The relations between agronomic parameters, such as Leaf Area Index (LAI), above ground biomass, fresh leaf mass, the contents of chlorophyll and carotenoid and nitrogen content of crops and their reflective spectra have been studied by monitoring their spectral changes in different growing stages (Shibayama and Akiyama, 1989; Blackburn, 1998; Wang et al., 1998; Tang et al., 2004). It provides a basis of growth monitoring and yields estimation for crops by remote sensing (Xu and Wang, 2001). The reflective spectra of crop canopies and leaves and their variables have been estimated their agronomic parameters, in particular pigment and nitrogen contents (Card et al., 1988; Wang et al., 1998), because they are mostly influenced by vegetation pigment (chlorophyll and carotenoid) and covering rate in the visible region and by the structure of canopy and leaf, above ground biomass, biochemical components such as protein, cellulose and starch in the near infrared region (Gausman et al., 1970). In recent years, the hyperspectral remote sensing has a mighty advantages on the vegetation remote sensing and its application because it could be used to quantitatively analyze a feebleness spectral difference (Vane, 1993). It is important for growth monitoring and yields estimation for rice that its spectral reflectance is used to estimate the chlorophyll content and nitrogen status (Takebe et al., 1990; Shibayama and Akiyama, 1991). Former tests have indicated that it is feasible for estimating the contents of leaf nitrogen or its crude protein from the canopy spectra by hyperspectyral remote sensing (Card et al., 1988; Niu et al., 2002). The contents of protein, cellulose and lignin of dry leaf are better to be estimated by their hyperspectra (Kokaly and Clark, 1999; Kokaly, 2001; Curran et al., 2001). There are some reports on estimating LAI, above ground biomass, yields, the contents of nitrogen, cellulose and lignin for rice, but very few for the contents of nitrogen, crude protein and crude starch of panicle and paddy on studying rice by hyperspectra. The nitrogen content or crude protein content is one of the important indices evaluating rice nutritious quality and one of the factors influencing on rice cooking quality (Martin and Fitzgerald, 2002).
In present study, hyperspectral reflectances of the canopy, flag leaf and panicle
of rice are measured by an ASD FieldSpec Pro FR in field and indoor. The
contents of chlorophyll and nitrogen of leaf, panicle and paddy corresponding
to the spectra are determined by the biochemical method. The correlation between
the nitrogen contents of grains (panicle, paddy) and their hyperspectra, the
canopy spectra are analyzed. This study can provide some basis of quality monitoring
of rice by hyperspectral remote sensing.
MATERIALS AND METHODS
Plant material: Rice (Oryza sativa L.) cv. Xiushui 110 (common Japonica rice with about 145 d of whole growing period, erect and compact), Xieyou 9308 (hybrid Indica rice with about 140 days of whole growing period, erect from disheveled), Jiayu 293 (common Indica rice with about 100 days, erect and compact), Jiazao 312 (common Indica rice with about 100d, erect and compact) and Jiazao324 (common Indica rice with about 105 days, erect and compact) were selected for investigation. They were seeded on June 2 and transplanted on June 25. The plants were provided with sufficient water during the whole growing period. Jiayu 293, Jiazao 312 and Jiazao 324 were harvested on Sept. 12, Xiushui 110 and Xieyou 9308 harvested on Oct. 15.
Testing site: A field experiment was conducted during June through October of 2002 at the Experiment Farm, Zhejiang University, Hangzhou (30°14/N, 120°10/E), China. A completely randomized design consisting of 60 plots of area 4.76x4.68 m was used. The single plant was transplanted with spacing by 0.13 m times 0.17 m.
Fertilizer treatments: The 60 plots represented three nitrogen levels (0, 120 and 240 kg N ha-1, symbol N0, N1 and N2, respectively) and four replications. The nitrogen fertilizer consists of 50% base fertilizer, 35% tiller fertilizer and 15% heading fertilizer to come into being no nitrogen fertilizer, proper one and superabundant one. In addition, 533.3 kg Ca (H2PO4)2 ha-1 is base fertilizer, 300 kg KCl ha-1 heading fertilizer. The sandy loam paddy soil has the following properties: total N with 1.15 g kg-1, quick-acting N with 188.5 mg kg-1, total P with 1.21 g kg-1, total K with 72.7 mg kg-1, organic matter with 9.96 g kg-1 and with pH 6.78.
Surveying instrument: The spectral radiometer was ASD (Analytical Spectral
Device) FieldSpec Pro FR with range 350 to 2500 nm. The spectral sampling
interval and resolution are 1.4 and 3 nm in the range 350-1000 nm and 2 and
10 nm in the range 1000-2500 nm, respectively. The spectrophotometer was BECKMAN
DU-600 with 1nm spectral sampling interval and one-ten thousands OD value. The
analytic software of data was SPSS10.0.
Canopy spectrum: The canopy spectral reflectance was determined at 10:00 am to 2:00 pm on a sunny day without cloud and wind at tiller stage, jointing stage, booting stage, heading stage, milking stage and maturing stage, respectively. The detector, about 0.70 m height from the canopy, is vertical down to opposite to a fixed position. A sampling spectrum consists of ten reading and ten sampling spectra averaged to represent the canopy mean spectral reflectance. Dark signal is subtracted and data compared to a standard white reference as spectral measurement.
The plants corresponding to the canopy spectra are sampled for determination of biophysical and biochemical parameters such as leaf, stem and panicle fresh mass, leaf area by leaf area meter, the contents of chlorophyll and carotenoid of leaves and panicles after the canopy spectral reflectance. Then they are killed green for 0.5 h at 105°C and dried at 70°C for determination of their dry mass.
Spectra of leaves and panicles: The leaves and panicles are sampled for spectral measurement and for their concentrations of chlorophyll and carotenoid at tiller stage, jointing stage, booting stage and 4 to 7 day interval after heading. The reflectance of uppermost unfolding leaves on the main stem of 3 group of rice plants with moderate growth selected from each plot are measured in laboratory by LC 1800 at a time. The panicle reflectance is measured in laboratory by ASD FieldSpec Pro FR. The detector fixed by a triangle-rack, about 0.10 m height above the canopy, is vertical down to opposite the center of the sample on a black rubber with about zero reflectance. The angle of VOF was 8°. The light source, 50 W halogenate lamp, is 0.45 m and 70° from the sample surface. The spectral radiometer, controlling computer and light source are linked to a UPS. Dark signal is subtracted and data compared to a standard white reference before spectral measurement. A sampling spectrum consists of ten reading and ten sampling spectra averaged to represent the sample mean spectral reflectance.
Analysis of the concentrations of chlorophyll, carotenoid and Nitrogen: Leaf and panicle chlorophyll contents are analyzed with Arnon method, Leaf and panicle carotenoid contents analyzed with China Standard GB12291-90 method and nitrogen contents determined with Kjeldahl method after the plant materials are over-dried at 70°C for 36 h, ground and digested with H2SO4-H2O2.
Spectra of canopy, leaf and panicle: From Fig. 1a,
the canopy reflectance increases in the red and blue region and decreases in
the infrared region because that the leaf nutrient transfer to panicles and
the lower leaves declined and fall off leading to change of leaf color and its
internal structure, the canopy chlorophyll content in continuous decrease and
LAI decrease after heading. As shown in Fig. 1b and
c, the reflectance of the flag leaf were gradually getting bigger in the
green region and bigger before milking and smaller after milking in the near
infrared region and panicle reflectance are gradually getting bigger in the
visible region and smaller in the near infrared region as the growth stages
went on. But it had a strange increase. This seems to be related to its structure
and change of biochemical components.
Correlative analysis of nitrogen content of rice panicle and paddy relating to the chlorophyll and nitrogen contents of leaves, stems and panicles: From Fig. 2a, the chlorophyll contents of flag leaves are gradually getting bigger from booting stage to heading stage (Aug. 22 ~ Sept. 4) and gradually getting smaller from the beginning of heading stage to the ending of heading stage (Sept. 4 ~ Sept. 8), then bigger from the ending of heading stage to milking stage (Sept. 8 ~ Sept. 20) and finally, smaller up to harvest. From Fig. 2b, the nitrogen contents of flag leaves are gradually getting smaller from the beginning of heading stage, but that of panicles are inverse.
Correlative analysis of nitrogen contents of rice panicle relating to the spectra of flag leaves and panicles: The correlation coefficients between the nitrogen contents (%) of panicles and hyperspectral reflectance of flag leaf and panicle, shown in Fig. 3, are calculated.
||Spectra of Xiushui 110 canopy, flag leaf and panicle at different
development stages after heading (N1). (a): Canopy, (b): Flag leaf and (c):
||Change of the contents of chlorophyll and nitrogen of flag
leaves and panicles for Xiushui 110
||Correlation coefficients between the nitrogen contents of
panicles and spectral reflectance of flag leaf and panicle
||Correlation coefficients between the nitrogen contents of
panicles and paddy and canopy spectral reflectance (a) Panicle and (b) Paddy
There are significant correlation between the nitrogen contents of panicles
and the spectra of flag leaves and panicles themselves in some wavebands. There
are also significant correlation between the nitrogen contents of panicles and
the first derivative spectra of flag leaves and panicles themselves in some
wavebands. The maximum correlation coefficients R between the panicle nitrogen
contents and the spectral reflectance ρ and their first derivative spectra
Dλ of flag leaf and panicles are -0.508** at 534 nm, -0.397** at 1957 nm
and 0.605** at 567 nm, 0.571** at 1654 nm, respectively.
Correlative analysis of nitrogen contents of rice panicle and paddy relating to the canopy spectra: The correlation coefficients between the nitrogen contents (%) of panicles, paddy and canopy spectral reflectance, shown in Fig. 4, are calculated at milking stage and maturing stage.
From Fig. 4a, there are preferably correlation between the nitrogen contents of panicles and their corresponding canopies spectra, where the correlation coefficients attained to positive significant levels in blue range, yellow range and short-wave infrared range. The maximum correlation coefficients R between the panicle nitrogen contents and the canopy spectral reflectance ρ, log (1/ρ), ratio vegetation index RVI and their first derivative spectra are 0.801** at 1452 nm, -0.794** at 2083 nm, -0.866** at ρ725/ρ1483 and 0.830** at 827 nm at milking stage, 0.760** at 1529 nm, -0.763** at 1777 nm, -0.854** at ρ725/ρ1529 and 0.798** at 675 nm at milking stage, respectively.
From Fig. 4b, preferred correlations can be found between the nitrogen contents of paddy and the canopies spectra at milking stage and maturing stage, where the correlation coefficients attained to positive significant levels in visible range and short-wave infrared range. But the correlation is bad at booting stage. The maximum correlation coefficients R between the panicle nitrogen contents and the canopy spectral reflectance ρ, log (1/ρ), ratio vegetation index RVI and their first derivative spectra are 0.798** at 1457 nm, -0.808** at 2076 nm, -0.866** at ρ739/ρ1483 and 0.853** at 827 nm at milking stage, 0.801** at 2413 nm, -0.807** at 2449 nm, -0.847** at ρ736/ρ1526 and 0.758** at 675 nm at milking stage, respectively.
The stepwise multiple regression equation between the nitrogen contents of panicles and paddy and the canopy spectral reflectance ρλ, the first derivative spectra Dλ and ratio vegetation index RVI are shown in Table 2.
From Table 2, it is known that over 80% of change of the nitrogen (or crude protein) contents of panicle and paddy are analyzed by stepwise multiple regression of ρλ, Dλ and RVI of canopy spectra at milking stage and over 75% of change of that at maturing stage. The estimating effect at milking stage is better than that at maturing stage. This is because the formation of nutrition component mainly depends on the milking process for rice seed.
|| Correlation coefficients between nitrogen contents of paddy,
panicles and that of leaves, stems (n = 45)
|**: Stand for significant difference at p<0.01 levels.
The same as below
|| The stepwise multiple regression equation between grain nitrogen
contents and canopies spectra
|NC, ρλ, Dλ
and ρλ1/ρλ2 Stand for grain nitrogen
contents, canopy spectral reflectance, the first derivative spectrum and
ratio vegetation index RVI, respectively. R2, F Stand for the
square of correlation coefficients r and statistical value F, respectively
From above analysis, there are significant correlation between the nitrogen contents of grain and the spectral reflectance ρλ and the first derivative spectra Dλ of flag leaves, panicles and canopy in some wavebands. There are also significant correlation between the nitrogen contents of grain and some RVI of the canopy spectra. The content of nitrogen or crude protein can be estimated by the spectral characteristics of these sensitive wavebands in theoretically. It is known that over 75% of change of the nitrogen (or crude protein) contents of panicle and paddy are analyzed by stepwise multiple regression of ρλ, Dλ and RVI of canopy spectra. The estimating effect by derivative spectra is better than that by original spectra. This is because the absorbing apices of different chemical components, such as protein, starch and cellulose in rice grain, overlap one another in infrared range (Curran, 1989). The derivative spectra, however, is in favor of separating these absorbing characteristics by excavating their change trend.
In addition, the estimation of nitrogen contents of rice grain from canopy
spectra at milking stage is better than that at maturing stage. This is because
the formation of nutrition component mainly depends on the milking process for
rice grain. The nutrition component of rice grain mostly comes from transformation
of nutrition saved in stems and leaves before heading. The most crude protein
has already been transferred to grain from stems and leaves after maturing.
That the contents of nitrogen and chlorophyll in leaf evidently reduce the results
in the relatively weak contribution for leaf chlorophyll is positively correlated
to the nitrogen content by spectral reflectance of canopy.
Of course, the nitrogen contents of rice grains are related not only to fertilization and growth but also to variety, development season and planting methods (Wopereis-Pura et al., 2002; Cheng et al., 2003). It is impossible to accurately estimate the nitrogen contents of rice grains only from its canopy spectra. This study would provide a basic idea and practical method to monitor rice quality by remote sensing.
This study is supported by the National Natural Science Foundation of China (40271078, 10664001), the National 863 Project of China (2007 AA10Z205) and the Nomarch Foundation and Science Foundation of Guizhou Province.