UAV-Based Multispectral Phenotyping and Machine-Learning Modeling Reveals Early Canopy Traits as Strong Predictors of Yield and Weed Competitiveness in Oat (Avena sativa L.).
Saved in:
| Title: | UAV-Based Multispectral Phenotyping and Machine-Learning Modeling Reveals Early Canopy Traits as Strong Predictors of Yield and Weed Competitiveness in Oat (Avena sativa L.). |
|---|---|
| Authors: | Benaragama, Dilshan1 (AUTHOR) dilshan.benaragama@umanitoba.ca, Hussain, Mujahid1,2 (AUTHOR), Senetza, Brianna1,2 (AUTHOR), Shirtliffe, Steve2 (AUTHOR), Willenborg, Chris2 (AUTHOR) |
| Source: | Remote Sensing. Apr2026, Vol. 18 Issue 8, p1211. 25p. |
| Subjects: | Multispectral imaging, Machine learning, Cultivars, Crop yields, Weed competition, Plant canopies, Crop improvement |
| Abstract: | Highlights: What are the main findings? UAV-based temporal multispectral and structural traits (NDVI, NDRE, canopy cover, height, and volume) revealed strong differences among oat cultivars in early canopy development and weed-competitive ability. Machine-learning models (RF, GBM, PLS) showed that early-season traits—particularly ground cover and NDRE at 3 WAP—accurately predicted the grain yield in both weed-free and weedy conditions. What are the implications of the main findings? Early canopy traits measured in weed-free conditions can reliably predict cultivar performance under weed pressure, reducing the need for labor-intensive weed trials. The combined UAV + ML framework provides a scalable approach for breeding programs to identify high-yielding, competitive oat cultivars suited to environments with increasing herbicide-resistant weeds. Understanding how oat (Avena sativa L.) cultivars differ in canopy development and competitive ability is essential for improving yield stability under increasing weed pressure. This study used unmanned aerial vehicle (UAV)-based multispectral imaging to characterize the temporal spectral and structural traits of sixteen oat cultivars grown under weed-free and weedy conditions across two locations for two years. Weedy conditions involved natural weed populations and pseudo-weeds where canola (Brassica napus) seeded as a weed. Weekly drone imaging was carried out using a multispectral sensor, which provided vegetation indices (NDVI, NDRE, ExG) and canopy metrics (ground cover, height, volume). Logistic and Gompertz models were fitted to cultivar traits to describe growth trajectories and obtain dynamic growth parameters. Cultivars showed clear differences in early canopy expansion, maximum NDVI, and canopy volume, with forage types expressing aggressive growth and several grain types combining high early growth rate with high yield potential. Machine-learning models integrating static and dynamic UAV-derived plant traits identified early ground cover and NDRE at three weeks after planting as the strongest predictors of grain yield. Models accurately predicted both weed-free (MAE = 262, R2 = 0.90) and weedy yield (MAE = 258, R2 = 0.90), demonstrating that early-season UAV traits capture the physiological and structural characteristics associated with competitive ability and grain yield. These findings show that high-throughput UAV phenotyping can reliably identify traits linked to yield formation and weed tolerance, providing a scalable approach for selecting competitive oat cultivars without relying solely on labor-intensive weedy field trials. [ABSTRACT FROM AUTHOR] |
| Copyright of Remote Sensing is the property of MDPI and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
|
Full text is not displayed to guests.
Login for full access.
|
|
| Abstract: | Highlights: What are the main findings? UAV-based temporal multispectral and structural traits (NDVI, NDRE, canopy cover, height, and volume) revealed strong differences among oat cultivars in early canopy development and weed-competitive ability. Machine-learning models (RF, GBM, PLS) showed that early-season traits—particularly ground cover and NDRE at 3 WAP—accurately predicted the grain yield in both weed-free and weedy conditions. What are the implications of the main findings? Early canopy traits measured in weed-free conditions can reliably predict cultivar performance under weed pressure, reducing the need for labor-intensive weed trials. The combined UAV + ML framework provides a scalable approach for breeding programs to identify high-yielding, competitive oat cultivars suited to environments with increasing herbicide-resistant weeds. Understanding how oat (Avena sativa L.) cultivars differ in canopy development and competitive ability is essential for improving yield stability under increasing weed pressure. This study used unmanned aerial vehicle (UAV)-based multispectral imaging to characterize the temporal spectral and structural traits of sixteen oat cultivars grown under weed-free and weedy conditions across two locations for two years. Weedy conditions involved natural weed populations and pseudo-weeds where canola (Brassica napus) seeded as a weed. Weekly drone imaging was carried out using a multispectral sensor, which provided vegetation indices (NDVI, NDRE, ExG) and canopy metrics (ground cover, height, volume). Logistic and Gompertz models were fitted to cultivar traits to describe growth trajectories and obtain dynamic growth parameters. Cultivars showed clear differences in early canopy expansion, maximum NDVI, and canopy volume, with forage types expressing aggressive growth and several grain types combining high early growth rate with high yield potential. Machine-learning models integrating static and dynamic UAV-derived plant traits identified early ground cover and NDRE at three weeks after planting as the strongest predictors of grain yield. Models accurately predicted both weed-free (MAE = 262, R2 = 0.90) and weedy yield (MAE = 258, R2 = 0.90), demonstrating that early-season UAV traits capture the physiological and structural characteristics associated with competitive ability and grain yield. These findings show that high-throughput UAV phenotyping can reliably identify traits linked to yield formation and weed tolerance, providing a scalable approach for selecting competitive oat cultivars without relying solely on labor-intensive weedy field trials. [ABSTRACT FROM AUTHOR] |
|---|---|
| ISSN: | 20724292 |
| DOI: | 10.3390/rs18081211 |
