begins.
During my PhD at Queensland University of Technology (QUT), I explored this connection using mathematical analysis, computer simulations, and statistical inference to examine and improve established whole plant models of growth. I was particularly interested in how plants allocate carbon and nitrogen between their shoots and roots, and in which experimental measurements allow us to understand these processes more reliably.
Through mathematical analysis and extensive simulations, we found that, under some conditions, an established plant model could produce biologically unrealistic behaviour, including highly irregular oscillations. We investigated why these behaviours occurred and proposed a modification that improved the model’s stability. This part of my research was published in the Journal of Theoretical Biology in 2025.
Improving a model, however, is only one part of the challenge. We must also determine which measurements are most useful for estimating its parameters and testing its predictions.
For example, total plant mass is relatively easy to measure, but on its own, it may not provide enough information to distinguish between different internal biological processes. Other measurements, such as internal nitrogen concentrations, are more difficult or costly to collect yet can offer much greater insight. In the later stages of my PhD, I used two statistical approaches to investigate how different types of plant data affect what a model can tell us. The first was Bayesian inference, which updates what we know about the parameters, including our uncertainty, as new data are added. The second was an analysis of model sloppiness (parameter identifiability), which reveals which combinations of parameters can be estimated well from the data, and which cannot.
This type of pre-experimental analysis can help researchers identify useful measurements before committing time and resources to collecting them. It can, therefore, support more targeted experimental design, particularly in biological systems where data collection is costly, time intensive, or limited.
My PhD was supported by a scholarship from the Centre, which also enabled me to present my research at conferences in Australia and overseas, attend research retreats and professional development workshops, and connect with researchers across different disciplines. These experiences reinforced the value of communication and collaboration between mathematicians, plant scientists and researchers working with data.
In a data driven world, we often focus on analysing and visualising data after they have been collected. However, even sophisticated analysis and attractive visualisations cannot recover information that was never measured. An important question must therefore come first: are we collecting the right data? My work shows how mathematical analysis helps answer this question before an experiment begins, supporting more informative and targeted experiments while helping to reduce the time, effort and cost involved.
I thank my supervisors, Chief Investigator Kevin Burrage, Postdoctoral Researcher Brodie Lawson, and Matthew Adams from QUT’s School of Mathematical Sciences, for their guidance and support during my PhD.
Ati Rostami
PhD graduate, Queensland University of Technology
READ MORE >
Related links:
Rostami, A., Lawson, B.A.J. and Burrage, K. (2025). An in-depth study of the dynamics of Thornley’s mathematical model in plant biology with a view to an improved model. Journal of Theoretical Biology. doi:10.1016/j.jtbi.2025.112071.
