《土壤学》课程教学资源（参考资料）A Leaf Layer Spectral Model for Estimating Protein

A Leaf Layer Spectral Model for Estimating Protein Content of Wheat Grains Chun-Hua Xiao2,Shao-Kun Li'2.,Ke-Ru Wang2,Yan-Li Lu2,Jun-Hua Bai2 Rui-Zhi Xie2,Shi-Ju Gao2,Qiong Wang",and Fang-Yong Wang' Key Labo ogy Agricu structio Crop/ Xinjiang. ence.Ch pees /N Chinese Academy es.Beijing 100081.P.R.China Tel:+86-010-68918891 lishkemail.caas.net.cn Abstract.The spectral signatures of crop canopies in the field provide much information relating morphological or quality characteristics of crops to their optical properties.This experiment was conducted using two winter-whea (Triticum aestivum)cultivars,Jingdong8(with erect leaves)and Zhongyou950 (with horizontal leaves). We analysiced the relation between the directior s and the laeves nitrogen c ontent(LNC).The result sho at th at the on th e upper canopy a d th ower laye r,the spe 60 the analys and the st leaf blade st with root mea and 0.6461 for Zhong ients of de measured and estimated grain p in contents were 0 7467 and 0 7590 Thus grain protein may be reliably predicted from the leaf layer spectral model. Keywords:wheat canopy,leaf distribution,direction spectra,view angle,model. 1 Introduction The spectral characteristics of a crop canopy are determined not only by biophysical and biochemical features and also plant structural attributes.Leaf optical properties are a main factor for canopy spectral. ·Corresponding author. D.Li,Y.Liu,and Y.Chen (Eds.):CCTA 2010.Part IV.IFIP AICT 347.pp.16-29.2011. IFIP International Federation for Information Processing 2011

D. Li, Y. Liu, and Y. Chen (Eds.): CCTA 2010, Part IV, IFIP AICT 347, pp. 16–29, 2011. © IFIP International Federation for Information Processing 2011 A Leaf Layer Spectral Model for Estimating Protein Content of Wheat Grains Chun-Hua Xiao1,2, Shao-Kun Li1,2,3,*, Ke-Ru Wang1,2, Yan-Li Lu2 , Jun-Hua Bai2 , Rui-Zhi Xie2 , Shi-Ju Gao2 , Qiong Wang2 , and Fang-Yong Wang1 1 Key Laboratory of Oasis Ecology Agriculture of Xinjiang Construction Crop/ Center of Crop High-Yield Research, Shihezi 832003, Xinjiang, China 2 Institute of Crop Science, Chinese Academy of Agricultural Sciences / National Key Facility for Crop Gene Resources and Genetic Improvement, NFCRI, Beijing 100081, China 3 Department of Crop Culture, Institute of Crop Science, Chinese Academy of Agricultural Sciences, Beijing 100081, P.R. China Tel.: +86-010-68918891 lishk@mail.caas.net.cn Abstract. The spectral signatures of crop canopies in the field provide much information relating morphological or quality characteristics of crops to their optical properties. This experiment was conducted using two winter-wheat (Triticum aestivum) cultivars, Jingdong8 (with erect leaves) and Zhongyou9507 (with horizontal leaves). We analysiced the relation between the direction spectral characteristics and the laeves nitrogen content(LNC). The result showed that the spectral information observed at the 0° angle mainly provided information on the upper canopy and the lower layer had little impact on their spectra. However, the spectral information observed at 30° and 60° angles reflected the whole canopy information and the status of the lower layer of the canopy had great effects on their spectra. Variance analysis indicated that the ear layer of canopy and the topmost leaf blade made greater contributions to CDS. The predicted grain protein content (GPC) model by leaf layers spectra using 0° view angle was the best with root mean squares (RMSE) of 0.7500 for Jingdong8 and 0.6461 for Zhongyou9507. The coefficients of determination, R2 between measured and estimated grain protein contents were 0.7467 and 0.7599. Thus, grain protein may be reliably predicted from the leaf layer spectral model. Keywords: wheat canopy, leaf distribution, direction spectra, view angle, model. 1 Introduction The spectral characteristics of a crop canopy are determined not only by biophysical and biochemical features and also plant structural attributes. Leaf optical properties are a main factor for canopy spectral. * Corresponding author

A Leaf Layer Spectral Model for Estimating Protein Content of Wheat Grains 17 Crop canopy spectral characteristics represent important info rmation needed to guide crop management.Crop composition and structure are difficult to assess by traditional spectral measurement with vertical canopy direction.The direction spectrum includes a great deal of crop canopy information.Multiangle data significantly improved the accuracy of recovering forest parameters when inverting 3-D optical models (Kimes et al.,2002)but forest vertical structure can not be accurately using angles with the 4 spect nde in the nadi vie all viev a singl band (Kimes ).Directi nal radiances in o near the principal plane of the sun provides informati on that leads to more accurate prediction of canopy structure parameters than from other azimuth planes(Gobron et al.,2000).Canopy emissivity increased with increasing view angle due to the greater proportion of vegetation observed at off-nadir view angles;when the proportion of leaves was lower than that of soil,canopy emissivity grew with increasing view angle (Sobrino et al 2005). It should be noted these models assume a Lambertian behaviour for soil and vegetation st Although vegetation surfaces show a near Lambertian behaviour bare soil surfaces do not and the angular variation on emissivity can not be neglected Lidars,multiangle radiometers,radars and imaging spectrometers have been identified as systems that can capture information in the vertical dimension.This requires a capability to remotely measure the vertical and spatial distribution of forest structural parameters that are needed for more accurate models of energy,carbon.and water flux ntal,and lobal scales.Thus. the utility of vertical w ea t structure note ser ing systems provided age o f the honzonta Iscope,but could not provide the vertical information on biochemical distribution in a crop canopy. thereby reducing the accuracy of measurement.Multi-angle data can increase the precision of forest parameters(Kimes,et al.,2006). The distribution of tissues in a crop canopy has certain characteristics-biochemical distribution is different because of transfer of matter during the growth stage.Leaves in a whe nopy are nder middle,upper layer nd layers(Wan et al.2004).The spectral characteristic differed among canopy leaves because of th different reflection and scatter,so their effect on canopy spectra was different (Wang etal.2004)】 Various biochemical (foliar lignin and nitrogen)and biophysical factors influencing canopy reflectance signatures have been studied in previous works.Information on biochemical paramete ers is important and the multi-angle spectra which provide informationo d directi ons can faci act pre parameters. there are n studies on the relative importar of vertica stribution of wheat leaves that determines canopy reflectance across the shortwave (350-2500nm)spectrum.The contribution of each leaf layer relative to all other factors has also not been adequately determined.Yet,it is the interaction of these factors. including their potential covariance or unique behavior,that must be understood if advances in remote sensing are to be achieved

A Leaf Layer Spectral Model for Estimating Protein Content of Wheat Grains 17 Crop canopy spectral characteristics represent important information needed to guide crop management. Crop composition and structure are difficult to assess by traditional spectral measurement with vertical canopy direction. The direction spectrum includes a great deal of crop canopy information. Multiangle data significantly improved the accuracy of recovering forest parameters when inverting 3-D optical models (Kimes et al., 2002) but forest vertical structure can not be captured accurately using only the 4 spectral bands in the nadir view or all view angles with a single spectral band (Kimes et al.,2006). Directional radiances in or near the principal plane of the sun provides information that leads to more accurate prediction of canopy structure parameters than from other azimuth planes (Gobron et al., 2000). Canopy emissivity increased with increasing view angle due to the greater proportion of vegetation observed at off-nadir view angles; when the proportion of leaves was lower than that of soil, canopy emissivity grew with increasing view angle (Sobrino et al., 2005). It should be noted that these models assume a Lambertian behaviour for soil and vegetation surfaces. Although vegetation surfaces show a near Lambertian behaviour, bare soil surfaces do not and the angular variation on emissivity can not be neglected. Lidars, multiangle radiometers, radars and imaging spectrometers have been identified as systems that can capture information in the vertical dimension. This requires a capability to remotely measure the vertical and spatial distribution of forest structural parameters that are needed for more accurate models of energy, carbon, and water flux over regional, continental, and global scales. Thus, we examined the utility of hyperspectral data for the quantitative characterization of vertical wheat structure. Most remote sensing systems provided an image of the horizontal scope, but could not provide the vertical information on biochemical distribution in a crop canopy, thereby reducing the accuracy of measurement. Multi-angle data can increase the precision of forest parameters (Kimes,et al., 2006). The distribution of tissues in a crop canopy has certain characteristics - biochemical distribution is different because of transfer of matter during the growth stage. Leaves in a wheat canopy are composed of under, middle, upper layer and ear layers (Wang et al.,2004). The spectral characteristic differed among canopy leaves because of the different reflection and scatter, so their effect on canopy spectra was different (Wang et al., 2004). Various biochemical (foliar lignin and nitrogen) and biophysical factors influencing canopy reflectance signatures have been studied in previous works. Information on biochemical parameters is important and the multi-angle spectra which provide information on different directions can facilitate a more exact prediction of biochemical parameters. To date, there are no studies on the relative importance of vertical distribution of wheat leaves that determines canopy reflectance across the shortwave (350–2500nm) spectrum. The contribution of each leaf layer relative to all other factors has also not been adequately determined. Yet, it is the interaction of these factors, including their potential covariance or unique behavior, that must be understood if advances in remote sensing are to be achieved

18 C.-H.Xiao et al In this study,the leaf slice method was used to characterize a wheat canopy,evaluate the spectral response of leaf vertical distribution in the canopy,determine the characteristics of spectral curve for different leaf layers and develop a methodology for predicting the biochemistry of the canopy. 2 Materials and Methods 2.1 Preliminary Experiment A field experiment was conducted in the Experimental Station of the Institute of Crop Science of the Chinese Academy of Agricultural Sciences,Beijing (3957'55"N. 11619'46"E)in the 2005-2006 growing season.The soil was a silt clay loam containing 1.16%organic matter,42.6 mg kg-lalkali-hydrolysable N.26.5 mg kg-1 available ph rus and 139.4 mg kg-1 a rail able pota Two winte at cultivars were used:Jingdong8,an erect leaf plant-type and Zhongyou9507,a lax leaf plant-type.Four N fertilizer (urea,46%N,)treatments were set up with four randomized replications of each cultivar:NO,no N fertilization,N1. 150 kg hm-2 pure N fertilization,N2,300 kg hm-2(rationally fertilized).N3,450 kg hm-2 (excessively fertilized).The N rates were applied in three splits at pre-sowing (50%of the total amount).reviving stage(25%of the total amount)and jointing stag (25%of the total an )All the tr tilized with the same PP205 144kghm-2) K (K20,75 kg h m-2)at pre-sowing The leaf slice method of wheat canopy:whole (plant)wheat samples 0.5m long and 0.8m wide were chosen;based on the vertical distribution of wheat canopy,the samples were measured off the whole wheat canopy (WWC).ear layer of canopy (ELC). inverse first leaves layer of canopy (ILLC-1).inverse second leaves layer of canopy(ILLC-2),inverse third leaves layers of canopy (ILLC-3)and inverse fourth leave s layer of canopy (ILLC-4).Canopy layers were s vered with a sciss ors from the ear layert lowe r layer (Fig.3). 2.2 Measured Traits and Methods All canopy spectral measurements were taken from a height of 1.3 m above ground (the height of the wheat was 90 cm at maturity),under clear sky conditions between 10:00 and 14:00(Beijing local time),usir 258ie0e r(Analytical Spectral Devic r.CO.USA optic operating in spectral region with a sampling interval of 14m between 350 and 1050 nm and 2 nm between 1050 and 2500 nm and with spectral resolutions of 3 nm at 700 nm and1o nm at 1400 nm.A 40 cm x40 cm BaSo4 calibration panel was used for calculation of reflectance.The spectra were measured with view angles of 0.30.60.90.120.150.and 180 to the line vertical to the wheat row using the Sim ple multi-angle sp ctral measurement e quipment(Fig.1)after eve was removed(N2 treatment).The model spectra were measured vith a view angle of

18 C.-H. Xiao et al. In this study, the leaf slice method was used to characterize a wheat canopy, evaluate the spectral response of leaf vertical distribution in the canopy, determine the characteristics of spectral curve for different leaf layers and develop a methodology for predicting the biochemistry of the canopy. 2 Materials and Methods 2.1 Preliminary Experiment A field experiment was conducted in the Experimental Station of the Institute of Crop Science of the Chinese Academy of Agricultural Sciences, Beijing (39º57'55" N, 116º19'46" E) in the 2005-2006 growing season. The soil was a silt clay loam containing 1.16% organic matter, 42.6 mg kg-1alkali-hydrolysable N, 26.5 mg kg-1 available phosphorus and 139.4 mg kg-1 available potassium. Two winter wheat cultivars were used: Jingdong8, an erect leaf plant-type and Zhongyou9507, a lax leaf plant-type. Four N fertilizer (urea, 46%N，) treatments were set up with four randomized replications of each cultivar: N0, no N fertilization, N1, 150 kg hm-2 pure N fertilization, N2, 300 kg hm-2 (rationally fertilized), N3, 450 kg hm-2 (excessively fertilized). The N rates were applied in three splits at pre-sowing (50% of the total amount), reviving stage (25% of the total amount) and jointing stage (25% of the total amount). All the treatments were fertilized with the same amounts of P (P2O5, 144 kg hm-2) and K (K2O, 75 kg hm-2) at pre-sowing. The leaf slice method of wheat canopy: whole (plant) wheat samples 0.5m long and 0.8m wide were chosen; based on the vertical distribution of wheat canopy, the samples were measured off the whole wheat canopy (WWC), ear layer of canopy (ELC), inverse first leaves layer of canopy (ILLC-1), inverse second leaves layer of canopy(ILLC-2), inverse third leaves layers of canopy (ILLC-3) and inverse fourth leaves layer of canopy (ILLC-4). Canopy layers were severed with a scissors from the ear layer to lower layer (Fig. 3). 2.2 Measured Traits and Methods All canopy spectral measurements were taken from a height of 1.3 m above ground (the height of the wheat was 90 cm at maturity), under clear sky conditions between 10:00 and 14:00 (Beijing local time), using an ASD FieldSpec Pro spectrometer (Analytical Spectral Devices, Boulder, CO,USA) fitted with a 258 field of view fiber optics, operating in the 350–2500 nm spectral region with a sampling interval of 1.4 nm between 350 and 1050 nm and 2 nm between 1050 and 2500 nm and with spectral resolutions of 3 nm at 700 nm and10 nm at 1400 nm. A 40 cm ×40 cm BaSO4 calibration panel was used for calculation of reflectance. The spectra were measured with view angles of 0, 30,60,90,120, 150, and 180° to the line vertical to the wheat row using the Simple multi-angle spectral measurement equipment (Fig.1) after every layer was removed (N2 treatment). The model spectra were measured with a view angle of 0°

A Leaf Layer Spectral Model for Estimating Protein Content of Wheat Grains 19 using other trea ments at ear layer (EL).upper leaves layer (ULL)and lower leaves layer(LLL)(Fig.2). Leaf samples from EL,ULL,LLL were taken almost synchronously with the spectral measurements.Measurements were conducted at jointing,heading,anthesis,milking and waxing stages.These samples were oven-dried at 70C and nitrogen content was determined by the Kieldahl technique (Bremneretal.1981)using a B-339 DistillationUnit (BUCHIAnalyticalLtd,Flawil,Switzerland).Wheat grain protein content estimated from the formula:Pro%=6.25x Nitr(%) ar laye pper lav Fig.1.Simple multi-angle spectral measurement Fig.2.Sketch of measured method by layer 2.3 Data Analysis The hyperspectral data were analyzed using the Matlab6.5 software and quantitative data were analyzed using an analysis of variance (ANOVA)procedure. 3 Results 3.1 The Spectral Curves Following Removal of Different Leaf Layers The lower leaves of the canopy changed the spectral reflectivity (Fig.3)with different view angles at 350-700nm 800-1300nm and1400-1800nm This is ortant for 0-10m ntent within the visible wav e band (350-700nm).The tivity of n The waveband (1400-1800nm)provide information on the water content.In this paper,we analyzed the characteristics of visible (350-700nm)and near infrared(800-1300nm)wavebands

A Leaf Layer Spectral Model for Estimating Protein Content of Wheat Grains 19 using other treatments at ear layer (EL), upper leaves layer (ULL) and lower leaves layer(LLL)(Fig.2). Leaf samples from EL,ULL,LLL were taken almost synchronously with the spectral measurements. Measurements were conducted at jointing, heading, anthesis, milking and waxing stages. These samples were oven-dried at 70℃ and nitrogen content was determined by the Kjeldahl technique (Bremneretal.,1981) using a B-339 DistillationUnit (BUCHIAnalyticalLtd, Flawil, Switzerland). Wheat grain protein content estimated from the formula : Pro% =6.25 × Nitr (% ). Fig. 1. Simple multi-angle spectral measurement equipment Fig. 2. Sketch of measured method by layer 2.3 Data Analysis The hyperspectral data were analyzed using the Matlab6.5 software and quantitative data were analyzed using an analysis of variance (ANOVA) procedure. 3 Results 3.1 The Spectral Curves Following Removal of Different Leaf Layers The lower leaves of the canopy changed the spectral reflectivity (Fig.3) with different view angles at 350-700nm, 800-1300nm and 1400-1800nm. This is important for pigment content within the visible wave band (350-700nm). The reflectivity of near infrared (800-1300nm) is influenced by canopy characteristics. The wavebands (1400-1800nm) provide information on the water content. In this paper, we analyzed the characteristics of visible (350-700nm) and near infrared (800-1300nm) wavebands. Lower layer Upper layer 1/2H 1/2H H sun Ear layer

20 C.-H.Xiao et al 0.50 0.50 =0.40 A Jingdong 0.40 0.30 o. A Zhongyou9507 0.20 0.20 0.10 0.00 0.00 300 300 800130018002300 39 二30 -90 —120 10 二180 —30 Fig.3.Spectra of removing the forth leaf layer at different angles for Jingdong8 and Zhongyou9507 wheat .60 0.40 Jingdong8 Zhongyou9507 0.30 0.00 300 300 2300 800 230 Fig.4.Spectra of removing different leaf layers at 60angles for Jingdong8 and Zhongyou9507 Comparing the direction spectral characteristic of the two wheat canopies,the respons of月 v angles more signif t tha that of Zhonyou5thlewayebands:the mat reeetivity o Jingdong8 was 31%higher than that of Zhongyou9507 at 350-700nm and 22.4% higher at 800-1300nm. The spectral curve of doing away with different leaves at the 60 view angle showed that the leaf influenced the curve of visible and near infrared wavebands(Fig.4).The change for Jingdong8 was more significant than for Zhongyou9507;the spectra reflectivity of Jingdong&was 11.higher than that of 350-70 and 26.8%highe at800-130 To further analyze the relation between canopy spectra and leaves,we selected six spectral reflectivities at 450,550,670,980,1090,1200nm. 3.2 Analysis of Spectral Reflectivity of Leaf Layers at Different View Angles As shown in Fig.5,for Jingdong8.the spectral reflec imila r with the if layer remova ILLCreduced the refetivity.The changes in reflectivity at550nm were more

20 C.-H. Xiao et al.            :DYHOHQJWKQP 5HIOHFWLRQ e e e e e e e            :DYHOHQWKQP 5HIOHFWLRQ e e e e e e e Jingdong8 Zhongyou9507 Fig. 3. Spectra of removing the forth leaf layer at different angles for Jingdong8 and Zhongyou9507 wheat                 :DYHOHQJWKQP 5HIOHFWLRQ ,//& ,//& ,//& ,//& ::& (/&            :DYHOHQJWKQP 5HIOHFWLRQ ,//& (/& ,//& ,//& ,//& ::& Jingdong8 Zhongyou9507 Fig. 4. Spectra of removing different leaf layers at 60°angles for Jingdong8 and Zhongyou9507 Comparing the direction spectral characteristic of the two wheat canopies, the spectral response of Jingdong8 with different view angles was more significant than that of Zhongyou9507 at visible and near infrared wavebands; the max reflectivity of Jingdong8 was 31% higher than that of Zhongyou9507 at 350-700nm and 22.4% higher at 800-1300nm. The spectral curve of doing away with different leaves at the 60° view angle showed that the leaf influenced the curve of visible and near infrared wavebands (Fig.4). The change for Jingdong8 was more significant than for Zhongyou9507; the spectral reflectivity of Jingdong8 was 11.2% higher than that of Zhongyou9507 at 350-700nm and 26.8% higher at 800-1300nm. To further analyze the relation between canopy spectra and leaves, we selected six spectral reflectivities at 450, 550, 670, 980, 1090, 1200nm. 3.2 Analysis of Spectral Reflectivity of Leaf Layers at Different View Angles As shown in Fig.5, for Jingdong8, the spectral reflectivity of the whole canopy was similar with the leaf layer removal treatment, ILLC-4; doing away with ELC and ILLC-1 reduced the reflectivity. The changes in reflectivity at 550nm were more

A Leaf Layer Spectral Model for Estimating Protein Content of Wheat Grains 2 WWC ELc 0 30 60 120 50 00 090120150180 -140 ILLC-1 111C2 10 060901801501 455+550+650-980+1090-+1200 0609012015010 050609012010180 45+ 55 20 45550680-980-1090→-10 Fig.5.Spectral characteristics following removal of different leaf layers in Jingdong8 obvious than at 450 and 670nm.In visible wave bands,changes were evident for 980 ed wave bands.At different view angles.the reflectivity o e lowest and with increasing distance from the vertical measurement,the reflectivity increased,especially in visible wave bands than in near infrared.The reflectivity changes due to ILLC-I and ELC were more obvious than those of other leaf layers. The reflectivity changes due to LLC-3 at 1090nm were 8.8%at 0 view angle and 0had more informtion on te o ange wa 150 ew angle

A Leaf Layer Spectral Model for Estimating Protein Content of Wheat Grains 21 WWC ELC ILLC-1 ILLC-2 Fig. 5. Spectral characteristics following removal of different leaf layers in Jingdong8 obvious than at 450 and 670nm. In visible wave bands, changes were evident for 980, 1090, 1200nm in near infrared wave bands. At different view angles, the reflectivity of the 90° view angle was the lowest and with increasing distance from the vertical measurement, the reflectivity increased, especially in visible wave bands than in near infrared. The reflectivity changes due to ILLC-1 and ELC were more obvious than those of other leaf layers. The reflectivity changes due to LLC-3 at 1090nm were 8.8% at 0° view angle and 48.6% for ILLC-1. At 90° view angle, the changes were 5.6 and 40.7%; at 30°, they were 12.9 and 27.7% and at 60° they were 12.2 and 34%, respectively. The change was less pronounced at the 180, 150 and 120° view angles. Compared to the 90° view angle, the reflectivity at 30 and 60° had more information on the lower leaves, which were

C.-H.Xiao et al important for reflectivity of the canopy.The reflectivity at 0 contained very important information on upper leaves. The canopy spectral reflectivity of Zhongyou9507 was lower than that of Jingdong8 (Fig.6):the reflectivity due to ILLC-3at view angle changed 4.4%.while that due to ILLC-1 changed14.view angle,the 5.9and46.8% at ,they wer 16.and 4.%and at 60 they were 16.8a thetraditionviewanglethe ower laves weremportant for canopy spectraa30 and 60 view angles;the upper leaves were important at 0 view angle although the upper leaves of Zhongyou9507 had less influence on canopy spectra than those of Jingdong8. WWC flc 主一士主 03060 9013D15030 03060 012015080 980-4 00 -1300 -60-阅0◆-109 ILLC-2 18 -455550一0009801090+120 。一4★—550一★一980一1090+一120 ILLC-3 罪 ILLC-4 北本上士 030 609012015018 0060901010100 一455 一980一1090+1200 Fig.6.Spectral characteristics following removal of different leaf layers in Zhongyou9507

22 C.-H. Xiao et al. important for reflectivity of the canopy. The reflectivity at 0° contained very important information on upper leaves. The canopy spectral reflectivity of Zhongyou9507 was lower than that of Jingdong8 (Fig.6) at1090nm; the reflectivity due to ILLC-3 at 0° view angle changed 4.4%, while that due to ILLC-1 changed14.7%; at 90° view angle, the changes were 5.9 and 46.8%; at 30°, they were 16.9 and 44.4% and at 60° they were 16.8 and 41.2%. Compared with the traditional 90° view angle, the lower leaves were important for canopy spectra at 30 and 60° view angles; the upper leaves were important at 0° view angle although the upper leaves of Zhongyou9507 had less influence on canopy spectra than those of Jingdong8. WWC ELC ILLC-2 ILLC-1 ILLC-3 ILLC-4 Fig. 6. Spectral characteristics following removal of different leaf layers in Zhongyou9507

A Leaf Layer Spectral Model for Estimating Protein Content of Wheat Grains 3.3 Analysis of Variance(ANOVA)of Spectral Reflectivity for Different Leaf Layers The ANOVA of mean reflectivities across six wavelengths showed that the reflectivity of Jingdong&was lower than that of Zhongyou9507 (Table 1).For Jingdong8,the response of canopy spectra due to ILLC-1 and ELC were significant at the 0view angle;at 30 as for6o°，the effects of ELC.LLC-l,LLC-2 and ILLC-3 were significant:at90°，thc ELC and ILLC-1 were important and at 120.150 and 180 view angles.the ELC and ILLC-1 were significant.The ELC and ILLC-1 were important for the influence of lower leaves were les s.For the the under leaves were less influential;at 30 and 60,the response of lower leaves were significant.The canopy spectral responses of different leaves for Zhongyou9507 differed from that of Jingdong8;at 0 and 90 view angles,the ELC and ILLC-1 effects were significant at 30 and 60,the ELC,ILLC-1,ILLC-2 effects were significant. At the same view angle.the ELC and ILLC-1 were more important than the lower leaves The spectral response of lo leaves was related to the view angle.With vertica and horizontal measurements,the influence of lower leaves were less than that of othe view angles Table 1.Canopy spectral characteristic of different leaf layers of Jingdong8 and Zhongyou9507 Average value Angles° Treatments Jingdong8 Zhongyou9507 0 wWC 0.2254 A 0.1657 A ILLC-3 0.2110 9 A 0.1643 A Π1C4 02046 AB 0.1662 ILLC-2 0.1974 ab AB 0.1573 A L.1 0.1690 bc BC 0.1354 h ELC 0.1508 c 0.1308 b B 30 wwC 0.2265 A 0.1713 A ILLC-4 0.2196 ab 0.1646 AB ILLC-3 0.1951 be AB 0.1619 ab ILLC-2 0.1886 AB 0.1398 be BC ILLC.I 0.1591 BC 0.1276 cd. ELC 0.1487 C 0.1143 d WWC 0.2064 a A 0.1784 A ILLC-4 0.2021 a A 0.1553 A ILLC.3 0.1816 ab AB 0.1503 ab AB ILLC-2 0.1645 bc ABC 0.1323 ABO ILLC-1 0.1412 BC 0.1052 c BC ELC 0.1225 C 0.0974 c C

A Leaf Layer Spectral Model for Estimating Protein Content of Wheat Grains 23 3.3 Analysis of Variance (ANOVA) of Spectral Reflectivity for Different Leaf Layers The ANOVA of mean reflectivities across six wavelengths showed that the reflectivity of Jingdong8 was lower than that of Zhongyou9507 (Table 1). For Jingdong8, the response of canopy spectra due to ILLC-1 and ELC were significant at the 0° view angle; at 30° as for 60°, the effects of ELC, ILLC-1, ILLC-2 and ILLC-3 were significant; at 90°, the ELC and ILLC-1 were important and at 120, 150 and 180° view angles, the ELC and ILLC-1 were significant. The ELC and ILLC-1 were important for canopy spectra and the influence of lower leaves were less. For the view angles of 0°, 90°, 120°, 150°, 180°), the under leaves were less influential; at 30 and 60°, the response of lower leaves were significant. The canopy spectral responses of different leaves for Zhongyou9507 differed from that of Jingdong8; at 0 and 90° view angles, the ELC and ILLC-1 effects were significant at 30 and 60°, the ELC, ILLC-1, ILLC-2 effects were significant. At the same view angle, the ELC and ILLC-1 were more important than the lower leaves. The spectral response of lower leaves was related to the view angle. With vertical and horizontal measurements, the influence of lower leaves were less than that of other view angles. Table 1. Canopy spectral characteristic of different leaf layers of Jingdong8 and Zhongyou9507 Average value Angles° Treatments Jingdong8 Zhongyou9507 WWC 0.2254 a A 0.1657 a A ILLC-3 0.2110 a A 0.1643 a A ILLC-4 0.2046 a AB 0.1662 a A ILLC-2 0.1974 ab AB 0.1573 a A ILLC-1 0.1690 bc BC 0.1354 b B 0 ELC 0.1508 c C 0.1308 b B WWC 0.2265 a A 0.1713 a A ILLC-4 0.2196 ab A 0.1646 a AB ILLC-3 0.1951 bc AB 0.1619 ab AB ILLC-2 0.1886 c AB 0.1398 bc BC ILLC-1 0.1591 d BC 0.1276 cd C 30 ELC 0.1487 d C 0.1143 d C WWC 0.2064 a A 0.1784 a A ILLC-4 0.2021 a A 0.1553 ab A ILLC-3 0.1816 ab AB 0.1503 ab AB ILLC-2 0.1645 bc ABC 0.1323 bc ABC ILLC-1 0.1412 bc BC 0.1052 c BC 60 ELC 0.1225 c C 0.0974 c C

24 C.-H.Xiao et al Table 1.(continued) Average value Angleso Treatments Jingdong8 Zhongyou9507 90 wwc 0.1588 A0.1766 a A ILLC-4 0.1726 AB 0.1657 a AB ⅡLC.3 0.1451 ab AB 0.1532 abc AB ILLC-2 0.1292 ab AB 0.1276 abed AB ELC 0.1081 B 0.1137 bed B ILLC-1 0.1040 b B 0.1084 d B 120 wwc 01645 a A 0.1765 A ILLC-4 0.1642 0.1692 6 48 ILLC-3 0.1511 B 0.1616 AB ILLC-2 0.1341 abc AB 0.1432 bed AB ⅡLC.1 0.1139 be AB 0.1255 AB ELC 0.1051 B 0.1325 w B 150 IIC4 0.1734 a 0.1679 ILLC-3 0.1637 ab AB 0.1692 Π1C.2 01547 AB 0.1568 A wwC 0.1454 abc ABC 0.1672 % AB ILLC-1 0.1325 BC 0.1515 c AB ELo 0.1166 d C 0.1555 B 180 WWC 0.2254 A 0.1678 ILLC-4 0.1879 山 B 0.1599 ILLC-3 0.1883 ab ABC 0.1555 A ILLC-2 0.1725 be BC 0.1464 ILLC-1 0.1388 c BC 0.1461 b B ELc 0.1360 c C0.1537 6 B Note:Means followed by different lower case letters differ significantly at P<0.05;those followed by different upper case letters differ significantly at P<0.01. 3.4 Correlation of Spectral Parameters of Leaf Layers with Leaf Nitrogen Content(LNC) The correlation between canopy spectral parameters,spectral grads and corresponding nitrogen content in different leaf layers (Table 2)was significant between ARVI[670. 890]and leaf layer grads at the flowering stage for Jingdong8 and milky stage for Zhongyou9507

24 C.-H. Xiao et al. Table 1. (continued) Average value Angles° Treatments Jingdong8 Zhongyou9507 WWC 0.1588 a A 0.1766 a A ILLC-4 0.1726 a AB 0.1657 a AB ILLC-3 0.1451 ab AB 0.1532 abc AB ILLC-2 0.1292 ab AB 0.1276 abcd AB ELC 0.1081 b B 0.1137 bcd B 90 ILLC-1 0.1040 b B 0.1084 d B WWC 0.1645 a A 0.1765 a A ILLC-4 0.1642 a A 0.1692 ab AB ILLC-3 0.1511 ab AB 0.1616 abc AB ILLC-2 0.1341 abc AB 0.1432 bcd AB ILLC-1 0.1139 bc AB 0.1255 d AB 120 ELC 0.1051 c B 0.1325 cd B ILLC-4 0.1734 a A 0.1679 ab A ILLC-3 0.1637 ab AB 0.1692 a A ILLC-2 0.1547 abc AB 0.1568 bc A WWC 0.1454 abc ABC 0.1672 ab AB ILLC-1 0.1325 cd BC 0.1515 c AB 150 ELC 0.1166 d C 0.1555 c B WWC 0.2254 a A 0.1678 a A ILLC-4 0.1879 ab AB 0.1599 a A ILLC-3 0.1883 ab ABC 0.1555 a A ILLC-2 0.1725 bc BC 0.1464 a A ILLC-1 0.1388 c BC 0.1461 b B 180 ELC 0.1360 c C 0.1537 b B Note: Means followed by different lower case letters differ significantly at P< 0.05; those followed by different upper case letters differ significantly at P<0.01. 3.4 Correlation of Spectral Parameters of Leaf Layers with Leaf Nitrogen Content (LNC) The correlation between canopy spectral parameters, spectral grads and corresponding nitrogen content in different leaf layers (Table 2) was significant between △RVI[670, 890] and leaf layer grads at the flowering stage for Jingdong8 and milky stage for Zhongyou9507

A Leaf Layer Spectral Model for Estimating Protein Content of Wheat Grains Table 2.Correlatio coefficients between nitrogen contents at different layers and spectral (SC) LSCP stage Cultivar ANDVE 人Rn ontent △680mm 18571210 (570.531)(550.540 △OSA7 1670.890 Jingdong8 ULNC -0.2058 0.6835 0.1929 -0.4270 0.6850° 0.6900 Anrhesis LLNC 0.2016 0.6437 02082 0.3300 -0.5669 0.5361 0.2779 0.6431 0.5443 0.5240 -03256 0.7710 Zhongyou ULN 0.062 -0.278 0.050 -0.6561 -0.0321 0.6171 9507 0.59 0022 0.214 0.37 02041 01320 04570 9507 LNC 04674 -04111 05394 06751" 02721 05118 △LNC 0.1453 0.5269 06252 0.4591 -0.1190 0.7888 and:significant correlations and 0.01 probability levels,respectively.ULNC:upper layer nitrogen content:LLNC:lower layer nitrogen content:ALNC:the difference between ULNC and LLNC.LSCP:spectral characteristics of leaf layers.NDVI:normalized difference vegetation index;PRI:pigment ratio index;PPR:plant pigment ratio:OSAVI:optical soil adjusted index:RVI ratio of vegetation index 3.5 Model Relating Grads of Spectral Parameter and Those of Nitrogen Content across Leaf Layers The following models were developed from the siginificant correlations between grads of spectral parameter and those of nitrogen content in leaf layers: Flowering stage of Jingdong8: △LNC(%)=35.2842△RVI1+0.3822 R=0.8267**,n=24 (1) Milking stage of Zhongyou9507: △LNC(%)=-8.0692△RVI+1.5506 R2-0.8099**,n=24 (2) 3.6 Leaf Layer Spectral Model for Protein Content in Wheat Grains A prediction model for GPC based on spectra of leaf layers was developed as follows a)Jingdong8 at the flowering stage: GPC1%)=12.0168△RVM+15.6975 R2=0.7727*,n=24 (3)

A Leaf Layer Spectral Model for Estimating Protein Content of Wheat Grains 25 Table 2. Correlation coefficients between nitrogen contents at different layers and spectral characteristics ( SC ) Growth LSCP stage Cultivar nitrogen content 680nm NDVI [857,1210] PRI (570,531) PPR (550,540) OSAVI RVI [670,890] ULNC -0.2058 0.6835* 0.1929 -0.4270 -0.6850* 0.6900* LLNC -0.2016 0.6437* 0.2082 -0.3300 -0.5669 0.5361 Jingdong8 LNC -0.2779 0.6431* 0.5443 -0.5240 -0.3256 0.7710** ULNC 0.0627 -0.2781 0.0508 -0.6561* -0.0322 0.6171* LLNC 0.5923 -0.0223 0.2141 -0.5069 0.1373 0.5710 Anrhesis Zhongyou 9507 LNC -0.2234 -0.4238 -0.6304* -0.4990 -0.1748 -0.6335* ULNC 0.3290 0.5788 0.6957* -0.5719 0.4431 0.5831 LLNC 0.3656 0.5690 0.4514 -0.5913 0.4368 0.6097* Jingdong8 LNC 0.5610 0.3882 -0.4454 -0.3284 -0.5557 0.6627* ULNC 0.2041 0.1279 0.1432 -0.5355 0.2822 0.4570 LLNC 0.4674 -0.4111 -0.5394 0.6751* -0.2721 0.5118 Milky ripe Zhongyou 9507 LNC -0.1453 -0.5269 -0.6252* -0.4591 -0.1190 -0.7888** * and ** : significant correlations at 0.05 and 0.01 probability levels, respectively. ULNC: upper layer nitrogen content; LLNC: lower layer nitrogen content; △LNC: the difference between ULNC and LLNC. LSCP: spectral characteristics of leaf layers. NDVI: normalized difference vegetation index; PRI: pigment ratio index; PPR: plant pigment ratio; OSAVI: optical soil adjusted index; RVI: ratio of vegetation index. 3.5 Model Relating Grads of Spectral Parameter and Those of Nitrogen Content across Leaf Layers The following models were developed from the siginificant correlations between grads of spectral parameter and those of nitrogen content in leaf layers: Flowering stage of Jingdong8: △LNC (%) = 35.2842△RVI+0.3822 R2 =0.8267**, n=24 (1) Milking stage of Zhongyou9507: △LNC (%) = -8.0692△RVI+1.5506 R2 =0.8099**, n=24 (2) 3.6 Leaf Layer Spectral Model for Protein Content in Wheat Grains A prediction model for GPC based on spectra of leaf layers was developed as follows: a) Jingdong8 at the flowering stage: GPC (%) = 12.0168△RVI + 15.6975 R2 =0.7727**, n=24 (3)