An assessment of background removal approaches for improved estimation of rice leaf nitrogen concentration with unmanned aerial vehicle multispectral imagery at various observation times
Graphical Abstract
Introduction
Rice (Oryza sativa L.) is one of the major cereal crops in the world, accounting for 29% of the cultivated area in China (Xiao, 2003). Its yield is not only closely related to international grain trade but also fundamental to global food security, especially in developing countries (Normile, 2008, Zhao et al., 2013, Zheng et al., 2019). Nitrogen (N) fertilizer, one of the most direct N nutrient sources in crop photosynthetic activity and productivity, plays a crucial role in improving rice yield and grain quality (Ju et al., 2009, Vitousek et al., 2009). However, excessive N fertilizer application will break the balance between supply and demand (Miao et al., 2011) and lead to a series of environmental problems such as water pollution (Chen et al., 2005) and the greenhouse effect (Sehy et al., 2003). Furthermore, the N status of rice is a crucial indicator to evaluate whether the N fertilizer application is reasonable (Miao et al., 2011). As a consequence, it is significant to monitor rice N status in order to achieve high yields and improve the N use efficiency (Shibayama and Akiyama, 1991, Zhang et al., 2011).
At present, leaf nitrogen concentration (LNC) has been commonly used as a crucial indicator for monitoring crop N status (Zhu et al., 2007). Furthermore, LNC estimation for critical growth stages is not only an important evidence for agronomists to make fertilization recommendations, but also an essential indicator to predict grain protein content (Wang et al., 2004). Compared to traditional measuring methods in the lab, remote sensing has been widely used to monitor crop LNC due to its simplicity and efficiency (Li et al., 2018, Zhou et al., 2018). It is a common practice to estimate LNC with the chlorophyll sensitive vegetation indices comprising near-infrared (NIR) and red edge bands based on the close linkage between LNC and leaf chlorophyll content (LCC) (Cheng et al., 2018, Zhou et al., 2018, Lu et al., 2019). Various remote sensing platforms have been used to monitor crop LNC, such as near-surface hyperspectral imaging systems (Tian et al., 2014, Zhou et al., 2018), unmanned aerial vehicles (UAV) (Lu et al., 2019, Wang et al., 2021), and satellites (Mutanga et al., 2015, Omer et al., 2017). Nevertheless, most of them pay more attention to the use of near-surface hyperspectral images and satellite images given the high spectral resolution or the wide field of view (Tian et al., 2014, Omer et al., 2017). Only a few studies have demonstrated that UAV platforms can be used to monitor LNC because of their low cost and ease of operation (Lu et al., 2019, Osco et al., 2020, Wang et al., 2021). However, these studies did not consistently consider the background effect (Table 1), which represents the impact of soil and water in the field on the canopy reflectance spectra because of the spectral mixing of plants and background materials in the field of view (Yao et al., 2014, Zhou et al., 2017, Zhou et al., 2018). Therefore, further research is needed to determine the necessity of background removal when estimating the LNC from UAV imagery.
The plant signal could be weakened because of the mixture of background signals in canopy reflectance, which often lead to worsened estimation of crop biochemical parameters. In this regard, some approaches were proposed to eliminate the background effect for improved monitoring of crop growth status (Gilabert et al., 2002, Jay et al., 2017a, Wang et al., 2021), such as the use of soil-adjusted vegetation indices (Gilabert et al., 2002) and extraction of green pixels for refining vegetation indices (Zhou et al., 2017, Zhou et al., 2018). Recently, (Wang et al., 2021) proposed a new approach to reducing the background effect by integrating a traditional VI and vegetation abundance through the use of linear spectral mixture analysis (LSMA). Since these approaches were used separately in different studies, it is unclear which one is the most effective and efficient to eliminate the background effect caused by soil and water, especially for the complex paddy fields. Therefore, it is beneficial to assess these approaches in reducing the background effect and determine the most useful approach for the estimation of rice LNC.
In addition, the studies on background removal paid more attention to the early growth stages (e.g., tillering and jointing) than the later growth stages (e.g., booting and heading) (Yao et al., 2014, Prudnikova et al., 2019, Wang et al., 2021). When the canopies become denser at later stages (e.g., booting and heading), the canopy reflectance is also influenced by the shadows cast by the leaves or stems in the upper layers on those in the lower layers (Jay et al., 2017a, Jay et al., 2019). Many studies have suggested that the sunlit and shaded components of the canopy should be distinguished because of the negative influence of shaded-leaf components on canopy reflectance when estimating N-related leaf traits (Zarco-Tejada et al., 2004, Jay et al., 2017a, Jay et al., 2017b). In contrast, a few studies have demonstrated that it is unnecessary to distinguish the sunlit and shaded components of the canopy in the near-ground hyperspectral images with an ultra-high spatial resolution (~ 1 mm) (Zhou et al., 2018, Jiang et al., 2021). Therefore, it remains unclear whether it is necessary to distinguish sunlit and shaded components for both early and later growth stages. It is also worth investigating whether the inclusion of shaded-leaf pixels derived from coarser-resolution UAV imagery (~ 5 cm) could affect the performance of background removal methods.
As an important factor affecting the sun-target-sensor geometry, observation time is critical to the fraction of sunlit components in the canopy and the estimation of crop biochemical parameters (Li et al., 2020). However, estimation of LNC from UAV-based multispectral imagery at multiple observation times is poorly understood. This is an important consideration because UAV flights can be completed within a short time but have to be in an appropriate time window with ideal weather conditions. Although some studies indicated the acquisition of UAV images at noon (12:00 h local time) (Jay et al., 2019, Jiang et al., 2021), others used various time windows within the period of 10:00–14:00 local time due to the stronger sunspots at noon (Liu et al., 2017, Zheng et al., 2018, Li et al., 2019, Fu et al., 2020, Osco et al., 2020, Lu et al., 2021) (Table 1). As a consequence, how the noon UAV imagery compares to off-noon UAV imagery for LNC estimation remains a valuable question to be answered. A recent study by Li et al. (2020) had demonstrated that the off-noon observation (15:00 local time) could yield better performance in LCC estimation as compared to the observation of noon (12:00 local time) at the jointing stage of rice and winter wheat. However, this task focused only on the early growth stages and it is uncertain whether their conclusion still holds for the later growth stages. Therefore, it is valuable for precision agriculture to assess the sensitivity of background removal approaches to observation time in LNC estimation.
Therefore, we compared the approaches adjusted for background removal and the traditional method without adjustment in the ability of LNC estimation for UAV-based multispectral imagery for individual and across growth stages. We also assessed the sensitivity of the background removal approaches to observation time and growth stage. The specific objectives of this study were: (1) to investigate the sensitivity of LNC ~ VI relationships for three background removal approaches to observation time, (2) to determine the best background removal approach for improving the accuracy in LNC estimation, and (3) to evaluate the effect of sunlit and shaded components on the abundance adjustment approach.
Section snippets
Experimental design
The rice field experiments were conducted at the experimental station of National Engineering and Technology Center for Information Agriculture (NETCIA) located in Xinghua, Jiangsu province, China, (119°53′E, 33°05′N). The rice experiments were conducted over two consecutive years (2018–2019) as the design of randomized block with three replications and the same treatments in rice cultivar, planting patterns and N rates (Fig. 1& Table 2). Two rice cultivars for planting were Yongyou 2640 (V1,
Sensitivity of VIs to observation time in LNC estimation
Fig. 5 shows the influence of observation times on the sensitivity of VIs in LNC estimation at individual and across growth stages. The r2 values over diurnal times generally followed a bowled-shape pattern with stronger relationships for off-noon times than noon times. All the VIs adjusted for background removal exhibited stronger relationships between LNC and VIs than CIre without adjustment. Meanwhile, AACIre-all yielded the highest r2 values among three VIs with background elimination at
The necessity of background removal
It is beneficial to alleviate the background effect and enhance the information of interest with simplified and effective approaches when using canopy reflectance to estimate crop biochemical parameters (Yao et al., 2014, Jay et al., 2019, Wang et al., 2021). This could be utilized to reveal the reason why background removal approaches outperformed the methods without removal. Here, this study reported solid findings that the AAVI approach outperformed the vegetation index from green pixels and
Conclusions
This study examined the importance of background removal in LNC estimation with UAV-based multispectral imagery and compared the sensitivities of three background removal approaches to the observation time and growth stage. The VIs adjusted for background removal yielded stronger relationships with LNC as compared to CIre without adjustment regardless of observation times and growth stages. The VIs for off-noon observation times showed better relationships with LNC than those for noon at
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This work was supported by the National Key Research and Development Program of China (2019YFE0125500), the National Natural Science Foundation of China (41871259, 32021004), Jiangsu Agricultural Science and Technology Innovation Fund (CX (21) 1006), and Collaborative Innovation Center for Modern Crop Production co-sponsored by Province and Ministry. We would like to thank Xiao Zhang, Penglei Li, Gaoxiang Yang, Xue Zhang, Ke Zhang, and Yanyu Wang for their help in field data collection.
References (70)
- et al.
Demonstrating the stability of support vector machines for classification
Signal Process
(2006) - et al.
Modelling leaf chlorophyll content in broadleaf and needle leaf canopies from ground, CASI, Landsat TM 5 and MERIS reflectance data
Remote Sens. Environ.
(2013) - et al.
The effect of spatial resolution on radiometric and geometric performances of a UAV-mounted hyperspectral 2D imager
ISPRS J. Photogramm. Remote Sens.
(2018) - et al.
Suitability of NDVI and OSAVI as estimators of green biomass and coverage in a semi-arid rangeland
Ecol. Indic.
(2018) - et al.
Wavelet analysis of MODIS time series to detect expansion and intensification of row-crop agriculture in Brazil
Remote Sens. Environ.
(2008) - et al.
A generalized soil-adjusted vegetation index
Remote Sens. Environ.
(2002) - et al.
Improved remote sensing of leaf nitrogen concentration in winter wheat using multi-angular hyperspectral data
Remote Sens. Environ.
(2016) A soil-adjusted vegetation index (SAVI)
Remote Sens. Environ.
(1988)- et al.
Exploiting the centimeter resolution of UAV multispectral imagery to improve remote-sensing estimates of canopy structure and biochemistry in sugar beet crops
Remote Sens. Environ.
(2019) - et al.
Estimating leaf chlorophyll content in sugar beet canopies using millimeter- to centimeter-scale reflectance imagery
Remote Sens. Environ.
(2017)
Retrieving LAI, chlorophyll and nitrogen contents in sugar beet crops from multi-angular optical remote sensing: Comparison of vegetation indices and PROSAIL inversion for field phenotyping
Field Crop. Res.
Improved estimation of leaf chlorophyll content of row crops from canopy reflectance spectra through minimizing canopy structural effects and optimizing off-noon observation time
Remote Sens. Environ.
Estimation of area- and mass-based leaf nitrogen contents of wheat and rice crops from water-removed spectra using continuous wavelet analysis
Plant Methods
Impact of the reproductive organs on crop BRDF as observed from a UAV
Remote Sens. Environ.
Multi-LUTs method for canopy nitrogen density estimation in winter wheat by field and UAV hyperspectral
Comput. Electron. Agric.
Combining plant height, canopy coverage and vegetation index from UAV-based RGB images to estimate leaf nitrogen concentration of summer maize
Biosyst. Eng.
Angle effects of vegetation indices and the influence on prediction of SPAD values in soybean and maize
Int. J. Appl. Earth Obs. Geoinf.
Solar zenith angle effects on vegetation indices in tallgrass prairie
Remote Sens. Environ.
Evaluating the robustness of models developed from field spectral data in predicting African grass foliar nitrogen concentration using WorldView-2 image as an independent test dataset
Int. J. Appl. Earth Obs. Geoinf.
Mapping leaf nitrogen and carbon concentrations of intact and fragmented indigenous forest ecosystems using empirical modeling techniques and WorldView-2 data
ISPRS J. Photogramm. Remote Sens.
Sun-view angle effects on reflectance factors of corn canopies
Remote Sens. Environ.
Random Forest classification of Mediterranean land cover using multi-seasonal imagery and multi-seasonal texture
Remote Sens. Environ.
Nitrous oxide fluxes from maize fields: relationship to yield, site-specific fertilization, and soil conditions
Agric. Ecosyst. Environ.
Estimating grain yield of maturing rice canopies using high spectral resolution reflectance measurements
Remote Sens. Environ.
The effect of spatial resolution on the accuracy of leaf area index estimation for a forest planted in the desert transition zone
Remote Sens. Environ.
Modeling bidirectional reflectance of forests and woodlands using boolean models and geometric optics
Remote Sens. Environ.
Optimization of endmembers for spectral mixture analysis
Remote Sens. Environ.
New calculation methods of diurnal distribution of solar radiation and its interception by canopy over complex terrain
Ecol. Model.
Prediction of grain protein content in winter wheat (Triticum aestivum L.) using plant pigment ratio (PPR)
Field Crops Res.
Coupling canopy functioning and radiative transfer models for remote sensing data assimilation
Agric. Meteorol.
Estimating impervious surface distribution by spectral mixture analysis
Remote Sens. Environ.
Estimating leaf nitrogen content in corn based on information fusion of multiple-sensor imagery from UAV
Remote Sens
Detecting leaf nitrogen content in wheat with canopy hyperspectrum under different soil backgrounds
Int. J. Appl. Earth Obs. Geoinf.
Hyperspectral indices and model simulation for chlorophyll estimation in open-canopy tree crops
Remote Sens. Environ.
A preliminary precision rice management system for increasing both grain yield and nitrogen use efficiency
Field Crop. Res.
Cited by (21)
Estimating leaf nitrogen and chlorophyll content in wheat by correcting canopy structure effect through multi-angular remote sensing
2023, Computers and Electronics in AgricultureAssessing a soil-removed semi-empirical model for estimating leaf chlorophyll content
2022, Remote Sensing of EnvironmentCitation Excerpt :These innovative strategies can be feasible on ground or UAV platforms where the observation angle and time can be easily adjusted with great flexibility. In addition, higher spatial resolution reflectance imagery can benefit the discrimination of vegetation and soil pixels and thus reduce the fraction of soil (Zarco-Tejada et al., 2013; Jay et al., 2017; Wang et al., 2022). Coupling the 3SV algorithm into these observation strategies could further reduce the soil effects and hence improve the LCC estimation accuracy.