Vertical farming is gaining momentum as a sustainable solution for urban agriculture but one of the biggest challenges for growers is determining which plant varieties perform best in controlled environments. Traditional plant breeding methods require extensive trials, time, and financial resources, but computational modeling offers a creative way to predict optimal plant architectures before a single seed is planted.
Michele Butturini, a researcher in the chair group of Horticulture and Product Physiology at Wageningen University & Research (WUR), is pioneering the use of Functional-Structural Plant Modeling (FSPM) to identify the best-performing dwarf tomato traits to help breeders optimize crops for vertical farming. His work aims to provide breeders and industry professionals with data-driven insights into the most productive plant architectures for controlled environments.
"It can be difficult to determine what makes a plant better or worse for vertical farming," Butturini explains. "Trials take time and money, but modeling provides an approximation that allows us to narrow down the most promising traits before committing to large-scale experiments."
© Yongran Ji Michele Butturini
How Functional-Structural Plant Modeling (FSPM) works
At the core of Butturini's research is 3D plant modeling, which allows him to simulate plant growth in a virtual environment. By using the open-source software, GroIMP, his model simulates:
- Light interception – Tracking how photons interact with plant structures.
- Photosynthesis rates – Estimating energy conversion efficiency.
- Biomass allocation – Predicting how resources are distributed within the plant.
- Yield potential – Identifying the most efficient plant traits for maximum productivity.
"With this model, I can test different plant morphologies and see which ones lead to the highest yield," Butturini says. "Instead of relying only on costly and time-consuming physical trials, we get a data-driven way to refine breeding targets much faster."
Why vertical farms need custom-bred varieties
Unlike open-field or greenhouse agriculture, vertical farms operate in highly controlled settings. Yet, many of the crops currently grown in these systems were initially bred for outdoor environments. According to Butturini, this approach leads to suboptimal plant performance and prevents vertical farms from reaching their full potential.
"We cannot keep using open-field or greenhouse varieties in vertical farms," Butturini says. "These environments are completely different. If we don't start breeding crops specifically for vertical farms, we risk staying suboptimal for too long, and that could limit the long-term success of the industry."
His research aligns with the growing recognition that vertical farming needs its own custom crops, specially bred varieties that prioritize architectural efficiency, light absorption, and optimized energy use over traits that are unnecessary in a controlled environment, like drought resistance.
The role of 3D modeling in future plant breeding
One of the most important insights from Butturini's research is that the best-performing individual plant is not necessarily the best for an entire crop system. His modeling approach considers above-ground competition between plants, helping breeders identify varieties that maximize yield per square meter, rather than just individual plant growth.
"Plants in vertical farms compete for light," Butturini explains. "A plant that grows too large might overshadow others, leading to lower overall yield. By using 3D modeling, we can simulate these interactions and design crops that perform best as a collective."
While his model provides valuable insights, it currently focuses only on plant architecture. Factors such as nutrient delivery, CO₂ levels, and dynamic lighting optimization are not yet incorporated but Butturini does see potential for expanding his work in these areas.
Butturini's 3D model
The role of AI in agriculture
While Butturini's current work does not use AI to generate models, he recognizes the potential of machine learning to enhance the analysis of plant architecture. "My model is not machine learning-based," Butturini explains. "However, I am using a machine learning technique to interrogate the 3D model and identify the optimal plant architecture for vertical farming." This approach exemplifies a hybrid modeling strategy, where a process-based 3D model (a 'white box' model) is used in synergy with a machine learning algorithm (a 'black box' model) to improve predictions and decision-making.
The machine learning component was developed as part of the master's thesis of Marloes Rodewijk, under the supervision of Butturini, and Dr Katarina Smolenova, a researcher specializing in 3D modeling in the Horticulture and Product Physiology group at WUR. Rodewijk's outstanding work earned her a grade of 9 for her thesis, highlighting the effectiveness of combining machine learning and traditional crop modeling to refine plant breeding for controlled environments.
Butturini is a strong advocate for combining process-based crop models with AI and machine learning. He believes that while artificial intelligence is often overhyped, hybrid approaches that integrate AI with traditional crop models hold the most promise. "AI is powerful, but in agriculture, we're dealing with sparse, inconsistent data," he says. "That makes it difficult to apply AI in the way that works for other industries. Instead of replacing crop models, I see AI as a tool to enhance them."
He credits professor Ioannis Athanasiadis, Chair of Artificial Intelligence at WUR, as an inspiration in this field. "His work highlights the importance of a hybrid approach, using AI in combination with structured crop models rather than as a standalone solution."
Paving the way for optimized vertical farming
As the vertical farming industry evolves, the need for specialized plant varieties will increase. Butturini's work offers a practical, data-driven approach to help breeders develop crops specifically tailored to indoor farming environments.
"The industry's short-term focus is on proving profitability, but long-term success will depend on breeding crops that are optimized for vertical farming," he says. "It's not just about technology and design, we need to focus on the biology of the plants themselves."
His research serves as a proof of concept that computational models can play a key role in the future of agriculture. Although still in its early stages, 3D modeling could change how vertical farms select and optimize their crops, ensuring greater efficiency, sustainability and long-term viability.
The FSPM Community
For those interested in learning more about Functional-Structural Plant Modeling, Butturini and fellow researchers have established a free, open community where scientists and industry professionals can exchange knowledge and resources.
Explore the FSPM community here: https://fspm.discourse.group/
For more information: 
© WURWageningen University & Research
Michele Butturini, Researcher
michele.butturini@wur.nl
www.wur.nl