Hyperspectral reflectance at canopy and leaf levels for predicting yield and physiological traits in spring wheat under Mediterranean conditions
High-throughput field phenotyping offers an efficient solution for identifying and selecting genotypes of interest in plant breeding. This study aimed to develop multivariate models using spectral reflectance data to estimate physiological and yield traits in spring wheat genotypes exposed to differ...
| Autores: | , , , , , , |
|---|---|
| Tipo de recurso: | artículo |
| Estado: | Versión publicada |
| Fecha de publicación: | 2025 |
| País: | España |
| Institución: | Consejo Superior de Investigaciones Científicas (CSIC) |
| Repositorio: | DIGITAL.CSIC. Repositorio Institucional del CSIC |
| OAI Identifier: | oai:digital.csic.es:10261/406001 |
| Acceso en línea: | http://hdl.handle.net/10261/406001 https://api.elsevier.com/content/abstract/scopus_id/105012431991 |
| Access Level: | acceso abierto |
| Palabra clave: | Carbon isotope composition Chlorophyll fluorescence High-throughput phenotyping Hyperspectral reflectance Leaf gas exchange Wheat |
| Sumario: | High-throughput field phenotyping offers an efficient solution for identifying and selecting genotypes of interest in plant breeding. This study aimed to develop multivariate models using spectral reflectance data to estimate physiological and yield traits in spring wheat genotypes exposed to different water regimes. Fifteen spring wheat varieties and one triticale genotype were evaluated in sixteen environments, which were generated by combining data from over four seasons in two Mediterranean locations in Chile, along with two water regimes (irrigated and water deficit). Measured traits were leaf pigments, leaf area index (LAI), leaf water potential (Ψ<inf>leaf</inf>), gas exchange, chlorophyll fluorescence, grain yield, and carbon isotope composition (δ<sup>13</sup>C). Hyperspectral reflectance was recorded at the leaf level and canopy level (45° and 90°) at anthesis and grain filling and used to generate predictive models using partial least squares (PLS), least absolute shrinkage and selection operator (LASSO), and elastic net (E.net) regression. Models explained over 60% of the trait variation (R<sup>2</sup>) for 70% of traits analysed. Fluorescence parameters (R<sup>2</sup> = 0.78–0.88), δ<sup>13</sup>C (R<sup>2</sup> = 0.80), leaf pigments (R<sup>2</sup> = 0.50–0.74), Ψ<inf>leaf</inf> (R<sup>2</sup> = 0.72), and LAI (R<sup>2</sup> = 0.68) had the most robust predictions. LASSO regression showed the highest R<sup>2</sup> and accuracy, while canopy-level spectra at 90° excelled in predicting grain yield and LAI, and leaf-level spectra were best for fluorescence traits. These methods facilitated the identification of genotypes with superior water-deficit adaptation and yield potential, accelerating breeding, enhancing crop resilience to climate change, and improving food security. |
|---|