Original papers
Phenotyping flag leaf nitrogen content in rice using a three-band spectral index

https://doi.org/10.1016/j.compag.2019.04.042Get rights and content

Highlights

  • TBDR (R755, R513, R508), a new three-band spectral index, effectively estimated rice leaf nitrogen content (LNC) even when the range of LNC was small.

  • TBDR (R755, R513, R508) performed much better than the 32 commonly used spectral indices in LNC estimation.

  • TBDR (R755, R513, R508) performed much better than the SPAD-502plus readings in LNC estimation.

  • TBDR (R755, R513, R508) may help design a practical and less expensive non-destructive real-time rice LNC sensor.

Abstract

Accurate, rapid and non-destructive measurements of rice flag leaf nitrogen content (LNC) are crucial for geneticists and breeders. To help design a less expensive, non-destructive, real-time LNC sensor, we developed a Three-Band Difference Ratio (TBDR) spectral index, TBDR (R755, R513, R508). This spectral index could accurately and rapidly estimate rice LNC in a population of chromosome segment substitution lines with small variation in LNC. The model estimating LNC was validated using a leave-one-out cross-validation technique; the achieved root mean square error was 0.13% and the relative error was 4.74%. In comparison with SPAD-502plus chlorophyll meter readings and commonly used spectral indices, including GreenSeeker- and Crop Circle-based indices, TBDR (R755, R513, R508) produced higher accuracy in LNC estimation.

Introduction

A major challenge for crop research in the 21st century is how to predict crop performance based on genetic background. Advances in “next generation” DNA sequencing have greatly enhanced the efficiency of genotyping while reducing its costs. Methods for characterizing plant traits (phenotypes), however, have progressed less rapidly over the decades, and constraints in phenotyping capability limit our ability to dissect the genetics of quantitative traits (Houle et al., 2010).

Chromosome segment substitution lines (CSSLs) can be used in genome-wide association studies to detect and fine-map quantitative trait loci (QTLs) and to study the interactions between those QTLs. Several populations of CSSLs have been developed for rice, and many QTLs for traits of agronomic importance have been detected in this way (Takai et al., 2007, Zhou et al., 2009, Zhu et al., 2009). These results have undoubtedly enhanced the understanding of the genetics of complex traits. However, efficient and accurate phenotyping of large populations is challenging for geneticists and breeders (Houle et al., 2010). This is particularly true when differences between CSSLs are relatively small but relevant.

Leaf nitrogen content (LNC) has been identified as a key plant trait determining photosynthetic capacity (Reich et al., 1998). Traditional measurements of LNC rely on destructive, costly and laborious methods (Ecarnot et al., 2013). Spectroscopic methods have been used to provide rapid and non-destructive estimates of leaf and canopy nitrogen contents of crops. Table 1 lists the vegetation spectral indices that have been commonly employed in remote sensing of LNC in crops (e.g., Tian et al., 2011, Li et al., 2014, He et al., 2016). In general, most existing spectral indices are based on two- or three-band combinations because including more bands may cause the problem of overfitting. The normalized difference vegetation index (NDVI) involving a near-infrared (NIR) band and a red band is one of the most widely used spectral indices in agricultural remote sensing (Rouse et al., 1974). Some NDVI-like two-band indices have been used to improve crop nitrogen management, e.g. the normalized pigment chlorophyll index (NPCI) (Peñuelas et al., 1994), the photochemical reflectance index (PRI) (Peñuelas et al., 1995), the green normalized difference vegetation index (Green-NDVI) (Gitelson et al., 1996), the normalised difference red edge index (NDRE) (Barnes et al., 2000), and the Crop Circle (CC) 730_670 (Erdle et al., 2011). Besides these two-band indices, a three-band index of modified normalized difference has been used to monitor leaf nitrogen status, because it effectively described changes in chlorophyll content (Sims and Gamon, 2002, Feng et al., 2008, Jin et al., 2014). Another three-band spectral index including reflectance at 491 nm, 705 nm and 717 nm was suggested as a good indicator of LNC in rice (Tian et al., 2011). However, these existing indices need further testing.

Based on these spectral indices, some apparatuses have been designed and widely applied to make non-destructive real-time diagnoses of crop nitrogen status and guide in-season nitrogen management (Cao et al., 2017). The handheld GreenSeeker (NTech Industries Inc., Ukiah, California, USA) measures canopy reflectance in the red (656 nm) region and the near infrared (NIR; 774 nm) region to calculate the GreenSeeker NDVI readings (Table 1). It has been applied for site-specific and need-based nitrogen management in rice (Bijay-Singh et al., 2011, Ali et al., 2014). Another widely used optical sensor is the Crop Circle sensor (Holland Scientific Inc., Lincoln, Nebraska, USA), which can produce NDVI-like spectral indices (Table 1). It has been applied for non-destructive and real-time nitrogen status assessment in rice (Cao et al., 2013), wheat (Cao et al., 2017) and potato (Giletto and Echeverria, 2016).

In contrast to the sensors of GreenSeeker and Crop Circle that assess reflectance, the SPAD-502plus chlorophyll meter (Minolta Camera Co., Osaka, Japan) measures transmittance at 650 nm and 940 nm to provide a relative measure of leaf chlorophyll content. The SPAD-502plus is designed as a light-weight hand-held chlorophyll meter; it has been employed as a low-cost, rapid, simple, and non-destructive apparatus to diagnose the nitrogen nutrition status of various crops (Yang et al., 2014).

Although these commonly used spectral indices have effectively estimated LNC in previous studies, it is unclear whether they are effective for accurate phenotyping because variations of LNC may be small among CSSLs. Therefore, the objectives of the present study were (1) to evaluate the applicability and reliability of commonly used spectral indices on estimating LNC of rice; (2) to develop a new LNC-sensitive spectral index that can estimate LNC more effectively than the commonly used spectral indices. The new spectral index may help designing a practical and less expensive, non-destructive, real-time LNC meter, and the anticipated outcome would help to provide efficient and accurate phenotyping of large populations for geneticists and breeders.

Section snippets

Experimental design and crop cultivation

In this study, a wide population consisting of 130 CSSLs was developed, derived from the crossing and back-crossing of two sequenced rice cultivars: 9311, an elite indica cultivar as the recipient and Nipponbare, a japonica cultivar as the donor (Xu et al., 2010). The experiment was conducted at the Experimental Farm of Yangzhou University, Yangzhou, Jiangsu, China (32o24′ N, 119° 23′E) in 2016 following a complete randomized block design, with three replications, three rows per plot, 12 hills

Relationship between LNC and SPAD readings

The SPAD values of rice flag leaves showed a significant linear relationship with LNC, with an R2 value of 0.38 and an SE value of 0.17% (Fig. 1).

Relationships between LNC and reflectance of single wavelength

Linear relationships between LNC and single reflectance were assessed by their correlation coefficient (r) (Fig. 2). The values of r varied notably with wavelength, and there were two peaks of absolute values of r located in the green and red edge regions, respectively. Specifically, the absolute values of r were higher than 0.60 only within the

Discussion and conclusions

SD (Rλ1, Rλ2) indices showed strong relationships with LNC within the wavelength range of 500–540 nm (Fig. 3a). Among them, SD (R513, R508) that represented the spectral reflectance increase within the short wavelength range of 508–513 nm was found to be the best SD index for LNC estimation of rice. TBDR (R755, R513, R508) introduced the reflectance at 755 nm into SD (R513, R508), and it produced a higher R2. Similarly, Liang et al. (2018) developed the first derivative ratio nitrogen

Acknowledgements

This work was supported by National Key Research and Development Program of China (2017YFD0200107); National Natural Science foundation of China (31872853); Natural Science Foundation of Jiangsu Province, China (grant no. BK20171286); National Science and Technology Ministry (project no. 2015BAD01B03); the Key Research and Development Programme of Jiangsu Province, China (project no. BE2015337-11); and the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD),

References (53)

  • R.K. Gupta et al.

    Comparative analysis of red edge hyperspectral indices

    Adv. Space Res.

    (2003)
  • P.M. Hansen et al.

    Reflectance measurement of canopy biomass and nitrogen status in wheat crops using normalized difference vegetation indices and partial least squares regression

    Remote Sens. Environ.

    (2003)
  • L. He et al.

    Improved remote sensing of leaf nitrogen concentration in winter wheat using multi-angular hyperspectral data

    Remote Sens. Environ.

    (2016)
  • A.R. Huete

    A soil-adjusted vegetation index (SAVI)

    Remote Sens. Environ.

    (1988)
  • Y. Inoue et al.

    Diagnostic mapping of canopy nitrogen content in rice based on hyperspectral measurements

    Remote Sens. Environ.

    (2012)
  • P.D. Jamieson et al.

    A test of the computer simulation model ARC-WHEAT1 on wheat crops grown in New Zealand

    Field Crop. Res.

    (1991)
  • X. Jin et al.

    Remote estimation of soil organic matter content in the Sanjiang Plain, Northeast China: the optimal band algorithm versus the GRA-ANN model

    Agr. Forest Meteorol.

    (2016)
  • F. Li et al.

    Reflectance estimation of canopy nitrogen content in winter wheat using optimised hyperspectral spectral indices and partial least squares regression

    Eur. J. Agron.

    (2014)
  • J. Peñuelas et al.

    Reflectance indices associated with physiological changes in nitrogen and water-limited sunflower leaves

    Remote Sens. Environ.

    (1994)
  • D.A. Sims et al.

    Relationships between leaf pigment content and spectral reflectance across a wide range of species, leaf structures and developmental stages

    Remote Sens. Environ.

    (2002)
  • D. Stroppiana et al.

    Plant nitrogen concentration in paddy rice from field canopy hyperspectral radiometry

    Field Crop. Res.

    (2009)
  • Y.C. Tian et al.

    Assessing newly developed and published vegetation indices for estimating rice leaf nitrogen concentration with ground- and space-based hyperspectral reflectance

    Field Crop. Res.

    (2011)
  • Barnes, E.M., Clarke, T.R., Richards, S.E., Colaizzi, P.D., Haberland, J., Kostrzewski, M., 2000. Coincident detection...
  • R.K. Sharma Bijay-Singh et al.

    Assessment of the nitrogen management strategy using an optical sensor for irrigated wheat

    Agron. Sustain. Dev.

    (2011)
  • Q. Cao et al.

    Developing a new crop circle active canopy sensor-based precision nitrogen management strategy for winter wheat in North China Plain

    Precis. Agric.

    (2017)
  • J. Dash et al.

    The MERIS terrestrial chlorophyll index

    Int. J. Remote Sens.

    (2004)
  • Cited by (22)

    • Estimating leaf nitrogen content by coupling a nitrogen allocation model with canopy reflectance

      2022, Remote Sensing of Environment
      Citation Excerpt :

      Since reflectance spectra are directly sensitive to the area-based content rather than to the mass-based content or concentration of absorbers, LNCA should be estimated from reflectance spectra and (if needed) converted to LNCM using the leaf mass per area (LMA): LNCM = LNCA/LMA (Wright et al., 2004; Li et al., 2018b). Although remote sensing has emerged as a powerful tool to quantify LNC (Cheng et al., 2018; Berger et al., 2020b), most studies have focused on the estimation of LNCM (Zhu et al., 2007; Wang et al., 2012; Wang et al., 2019), and only a few studies have revealed the difference in estimating LNCA and LNCM from reflectance spectra (e.g. Li et al., 2018b). In addition, as pointed out by (Kattenborn et al., 2019) for pigments, the retrieval of the area-based content from canopy reflectance spectra is superior to the mass-based concentration, and this statement is also suitable for the estimation of nitrogen, as proven by (Li et al., 2018b).

    • Mapping cotton cultivated area combining remote sensing with a fused representation-based classification algorithm

      2021, Computers and Electronics in Agriculture
      Citation Excerpt :

      As a second, the number of pixels labeled as cotton for each municipality was counted from the MODIS-derived map, which was then multiplied by the MODIS pixel size. The comparison between the cotton areas detected from MODIS images and municipal-level statistics of sown areas was performed by calculating the coefficient of determination (R2) and root mean square error (RMSE) (Wang et al., 2019a). The annual time series EVI from six sites A, B, C, D, E and F are shown in Fig. 5.

    View all citing articles on Scopus
    View full text