Knowledge Center Catalog

Local cover image
Local cover image

UAV-based high-throughput phenotyping to increase prediction and selection accuracy in maize varieties under artificial MSV inoculation

By: Contributor(s): Material type: ArticleArticleLanguage: English Publication details: Amsterdam (Netherlands) : Elsevier, 2021.ISSN:
  • 0168-1699
Subject(s): In: Computers and Electronics in Agriculture v. 184, art. 106128Summary: The use of unmanned aerial vehicles’ (UAV) remotely sensed data in crop evaluation is revolutionizing the field of plant phenotyping. This study was conducted to (1) develop protocol to predict maize streak virus (MSV) and grain yield using UAV-derived multispectral data; and (2) identify the suitable predictor variables and ideal phenological stages for MSV and grain yield prediction. Twenty-five maize varieties were evaluated under artificial MSV inoculation. Manual scoring and multispectral imaging measurements were performed at mid-vegetative, flowering and mid-grain filling stages. UAV-derived data were acquired in the multispectral bands of Green (0.53–0.57 μm), Red (0.64–0.68 μm), Red-edge (0.73–0.74 μm) and Near-Infrared (0.77–0.81 μm). Eight vegetation indices were determined: NDVI (normalized difference vegetation index), NDVIred-edge, GNDVI (green normalized difference vegetation index), SR (simple ratio), CIgreen (green chlorophyll index), CIred-edge (red-edge chlorophyll index), SAVI (soil-adjusted vegetation index) and OSAVI (optimized SAVI). Finally, predictions of MSV and grain yield were performed with 36 models using multiple regression, decision trees and linear regression. Frequently selected variables for MSV prediction were Green band at vegetative (61.5%), Red band at vegetative (68.4%) and flowering (80.4%), and GNDVI at mid-vegetative (88.7%). The best MSV predictors were GNDVI (r = 0.84; RMSE = 0.85), CIgreen (r = 0.83; RMSE = 0.86) and Red band (r = 0.77; RMSE = 0.99) measured at mid-vegetative stage. Six out of 36 models were selected as ideal for predicting maize grain yield: RF-REF-NIRF (r = 0.69; RMSE = 0.65); NDVIREG-GNDVIG (r = 0.74; RMSE = 0.56); RV-NIRV (r = 0.84; RMSE = 0.37); and the tree with the largest correlations are RV-NIRV-RF (r = 0.86; RMSE = 0.32); GNDVIV-OSAVIV (r = 0.84; RMSE = 0.36); GV-RV-NIRV (r = 0.84; RMSE = 0.35); the last two of which were at mid-vegetative stage. We conclude that UAV-based multispectral remote sensing is a reliable tool for phenotyping MSV disease and grain yield prediction, and mid-vegetative appear to be the most ideal phenological stage for MSV and grain yield prediction.
Tags from this library: No tags from this library for this title. Log in to add tags.
Star ratings
    Average rating: 0.0 (0 votes)
Holdings
Item type Current library Collection Call number Status Date due Barcode Item holds
Article CIMMYT Knowledge Center: John Woolston Library CIMMYT Staff Publications Collection Available
Total holds: 0

Peer review

The use of unmanned aerial vehicles’ (UAV) remotely sensed data in crop evaluation is revolutionizing the field of plant phenotyping. This study was conducted to (1) develop protocol to predict maize streak virus (MSV) and grain yield using UAV-derived multispectral data; and (2) identify the suitable predictor variables and ideal phenological stages for MSV and grain yield prediction. Twenty-five maize varieties were evaluated under artificial MSV inoculation. Manual scoring and multispectral imaging measurements were performed at mid-vegetative, flowering and mid-grain filling stages. UAV-derived data were acquired in the multispectral bands of Green (0.53–0.57 μm), Red (0.64–0.68 μm), Red-edge (0.73–0.74 μm) and Near-Infrared (0.77–0.81 μm). Eight vegetation indices were determined: NDVI (normalized difference vegetation index), NDVIred-edge, GNDVI (green normalized difference vegetation index), SR (simple ratio), CIgreen (green chlorophyll index), CIred-edge (red-edge chlorophyll index), SAVI (soil-adjusted vegetation index) and OSAVI (optimized SAVI). Finally, predictions of MSV and grain yield were performed with 36 models using multiple regression, decision trees and linear regression. Frequently selected variables for MSV prediction were Green band at vegetative (61.5%), Red band at vegetative (68.4%) and flowering (80.4%), and GNDVI at mid-vegetative (88.7%). The best MSV predictors were GNDVI (r = 0.84; RMSE = 0.85), CIgreen (r = 0.83; RMSE = 0.86) and Red band (r = 0.77; RMSE = 0.99) measured at mid-vegetative stage. Six out of 36 models were selected as ideal for predicting maize grain yield: RF-REF-NIRF (r = 0.69; RMSE = 0.65); NDVIREG-GNDVIG (r = 0.74; RMSE = 0.56); RV-NIRV (r = 0.84; RMSE = 0.37); and the tree with the largest correlations are RV-NIRV-RF (r = 0.86; RMSE = 0.32); GNDVIV-OSAVIV (r = 0.84; RMSE = 0.36); GV-RV-NIRV (r = 0.84; RMSE = 0.35); the last two of which were at mid-vegetative stage. We conclude that UAV-based multispectral remote sensing is a reliable tool for phenotyping MSV disease and grain yield prediction, and mid-vegetative appear to be the most ideal phenological stage for MSV and grain yield prediction.

Text in English

Click on an image to view it in the image viewer

Local cover image

International Maize and Wheat Improvement Center (CIMMYT) © Copyright 2021.
Carretera México-Veracruz. Km. 45, El Batán, Texcoco, México, C.P. 56237.
If you have any question, please contact us at
CIMMYT-Knowledge-Center@cgiar.org