This paper explores three case studies that demonstrate how digital twins and machine learning can be applied to smart farming in space and extreme environments. In the first case study, we tested the performance of commercial machine learning models specifically using Azure Machine Learning to predict plant growth and assess their performance. Lettuce growth was observed in two distinct environments, including one on a mountain with extreme weather conditions (acting as the physical twin-sender) and the other in a controlled growth chamber that replicated those conditions (serving as the physical twin-receiver). The mountain setting functioned as a “space surrogate,” while the growth chamber represented Earth-like conditions. Lettuce was cultivated using both soil-based pots and hydroponic solutions to assess different growing methods. In the second study, the growth of lettuce under different stresses in six climate zones on Earth was considered. This study clearly demonstrated the practical application of machine learning algorithms for predicting lettuce growth under a range of climate conditions from six different global regions. Particularly, linear regression, ridge regression, lasso regression, polynomial regression, random forest regression, and boosted decision tree regression modeled by Python were used for predicting crop yield. Models consistently delivered the most accurate and reliable biomass predictions across all environmental scenarios. The results show that real-time experimental data, when paired with well-trained machine learning models, can successfully capture complex plant growth dynamics in varying climates. This work represents a meaningful step toward making digital twin technology a practical tool for climate-resilient agricultural applications. Finally, this paper discusses the development of a digital twin using XMPro software for the growth of lettuce under abiotic stress conditions (temperature, light, water, oxygen in water) and different cultivation methods (simulated regolith as lunar soil, hydroponics).
Acknowledgements
The authors thank The ARC Centre of Excellence “Plants for Space” (P4S) with the grant number CE230100015.
