Functional data analysis-based yield modeling in year-round crop cultivation

• Published in: Horticulture research 2024-07, Vol.11 (7)
• Main Authors: Matsui, Hidetoshi, Mochida, Keiichi
• Format: Article
• Language: English
• Subjects: Method
• ISSN: 2052-7276; 2662-6810; 2052-7276
• DOI: 10.1093/hr/uhae144

Crop yield prediction is essential for effective agricultural management. We introduce a methodology for modeling the relationship between environmental parameters and crop yield in longitudinal crop cultivation, exemplified by strawberry and tomato production based on year-round cultivation. Employing functional data analysis (FDA), we developed a model to assess the impact of these factors on crop yield, particularly in the face of environmental fluctuation. Specifically, we demonstrated that a varying-coefficient functional regression model (VCFRM) is utilized to analyze time-series data, enabling to visualize seasonal shifts and the dynamic interplay between environmental conditions such as solar radiation and temperature and crop yield. The interpretability of our FDA-based model yields insights for optimizing growth parameters, thereby augmenting resource efficiency and sustainability. Our results demonstrate the feasibility of VCFRM-based yield modeling, offering strategies for stable, efficient crop production, pivotal in addressing the challenges of climate adaptability in plant factory-based horticulture.