Challenges in modelling photosynthesis as a function of spectral photon flux density

To establish a sustainable and efficient crop cultivation, independent of season and climate, artificial lighting remains a critical factor. Artificial lighting can incur a significant environmental and financial cost. Thus, a key challenge is reducing power consumption for lighting without compromising crop quality and yield. In other terms, lighting needs to be optimized with regard to net photosynthetic activity, which is directly linked to crop quality and yield. Determining an optimal lighting strategy is not trivial, as the photosynthetic process and its influencing factors have not been fully understood yet. Furthermore, interaction between the influencing factors is not only possible, but can occur non-linearly. This is the reason why using data-driven approaches and machine learning to model photosynthetic behaviour as an abstract black-box has become one of many focal points in photosynthesis research. Currently, studying photosynthesis requires the examination of the net CO2-assimilation-rate, which is labour- as well as resource-intensive. The aim of this work is to propose a framework for modelling the photosynthetic activity of C3- and C4-plants as a function of the spectral photon flux density, with the goal of eliminating the assimilation-rate measurement altogether. Challenges arising from the biochemical complexities are highlighted and possible workarounds determined. A special focus is placed on data pre-processing and augmentation steps, as well as proposed machine learning approaches.