EDITORIAL article
Front. Plant Sci.
Sec. Technical Advances in Plant Science
Volume 16 - 2025 | doi: 10.3389/fpls.2025.1626622
This article is part of the Research TopicLeveraging Phenotyping and Crop Modeling in Smart AgricultureView all 29 articles
Editorial: Leveraging Phenotyping and Crop Modeling in Smart Agriculture
Provisionally accepted- 1Jiangsu Academy of Agricultural Sciences Wuxi Branch, Wuxi, China
- 2Nanjing Agricultural University, Nanjing, China
- 3Aarhus University, Aarhus, China
- 4China Agricultural University, Beijing, China
Select one of your emails
You have multiple emails registered with Frontiers:
Notify me on publication
Please enter your email address:
If you already have an account, please login
You don't have a Frontiers account ? You can register here
In recent years, the agricultural sector has witnessed a significant transformation driven by the integration of sensing technologies, big data analytics, and artificial intelligence (Ahmed and Shakoor, 2025). Cutting-edge innovations, notably high-throughput phenotyping and crop modeling, have fundamentally altered our understanding and management of crop systems (Keating and Thorburn, 2018;Yang et al., 2020). In many cases, phenotyping and modeling are closely intertwined: phenotyping provides accurate characterization of plant traits, forming the basis for reliable crop models, while modeling elucidates interactions among phenotypes, genotypes, and the environment, and enables prediction of phenotypic outcomes (Yu et al., 2023;Zhang et al., 2023b). Despite their natural synergy, phenotyping and modeling are still frequently treated as separate domains, limiting their full potential. This research topic aims to close that gap by promoting the development of integrated phenotyping-modeling frameworks to advance smart agriculture. The following sections provide a categorized overview of the contributions to this research topic, highlighting key findings and identifying future directions for this rapidly advancing field.Crop phenotyping, which plays a vital role in gene function analysis, plant breeding, and smart agriculture, can be broadly categorized based on the traits measured.Morphological and structural traits include leaf length, leaf width, leaf area, and leaf angle, while physiological and biological traits encompass chlorophyll content, nitrogen levels, transpiration, and photosynthetic parameters.2D imaging combined with machine vision remains the most widely adopted technique for acquiring plant morphological and structural phenotypes. In this topic, a range of studies have explored deep learning-based approaches tailored for specific plant phenotyping applications, with a particular focus on refining model architectures and technical strategies to enhance detection accuracy, computational efficiency, and adaptability to complex field conditions. Among them, semantic segmentation A region-growing algorithm was used for stem and leaf segmentation, though substantial leaf overlap during the tillering, jointing, and booting stages made the process particularly challenging. Plant height, convex hull volume, plant surface area, and crown area were extracted, enabling a detailed analysis of dynamic changes in wheat throughout its growth cycle. In recent years, ultra-low-altitude UAV-based crosscircling oblique imaging has become a more efficient and cost-effective approach for in-field 3D reconstruction (Fei et al., 2025;Sun et al., 2024). Unlike indoor multi-view imaging systems, 3D phenotyping conducted directly in the field more accurately reflects real-world agricultural conditions and population-level dynamics. and the Nitrogen Balance Index (NBI), measured by a Dualex sensor, alongside machine learning models for nitrogen status assessment. Data from 15 rice varieties under varying nitrogen rates showed chlorophyll saturation at high nitrogen levels, while Flav and NBI remained reliable. Random Forest and Extreme Gradient Boosting achieved high prediction accuracy, with SHAP analysis identifying NBI and Flav from the top two leaves as critical predictors. In recent years, these technologies have been widely applied to precision farmland management. For example, on farms in Brazil, Castilho Silva et al. (2025) used UAV-based multispectral remote sensing to monitor leaf nitrogen content in maize and applied variable-rate fertilization accordingly. Compared to conventional methods, this approach reduced nitrogen input by 6.6% to 35% without compromising yield.Phenotyping equipment is essential for the precise monitoring of plant traits and environmental growth conditions. Liu et al. developed a portable vegetation canopy reflectance (VCR) sensor for continuous operation throughout the day, featuring optical bands at 710 nm and 870 nm. The sensor was calibrated using an integrating sphere and a solar altitude correction model, with validation against a standard reflectance gray scale board. Field measurements taken at 14 sites using both the VCR sensor and an ASD spectroradiometer showed closely aligned reflectance values. In Bermuda grass measurements, the intra-day reflectance range narrowed and the coefficient of variation decreased after solar altitude correction, demonstrating the sensor's effectiveness for precise vegetation monitoring. Compared to remote sensing, recent developments in flexible sensors enable direct, continuous, and high-resolution monitoring of plant physiological traits and environmental conditions (Zhang et al., 2024). These innovative sensing technologies are poised to significantly enhance phenotyping applications.While various models for the direct extraction or inversion of crop phenotypes have been explored in the crop phenotyping section, crop modeling in this context specifically refers to growth modeling designed to predict crop development and growth. Depending on the approach, crop growth models may be data-based, incorporating machine learning techniques, or mechanistic, based on process-based simulations of crop physiological processes (Maestrini et al., 2022). In this topic, process-based models are limited, with more researchers focusing on simpler modeling approaches. The barrier-free fruit selection algorithm identifies the largest, non-occluded fruit as the optimal target. This approach effectively detects and locates barrier-free fruits, providing a reliable solution for harvesting robots, applicable to other fruits and vegetables as well.In conclusion, we propose an integrated framework that links plant phenotype, genotype, and environment (Fig. 1), aiming to better synthesize current research efforts.Environmental parameters are commonly obtained via in-situ sensing, where sensors capture electrical signals (e.g., capacitance, resistance) and convert them into quantitative data such as air temperature, humidity, atmospheric pressure, photosynthetically active radiation, and soil temperature and moisture. These parameters facilitate the development of microclimate models, which can be further coupled with other simulation models. Phenotypic information is generally acquired through remote sensing and 3D reconstruction. Multispectral or hyperspectral imagery is processed through feature extraction and inversion to retrieve physiological and biochemical traits, while RGB imagery enables extraction of morphological and structural features at 2D level. Additionally, 3D point clouds derived from LiDAR or multi-view image reconstruction are processed through segmentation and surface modeling to obtain 3D structural traits. Therefore, functional and structural models are established through system analysis and dynamic modeling based on these phenotypes.In recent years, such models have been widely applied to investigate the impacts of climate change on crop productivity, identify potential yield gaps, and explore targeted improvement pathways (Gavasso-Rita et al., 2023). On the genetic level, reference genomes from de novo sequencing and genomic variations from resequencing support the development of genotype-based models (Zhang et al., 2023a).The integration of big data and artificial intelligence further enables hybrid modeling approaches-such as knowledge-guided machine learning (KGML) (Li et al., 2025) and improved phenotype-model data assimilation techniques (Jin et al., 2018). KGML leverages mechanistic knowledge of biological processes to guide the learning process, enhancing model interpretability and generalization capacity. Meanwhile, data assimilation techniques dynamically update model states and parameters using real-time phenotypic observations, thereby allowing high-throughput phenotyping data acquired by modern sensing technologies to be effectively integrated into the modeling framework. This unified phenotyping-modeling framework creates a digital twin by linking physical plants to their virtual counterparts, offering a promising pathway to integrate phenotyping with modeling for intelligent breeding and smart agriculture. However, the current framework remains incomplete, as it primarily emphasizes the virtual simulation of physical plants. Achieving a true digital twin requires establishing reverse control mechanisms that enable real-time feedback from the virtual twin to the physical system-a process that depends on further advancements in intelligent agricultural equipment and the seamless integration of agronomic practices with agricultural machinery.
Keywords: phenotyping, modeling, smart agriculture, functional-structural plant models, environment, Genotype
Received: 11 May 2025; Accepted: 21 May 2025.
Copyright: © 2025 Sun, Xiao, Ata-Ul-Karim, Ma and Zhang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence: Wenyu Zhang, Jiangsu Academy of Agricultural Sciences Wuxi Branch, Wuxi, China
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.