Spectral Vegetation Indices to Track Senescence Dynamics in Diverse Wheat Germplasm - PubMed Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2020 Jan 28:10:1749.
doi: 10.3389/fpls.2019.01749. eCollection 2019.

Spectral Vegetation Indices to Track Senescence Dynamics in Diverse Wheat Germplasm

Jonas Anderegg  1 , Kang Yu  1   2 , Helge Aasen  1 , Achim Walter  1 , Frank Liebisch  1 , Andreas Hund  1
Affiliations

Spectral Vegetation Indices to Track Senescence Dynamics in Diverse Wheat Germplasm

Jonas Anderegg et al. Front Plant Sci. .

Abstract

The ability of a genotype to stay green affects the primary target traits grain yield (GY) and grain protein concentration (GPC) in wheat. High throughput methods to assess senescence dynamics in large field trials will allow for (i) indirect selection in early breeding generations, when yield cannot yet be accurately determined and (ii) mapping of the genomic regions controlling the trait. The aim of this study was to develop a robust method to assess senescence based on hyperspectral canopy reflectance. Measurements were taken in three years throughout the grain filling phase on >300 winter wheat varieties in the spectral range from 350 to 2500 nm using a spectroradiometer. We compared the potential of spectral indices (SI) and full-spectrum models to infer visually observed senescence dynamics from repeated reflectance measurements. Parameters describing the dynamics of senescence were used to predict GY and GPC and a feature selection algorithm was used to identify the most predictive features. The three-band plant senescence reflectance index (PSRI) approximated the visually observed senescence dynamics best, whereas full-spectrum models suffered from a strong year-specificity. Feature selection identified visual scorings as most predictive for GY, but also PSRI ranked among the most predictive features while adding additional spectral features had little effect. Visually scored delayed senescence was positively correlated with GY ranging from r = 0.173 in 2018 to r = 0.365 in 2016. It appears that visual scoring remains the gold standard to quantify leaf senescence in moderately large trials. However, using appropriate phenotyping platforms, the proposed index-based parameterization of the canopy reflectance dynamics offers the critical advantage of upscaling to very large breeding trials.

Keywords: canopy reflectance; feature selection; field-based phenotyping; high-throughput phenotyping; hyperspectral remote sensing.

PubMed Disclaimer

Figures

Figure 1
Figure 1
Scaled visual scorings of canopy greenness (Sc) and a scaled spectral index (SI) as a function of thermal time after heading for one experimental plot. Linear interpolation was used to derive the onset (Onsen), mid (Midsen) and end (Endsen) of the rapid senescence phase, its duration (Tsen) and the deviation of the SI curve from the Sc curve (error; shaded area). Black arrows represent the difference between SI- and Sc-derived parameters. The mean of these differences across all plots represents a measure of bias.
Figure 2
Figure 2
Overview of the objectives of this study and the implemented workflow: pre-processing of reflectance spectra and conversion to spectral indices (SI) [1]; full-spectrum models (Mod) to obtain predictions (Pred) of visual senescence scorings (Sc) based on reflectance spectra [4]; fitting of SI, Sc and Pred against thermal time and extraction of corresponding dynamics parameters (DynPars) [2]; Unsupervised DynParsSI subset selection [3]; Model and SI evaluation based on DynPars [5]; Spatial correction and calculation of best linear unbiased estimators (BLUEs) [6]; Modelling of primary traits (i.e. grain yield (GY) and grain protein concentration (GPC)) and supervised feature selection by recursive feature elimination [7] to determine the most predictive features and estimate the potential benefits of a high spectral resolution.
Figure 3
Figure 3
Daily mean temperatures (black solid line), daily maximum temperatures (red dotted line) and rainfall measured at 2 m above the ground for the main growing period of the experiment at the field phenotyping platform of ETH Zurich. Temperature data was retrieved from an on-site weather station. Rainfall data was obtained from a nearby weather station of the federal Swiss meteorological network Agrometeo (www.agrometeo.ch).
Figure 4
Figure 4
General reflectance patterns of senescing wheat canopies. (A) Mean reflectance spectrum of wheat genotypes through the process of senescence (10 denotes completely green canopies, 0 denotes complete senescence, based on visual scorings). Data from all time points and all years was used to calculate the mean reflectance spectrum per scoring. Vertical lines mark the wavebands constituting the NDVI and the PSRI. (B) Pearson correlation between reflectance at each wavelength and visual senescence scorings. Positive correlations indicate increasing reflectance as senescence progresses, negative correlations indicate decreasing reflectance as senescence progresses; Year-specific correlation coefficients; (C) Separate analyses for early senescence (scorings = [0:3]), intermediate senescence (scorings = [4:7] and late senescence (scorings = [8:10]). This part of the graph is based on data of the 2018 experiment.
Figure 5
Figure 5
Dynamic pattern of scaled spectral indices (in grey) and visual canopy senescence scores (in green; identical in all subplots) over thermal time after heading. Mean linearly interpolated values over all experimental plots of the 2016 experiment (thick lines) and their standard deviations (thin lines) are shown. Dashed lines mark the thresholds defined as onset, midpoint and end of senescence.
Figure 6
Figure 6
Time courses of PSRI, NDVI (nadir view) and visual scorings (whole plot, 45° viewing angle) for two experimental plots of the 2018 experiment. Left: Genotype with changing spike orientation during grain filling. With time, spikes make up an increasingly dominant part of the image. Concomitantly, NDVI values decrease early in the grain filling phase, while visual scorings indicate no change in canopy greenness (evidenced by red arrow). Right: Genotype characterized by relatively stable spike orientation during grain filling and comparable NDVI, PSRI and scoring time courses (evidenced by red arrow); Letters A-F in the upper part of the Figure represent time points when corresponding images were taken. Images were taken by the field phenotyping platform (FIP, Kirchgessner et al., 2017).
Figure 7
Figure 7
Example of model within-year and across-year validation results. Predictions of senescence scorings obtained from full-spectrum models are plotted against the visual scorings (observed). Here, averaged reflectance spectra were used, and the data was mean-centered and scaled to unit variance prior to modeling. Data from the 2017 experiment was used for model training. Models were validated on held-out samples of the same year (within-year validation, red) as well as on samples from the 2018 experiment (across-year validation, cyan). The full dataset was used for model training, i.e. no down-sampling was performed, whereas validation datasets were randomly down-sampled. (A) Results for partial least squares regression; (B) Results for cubist regression.
Figure 8
Figure 8
(A) Performance of the cubist regression models to predict visual senescence scorings depending on the number of wavelengths used as predictors. Mean performance as measured by the RMSE of predictions and standard deviations are shown based on 30 resamples of the data. (B) Frequency of wavelengths resulting among the most informative to predict visual scorings of canopy senescence. Frequencies denote the number of times out of 30 resampling iterations in which a given wavelength was retained in the cubist regression model down to a subset size of 12 wavelengths during recursive feature elimination. Only wavelengths which were among the top 12 predictors in at least 10% of the resamples (i.e. in at least 3 resamples, marked by the dashed horizontal line) are shown. The grey line represents the mean reflectance spectrum of canopies with a visual scoring of 8 (early senescence).
Figure 9
Figure 9
Correlation-based sensitivity analysis of the spectral bands (500, 678 and 756) constituting the PSRI. The “x” in the SI formula denotes the reflectance at the waveband that was varied in the depicted range.
Figure 10
Figure 10
(A) Performance of the random forest regression models to predict grain protein concentration (GPC) and grain yield (GY). Mean performance and standard deviation are shown based on 30 resamples of the data for models containing a decreasing number of features selected by recursive feature elimination. (B) Feature ranks as determined by recursive feature elimination. Mean feature rank and standard deviation are shown based on 30 resamples of the data for the top and lowest 15 features, separated by the broken line. Features are plotted according to their descending mean rank in the 2016 models.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Aasen H., Bolten A. (2018). Multi-temporal high-resolution imaging spectroscopy with hyperspectral 2D imagers – From theory to application. Remote Sens. Environ. 205, 374–389. 10.1016/j.rse.2017.10.043 - DOI
    1. Aasen H., Honkavaara E., Lucieer A., Zarco-Tejada P. J. (2018). Quantitative remote sensing at ultra-high resolution with uav spectroscopy: a review of sensor technology, measurement procedures, and data correction workflows. Remote Sens. 10, 1091. 10.3390/rs10071091 - DOI
    1. Ambroise C., McLachlan G. J. (2002). Selection bias in gene extraction on the basis of microarray gene-expression data. PNAS 99, 6562–6566. 10.1073/pnas.102102699 - DOI - PMC - PubMed
    1. Asrar G., Fuchs M., Kanemasu E. T., Hatfield J. L. (1984). Estimating absorbed photosynthetic radiation and leaf area index from spectral reflectance in Wheat 1. Agron. J. 76, 300–306. 10.2134/agronj1984.00021962007600020029x - DOI
    1. Barmeier G., Schmidhalter U. (2017). High-Throughput Field Phenotyping of Leaves, Leaf Sheaths, Culms and Ears of Spring Barley Cultivars at Anthesis and Dough Ripeness. Front. Plant Sci. 8. 10.3389/fpls.2017.01920 - DOI - PMC - PubMed

LinkOut - more resources