

small (250x250 max)
medium (500x500 max)
Large
Extra Large
large ( > 500x500)
Full Resolution


RED EDGE AS A POTENTIAL INDEX FOR DETECTING DIFFERENCES IN PLANT NITROGEN STATUS IN WINTER WHEAT By YUMIKO KANKE Bachelor of Plant and Soil Science Oklahoma State University Stillwater, Oklahoma 2009 Submitted to the Faculty of the Graduate College of the Oklahoma State University in partial fulfillment of the requirements for the Degree of MASTER OF SCIENCE December, 2009 ii RED EDGE AS A POTENTIAL INDEX FOR DETECTING DIFFERENCES IN PLANT NITROGEN STATUS IN WINTER WHEAT Thesis Approved: Dr. William Raun Thesis Adviser Dr. John Solie Dr. Randy Taylor Dr. A. Gordon Emslie Dean of the Graduate College iii ACKNOWLEDGMENTS First of all I would like to thank Dr. William R. Raun who gave me many opportunities in Oklahoma, and also abroad. Dr. Raun was the person who taught me how science should be and what wisdom is. I also appreciate time with Dr. John Solie, Dr. Marvin Stone and Dr. Randy Taylor who gave thoughtful advice and support for my academic career especially from an engineering prospective. I thank soil fertility friends; Jacob Vossenkemper, Jerry May, Guilherme Torres, Emily Rutto, Birehane Desta, and Jonathan Kelly who are members of the soil fertility team. iv TABLE OF CONTENTS Chapter Page I. ABSTRACT ......................................................................................................... 1 II. INTRODUCTION ................................................................................................ 2 III. REVIEW OF LITERATURE ................................................................................. 3 IV. OBJECTIVES AND HYPOTHES ......................................................................... 7 V. MATERIALS AND METHODS ............................................................................ 8 Location .................................................................................................... 8 Data Measurements.................................................................................. 9 Spectral Calculation .................................................................................10 VI. RESULTS ..........................................................................................................14 Stillwater, 20072008 ...............................................................................14 Stillwater, 20082009 ...............................................................................16 Summary for Stillwater .............................................................................17 Perkins, 20072008 ..................................................................................18 Perkins, 20082009 ..................................................................................19 Summary for Perkins ...............................................................................20 VII. DISCUSSION AND CONCLUSIONS .................................................................22 VIII. REFERENCES ..................................................................................................26 IX. APPPENDIX ......................................................................................................30 v LIST OF TABLES Table Page 1. Analysis of variance for GreenSeeker NDVI at different N rates and different growth stages, Stillwater, OK, 20072008. ..................................................................31 2. Analysis of variance for SPAD readings at different N rates and different growth stages, Stillwater, OK, 20072008. ..................................................................32 3. Analysis of variance for rededge position (REP) at different N rates and different growth stages, Stillwater, OK, 20072008. ......................................................33 4. Analysis of variance for AREA2 at different N rates and different growth stages, Stillwater, OK, 20072008. ..............................................................................34 5a. Correlation matrices and simple correlation coefficients, Feekes growth stage 4, Stillwater, OK, 20072008. ..............................................................................35 5b. Correlation matrices and simple correlation coefficients, Feekes growth stage 5, Stillwater, OK, 20072008.….....................................................................….... 36 5c. Correlation matrices and simple correlation coefficients, Feekes growth stage 7, Stillwater, OK, 20072008.….....................................................................….... 37 5d. Correlation matrices and simple correlation coefficients, Feekes growth stage 10, Stillwater, OK, 20072008.….....................................................................….... 38 6. Analysis of variance for GreenSeeker NDVI at different N rates and different growth stages, Stillwater, OK, 20082009 ...................................................................39 7. Analysis of variance for SPAD readings at different N rates and different growth stages, Stillwater, OK, 20082009 ...................................................................40 8. Analysis of variance for REP at different N rates and different growth stages, Stillwater, OK, 20082009. ..............................................................................41 9. Analysis of variance for AREA2 at different N rates and different growth stages Stillwater, OK, 20082009. ..............................................................................42 10a. Correlation matrices and simple correlation coefficients, Feekes growth stage 4, Stillwater, OK, 20082009. ..............................................................................43 10b. Correlation matrices and simple correlation coefficients, Feekes growth stage 5, Stillwater, OK, 20082009. ..............................................................................44 10c. Correlation matrices and simple correlation coefficients, Feekes growth stage 7, Stillwater, OK, 20082009. ..............................................................................45 10d. Correlation matrices and simple correlation coefficients, Feekes growth stage 10, Stillwater, OK, 20082009. ..............................................................................46 11. Analysis of variance for GreenSeeker NDVI at different N rates and different growth stages, Perkins, OK, 20072008......................................................................47 vi 12. Analysis of variance for SPAD readings at different N rates and different growth stages, Perkins, OK, 20072008......................................................................48 13. Analysis of variance for REP at different N rates and different growth stages,Perkins, OK, 20072008. ............................................................................................... 49 14. Analysis of variance for AREA2 at different N rates and different growth stages, Perkins, OK, 20072008. .................................................................................50 15a. Correlation matrices and simple correlation coefficients, Feekes growth stage 4, Perkins, OK, 20072008. .................................................................................51 15b. Correlation matrices and simple correlation coefficients, Feekes growth stage 5, Perkins, OK, 20072008. .................................................................................52 15c. Correlation matrices and simple correlation coefficients, Feekes growth stage 7, Perkins, OK, 20072008. .................................................................................53 15c. Correlation matrices and simple correlation coefficients, Feekes growth stage 10, Perkins, OK, 20072008. .................................................................................54 16. Analysis of variance for GreenSeeker NDVI at different N rates and different growth stages, Perkins, OK, 20082009. .....................................................................55 17. Analysis of variance for SPAD readings at different N rates and different growth stages, Perkins, OK, 20082009. .....................................................................56 18. Analysis of variance for REP at different N rates and different growth stages, Perkins, OK, 20082009. .................................................................................57 19. Analysis of variance for AREA2 at different N rates and different growth stages, Perkins, OK, 20082009. .................................................................................58 20a. Correlation matrices and simple correlation coefficients, Feekes growth stage 4, Perkins, OK, 20082009. .................................................................................59 20b. Correlation matrices and simple correlation coefficients, Feekes growth stage 7, Perkins, OK, 20082009. .................................................................................60 20c. Correlation matrices and simple correlation coefficients, Feekes growth stage 10, Perkins, OK, 20082009……………………………………………………………..61 vii LIST OF FIGURES Page 1. The rededge position determined the linear method. ................................................10 2. The visula description of Area 1 and 2 .......................................................................13 3. Plot plan for Experiment 222, Stillwater, OK. ............................................................62 4. Plot plan for N & P Study, Perkins, OK. .....................................................................63 5. GreenSeeker NDVI plotted against N rates, Stillwater, OK 20072008 ......................64 6. SPAD reading plotted against N rates, Stillwater, OK 20072008. .............................65 7. REP plotted against N rates, Stillwater, OK 20072008. ............................................66 8. AREA2 plotted against N rates, Stillwater, OK 20072008. ........................................67 9. GreenSeeker NDVI plotted against N rates, Stillwater, OK 20082009. .....................68 10. SPAD reading plotted against N rates, Stillwater, OK 20082009. ...........................69 11. REP plotted against N rates, Stillwater, OK 20082009. ..........................................70 12. AREA2 plotted against N rates, Stillwater, OK 20082009. ......................................71 13. Shape of rededge and its position at Feekes 4 in the 0 kg N/ha and 135 kg N/ha plots, Stillwater, OK, 2008 ...............................................................................72 14.The first derivative reflectance in the 0 kg N/ha and 135 kg N/ha plots, Stillwater, OK, 2008. ...............................................................................................................73 15. Filtering the first derivative reflectance by averaging over 10 nm widths in the 0 kg N/ha and 135 kg N/ha plots, Stillwater, OK, 2008. ...........................................74 16. Filtering the first derivative reflectance by weight averaging over 10 nm widths in the 0 kg N/ha and 135 kg N/ha plots, Stillwater, OK, 2008. ...................................75 17. Filtering the first derivative reflectance by using trend fitting in the 0 kg N/ha and 135 kg N/ha plots, Stillwater, OK, 2008. .................................................................76 18. Visual image of winter wheat at Feekes 4, 0 kg N/ha and 135 kg N/ha under conventional tillage, Stillwater, OK, 2008. .......................................................77 19. Visual image of winter wheat at Feekes 4, 0 kg N/ha and 168 kg N/ha under notill system, Perkins, OK, 2008. .............................................................................78 20. Linear relationship between REP and GreenSeeker NDVI at Feekes 4 at different N rates, Stillwater, OK, 20072009......................................................................79 21.Linear relationship between GSNDVI and REP under the conbentional tillage or notill system at all growth stages, Stillwater and Perkins, OK, 2008200…………… 80 22. Linear relationship between REP and GreenSeeker NDVI at Feekes 4 and Feekes 5, Stillwater, OK, 20072009………………………………………………………….. 81 23. Linear relationship between SPAD and GreenSeeker NDVI at Feekes 4 and Feekes 5, Stillwater, OK, 2007009………………………………………………………. 82 24. Linear relationship between SPAD and REP at Feekes 4 and Feekes 5, Stillwater, OK, 20072009………………………………………………………………………. 83 1 ABSTRACT Alternative indices have been proposed to better detect N status in cereals. Normalized difference vegetation index (NDVI) is a key input for the OSU Nrate algorithm that is now in commercial use. However, under certain conditions NDVI has low sensitivity; therefore rededge position (REP) has been tested as a new index. The objective of this paper was to determine whether the rededge index could be useful for detecting differences in N status for winter wheat compared to NDVI. Field research was conducted at various locations in Oklahoma. Indirect measurements were collected by using a GreenSeeker sensor, a passive spectrometer, and the SPAD meter to measure plant N status in 4 different N treatments. NDVI was sensitive to plant N response as well as different plant growth stages but the sensitivity tended to decrease as N rate increased. REP sensitivity to plant N response increased as N rate increased and with advancing plant growth. NDVI and REP were linearly correlated at all growth stages (r2=0.79). REP and SPAD meter readings were also highly correlated (r2= 0.62) as were NDVI and SPAD (r2=0.56). REP sensitivity was expected to be higher than NDVI sensitivity at high plant biomass or N concentration, but this was not obvious for winter wheat. Concerning the technical problems of measuring REP in the field, further engineering will be required to evaluate REP as an alternative N index in agronomic fields. 2 INTRODUCTION Nitrogen (N) is one of the major limiting mineral nutrients for plant growth. Due to the continuous increase in fertilizer costs and growing environmental concerns associated with fertilizer use, application of N fertilizer according to plant need becomes increasingly popular due to its potential for increasing NUE and reducing input costs. To determine the optimum N rate based on plant need, optical sensing technologies have been developed to detect N status in plants. Normalized difference vegetative index (NDVI) computed from optical sensing is one of the most widely used indices for the evaluation of plant N status. There are some drawbacks to NDVI, however. It has been reported to have low sensitivity at high chlorophyll content or abundant biomass. Gitelson et al. (2002) listed several possible reasons for low senstivity of NDVI. Decreased NIR reflectance was associated with changes in leaf orientation from one growth stage to the next, reduction in chlorophyll content at senescence, and increasing soil moisture. This also causes the poor estimation of biomass once soil is covered by vegetation (Clevers and Jongschaap, 2001). To overcome these limitations, wavebands called “rededge” were employed as new spectra used to evaluate plant N conditions. Rededge wavebands are between RED and NIR and correspond to wavelengths between 680 nm to 740 nm. These bands were shown to have greater sensitivity at higher chlorophyll content, which was detected as greener biomass. Oklahoma State University (OSU) has been developing algorithms for N fertilization for various crops since the early 1990’s. The algorithms are based on the use of an optical, active light, handheld GreenSeekerTM sensor which detects the fraction of light being reflected off of the plant. The OSU algorithm uses GreenSeekerTM NDVI values as the key input for calculation of the optimum midseason N fertilization rate. Due to reported limitations of NDVI at high chlorophyll concentrations or plant biomass, it was necessary to evaluate the potential of red edge for detecting N differences.. 3 REVIEW OF LITERATURE Raun and Solie (1999) estimated worldwide cereal NUE to be approximately 33 %. Efficient N fertilization based on seasonspecific plant need is one of the methods used to increase NUE. Optical, remote sensing technologies for assessing plant N status midseason have promoted more efficient utilization of N. The optical realtime active light GreenSeekerTM sensor developed by NTech Industries, Inc., in conjunction with OSU, allows for the determination of plant responses to N fertilization. NDVI is used to estimate the Response Index (RI) and yield potential (YP) which facilitates more informed decision making associated with midseason N rate application (Raun et al., 2005). This technology has shown to deliver up to US$25 per hectare (US$10 per acre) in fertilizer savings to winter wheat producers. NDVI is highly correlated with chlorophyll content which is directly associated with photosynthetic capacity (Myneni et al., 1995). Also, NDVI is correlated with above ground biomass as well as N uptake in grain (Solie et al., 1996; Boelman et al., 2003). However, NDVI has been reported to have low sensitivity at high chlorophyll content or abundant biomass due to variation in leaf orientation, as well as the effect of soil background (Hatfield et al., 1984; Rondeaux et al., 1996; Elmore et all., 2000). To overcome these limitations, rededge has been proposed as an alternative index to sense plant conditions. Rededge is referred to as the region defined by 680 nm to 740 nm, which is between RED and NIR wavelengths. The main difference between NDVI and rededge is that NDVI reflectance stays horizontal even though the intensity changes and it is the ratio of reflectance, but rededge encumbers the study of inflection positions on the slope of reflectance, or so called rededge position (REP). Due to the slope changing and is expanded, rededge can give more information by analyzing the shape and the maximum point of the first derivative reflectance on the rededge waveband. 4 Several methods have been developed to find the REP. One is detecting the inflection point which is the maximum slope between RED and NIR (Clevers and Jongschaap, 2006). It uses the first derivative analysis to detect REP (Chen and Elvidge, 1993). Another method is the linear method. This method was introduced by Guyot and Baret (1988). The estimation of reflected maximum inflection point (Rre) is calculated by using reflectance of 670 nm and 780nm. Then the wavelength of rededge position is calculated as 700+40((Rre R700)/(R740R700)). Shafri et al (2006) reported this linear model as an estimator for REP and easier but has more soil background noise. With the linear method, the overestimation of REP 10 nm wide compared with the first derivative method was found. (Dwason and Curran, 1998) However, Dwason and Curran also reported that both methods were correlated at different chlorophyll content and the correlation coefficient of REP determined by different methods was high (R2>0.99). Mutanga and Skidmore (2007) drew attention to the double points for the rededge especially in the high N treated plant using the first derivative system. If double REP exists, the linear method is not appropriate to detect the rededge position. Cho and Skidmore (2006) also developed another method where REP is determined by the intersection of the farred and NIR lines on the first derivative reflectance. REP determined by this method increased the linear relationship with N concentration compared with the first derivative method or linear method. Clevers and Jongschaap (2006) summarized sensitivity analysis for REP. REP is influenced by chlorophyll content, leaf mesophyll structure, and LAI. On the other hand, leaf orientation, solar angle and soil background had a small influence on REP. Also by combining plant growth models with REP, they increased the estimation of yield in sugarcane. Deviation of rededgeshifts and rededgeshape associated with chlorophyll content was studied by Collin (1978), who discovered rededge shifts associated with different growth stages. Maximum derivative rededge parameters were found to be highly correlated with C/N ratio in corn leaves (Middletion et al., 2002). Rededge has the potential for accurate determinations of variation in plant biomass and chlorophyll content in winter wheat (Boochs and Kupfer, 1990). Also, the rededge index enables the user to pick up the variation of red (703nm) amplitude corresponding to different N rate responses in two wheat varieties (Boochs and Kupfer, 1990). Mutanga and Skidmore (2004) investigated band depth analysis methodology which used red edge 5 bands for better biomass estimation at high canopy density. This research showed high correlation between biomass and band depth (0.81), band depth ratio (0.83), and normalized band depth index (0.85) but the coefficient between biomass and NDVI was low (0.31) . Following previous research that LAI or plant biomass and chlorophyll concentration was related to REP, Mutanga and Skidmore (2007) showed that foliar N concentration and REP were correlated. It was discovered that REP which was established by calculating the inflection point between 680 nm and 750 nm, was highly correlated with nitrogen concentration (r =0.89). From previous research, it was required to test whether red edge has the potential to better detect N response. However, there are also some drawbacks with agronomic applications. In general, complicated methodology and the high cost of spectrometers or sensing tools restrict their use among crop producers (Daberkow and McBride, 2003). Spectrometer indices are widely available, but their use is restricted due to their high sensitivity to sunlight intensity (Kollenkark et al., 1982). REP has the potential to evaluate plant N status, but technologies employing REP tend to be costprohibitive. GreenSeeker is a commercially available, optical activelight sensor (independent of incident light) and is user friendly. However, the cost of the GreenSeeker handheld sensor is approximately US$4000, and is still considerably high for most farmers to use. To manufacture an activelight sensor employing REP would be more expensive because it requires much narrower and sensitive light bands compared to those used in the GreenSeeker sensor. For example, a hyperspectrometer must precisely detect within ± 5 nm. Research scientists at the Department of Biosystems and Agricultural Engineering at OSU have estimated the cost of such a sensor at approximately $8600, which simply employs the ratio of reflectances. From an agronomic perspective, it is crucial that timing of N application as well as the rate of N fertilizer be considered. Good application comes from a good decision. The best time to make a decision for midseason N application in winter wheat is at Feekes 4. It was reported that when midseason N application was made between Feekes 3 to 4, there was no yield loss (Boman et al., 1995). After Feekes 4, tissue damage and lower forage yields were detected from having applied foliar N. Rapid N uptake occurs between Feekes 2 to 4 and by Feekes 7, wheat takes up more than a third of the total accumulated (Waldren and Flowerday, 1979). Therefore, in winter wheat, it is essential to determine the N rate for midseason application at or before 6 Feekes 4. At this time, the wheat plant does not completely cover the ground, so NDVI is still sensitive. It is, thus, essential to investigate how NDVI and REP behave differently for early season growth of winter wheat. As an N rate recommendation tool, the SPAD meter is also commercially available. This device emits light at 650 and 940 nm. The transmittance ratio is then used for estimation of chlorophyll content. It has also been used for determining optimum N fertilization rate in wheat (Fox R.H. et al., 1994). Because it is one of the decision making tools for mid season N rate, it is essential to evaluate how it behaves differently with REP. In summary, REP should be an indicator of N status in plants. However, little research has been conducted to compare REP with the indices computed from commercially available tools, which determine midseason N rate application. Considering all the factors mentioned above it was considered prudent to investigate rededge and investigate its potential for detecting plant N status and ultimately identify optimum midseason N fertilization compared to NDVI in winter wheat. 7 OBJECTIVES AND HYPOTHES The objective of this paper was to determine whether the rededge index has the potential to be a useful index for detecting difference in N status for winter wheat compared to NDVI or commercially available instruments such as the SPAD meter. • Determine whether the REP has different behavior compared with NDVI measured from GreenSeeker at different growth stages and different N rates. • Determine whether the REP has different behavior compared with SPAD values measured from Minolta SPAD 502 chlorophyll meter at different growth stages and different N rates. • Determine whether REP could be a better index for detecting differences in plant N status. 8 MATERIALS AND METHODS A spectrometer, a chlorophyll meter (SPAD502), and a GreenSeeker handheld sensor were used to collect data in winter wheat. Measurements were taken at different growth stages (Feekes growth stages 4, 5, 7, and 10) for two cropping seasons. Location Data were collected from longterm winter wheat experimental plots located at Stillwater (Experiment # 222; Fig.3) and Perkins (N & P Study; Fig.4), Oklahoma. Experiment # 222 was established in 1969 under conventional tillage on a Kirkland silt loam (fine, mixed, superactive, thermic Udertic Paleustoll). The N & P study was initiated in 1996, also under conventional tillage on a Teller sandy loam (fineloamy, mixed, thermic Udic Argiustoll). These experiments are longterm NPK trials consisting of thirteen treatments (Experiment # 222) with four replications, and twelve treatments (N & P study) with three replications, respectively. Both were arranged in a randomized complete block design (RCBD). Five (treatments 14, and 10) and four (treatments 3,6,9 and 12) treatments were used in Experiment # 222 and in the N & P study, respectively. 9 Data Measurements Three instruments were used to obtain data for this study: the Minolta SPAD 502 meter, an Ocean Optics USB4000 spectrometer, and the GreenSeeker NTech handheld optical sensor. All of the readings were taken from a 1 m2 area in each treatment. SPAD Meter The Minolta SPAD 502 chlorophyll meter determines the relative amount of chlorophyll by measuring light transmitted or absorbed by plant leaves. The SPAD 502 is a compact meter that measures chlorophyll using optical density differences at two wavelengths (650 nm and 940 nm) with a measurement area of 2 mm x 3 mm. Twenty SPAD readings were randomly taken from winter wheat plant leaves within the 1 m2 sampling area, and subsequently averaged. Spectral Measurements The Ocean Optics USB4000 spectrometer operates with Spectrasuite (crossplatform Spectroscopy software) to measure reflectance. This spectrometer can detect reflectance from 2001100 nm at a high resolution (optical resolution of 1.5 nm full width half maximum). Reflectance of the plant canopy was computed by (the reflected light from the surface of the plant canopy minus black measurement to eliminate noise)/(incident light minus black measurement). Incident light was taken measuring reflectance of a 1m2 white board composed of Barium sulfate. Black measurement was taken measuring reflectance by covering the sensor with a cap and fabric material. Greenseeker® Sensor The GreenSeeker handheld optical sensor is an active sensor that measures reflectance in both red (671±10 nm) and near infrared (NIR; 780±10 nm) wavebands at a distance of 0.6 to 1.0 m from the canopy. It then calculates NDVI using the equation: NIR d NDVI NIR d Re Re  ρ ρ ρ ρ + = Where: ρ NIR = fraction of emitted NIR radiation from the sensed area ρRED = fraction of emitted red radiation from the sensed area Two readings per plot were these was calculated. One reading con Spectral Calculation Spectrometer readings indicies. For the REP, two methods were techniques and the linear method. because with this method, there is potential to compute REP the derivative method, spectrometer reflectance from 650 nm to 750 nm were and transported into Table Curve 2.D formula . By using the formula, the maximum point of the REP. For the linear method Figure 1. Red 10 taken with the GreenSeeker sensor, and an average of consists of 10 observations per second. were used to compute NDVI, REP, and simple ratio , applied: Derivative method by curve fitting In this study, the linear method was more focused using an active sensor. D. software and interpolated a using curve fitting first derivative was recorded as . (Figure 1), the interpretation by Clevers (1994) w Rededge position determine by the linear method , sists : For collected , was used. 11 2 ρ670 ρ780 ρ + re = (  ) (  ) 700 40* 740 700 700 ρ ρ ρ ρ λ re = + re Where: re ρ = Reflectance at estimated rededge ρ 670 = Reflectance at 670 nm ρ 700 = Reflectance at 700 nm ρ 740 = Reflectance at 740 nm ρ 780 = Reflectance at 780 nm ρ 780 = Reflectance at 780 nm re λ = Waverength of rededge position For the NIR, RED and GREEN spectral reflection, the following wave bands were selected and average reflectances were computed as the point of reflectance. NIR1 ρ = Average reflectance between 750775 nm NIR2 ρ =Average reflectance between 780805 nm ρ RED = Average reflectance between 740765 nm GREEN ρ = Average reflectance between 540565 nm Indices formulas are listed as following. 12 1. NIR RED NDVI NIR RED ρ ρ ρ ρ + = 1 1  1 2. NIR RED NDVI NIR RED ρ ρ ρ ρ + = 2 2  2 3. NIR GREEN GRNNDVI NIR GREEN ρ ρ ρ ρ + = 2 2  2 4. 2 2 IGREEN GRNRATIO NIR ρ ρ = 5. GREEN REDRATIO NIR ρ ρ 2 = Also Rededgearea was computed. Rededgearea was determined as the area between continuum line from 550 nm to 750 nm and reflectance. The reflectance at 550 nm and 750 nm was used to determine slope and intercept of the continuum line. Rededge area was determined by subtracting reflectance area from the area under the continuum line. Continuum line (λ ) y =a*λ+b ‘a’ and ‘b’ is computed from following formulas ρ 550 =a*550+b ρ 750 =a*750+b Area1 = *( ) 2 ( ) *( ) ( 1) ( ) ( 1) ( ) 750 ( ) ( 1) ( ) 550 750 550 i i i i i i y λ λ ρ ρ λ λ λ − + − − + + Σ + Σ Area2 = *( ) 2 ( ) *( ) ( 1) ( ) ( 1) ( ) 750 ( ) ( 1) ( ) 650 750 650 i i i i i i y λ λ ρ ρ λ λ λ − + − − + + Σ + Σ 13 0.2 0.3 0.4 0.5 0.6 0.7 0.8 500 550 600 650 700 750 800 Reflectance Wavelength (nm) Continuum Line Continuum Line Area1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 500 550 600 650 700 750 800 R e f l e c t a n c e Wavelength (nm) Continuum Line Area2 Figure 2. Visual description of Area1 and 2 14 RESULTS The following details results by location and year. In 2007, spectrometer measurements at Feekes 10 were excluded due to measurement errors. Stillwater 222, 20072008 GreenSeeker There were significat influences of N rate on GreenSeeker NDVI (GSNDVI) at Feekes 4, 5, 7 and 10 (Table 1). The NDVI values for this location ranged from 0.54 to 0.78, 0.50 to 0.77, 0.50 to 0.79, and 0.36 to 0.70 at Feekes 4, 5, 7, and 10 respectively. At Feekes 4, the highest GSNDVI was recorded. In general, as plant growth progressed GSNDVI decreased, especially at Feekes 10. The CV (coefficient of variation) for NDVI stayed constant and ranged from 7.4 to 7.7 between Feekes 4 to Feekes 7 but increased to 10.9 at Feekes 10. Plant N response was obvious at mature growth stages and the GreenSeeker sensor could detect these differences. Nitrogen rate and GSNDVI had significant linear relationships at all growth stages. Quadratic relationships were found only at Feekes 4 (Fig. 5). SPAD Meter There were significant influences of N rate on SPAD meter readings at Feekes growth stages 4, 5, 7 and 10. The SPAD readings for this location ranged from 32 to 48, 39 to 47, 30 to 37 and 33 to 38 at Feekes 4, 5, 7, and 10 respectively (Table 2). The CV for the SPAD readings dramatically decreased across different N rates with advancing plant growth, especially between Feekes 4 and Feekes 5 (12.7, and 4.1). It indicates that as plant growth progressed, small variability is detected by SPAD meter in different N rates and the sensitivity of the SPAD meter decreased with plant growth. Nitrogen rate 15 and SPAD readings had a significant linear relationship at all growth stages. Figure 6 shows a quadratic relationship after Feekes 7 and 10 at low N rates (0 and 45 kg N/ha). Spectral Indices There were significant influences of N rate on rededge position (REP) at all growth stages (Table 3). The range of REP increased as plant growth progressed (4.33, 3.88 and 5.39 nm at Feekes 4, 5, and 7 respectively). This trend was the same for GSNDVI. REP shifted to longer wavelengths with an increase in N rate (Fig. 7). At high N rates, shifts of REP to longer wavelength were clearly detected as growth stage increased. This means that the linear relationship between REP and N rates increased with advancing plant growth (r2=0.87, 0.98 and 0.99 at Feekes 4, 5, and 7). It indicates that REP has sensitivity to high N rates at mature growth stages, which was expected with high chlorophyll concentration and biomass. With GSNDVI and SPAD readings, N rate and REP had a significant linear relationship at all growth stages. A quadratic relationship was detected after Feekes 5. There were significant influences of N rate on rededgearea 2 (AREA2) as well as linear relationships at all growth stages (Table 4). AREA2 behaved similarly as REP increased with advancing plant growth as well as N rates (Fig. 8). Relationships Between Indices Overall, each index was highly correlated with all others at all growth stages (Tables 5a, b, c, and d). There was slightly higher correlation between REP and SPAD (r2=0.74, 0.83, and 0.81 at Feekes 4, 5, and 7 respectively) compared with correlation between GSNDVI and SPAD (r2=0.69, 0.76, and 0.66 at Feekes 4, 5, and 7 respectively) at all growth stages. It possibly indicates that REP has higher sensitivity to chlorophyll concentration than NDVI. Correlation between GSNDVI and REP was also relatively high and the correlation increased with advancing plant growth (r2=0.74, 0.77, and 0.81 at Feekes 4, 5, and 7 respectively). Similar relationships were found between the spectrometer based NDVI. There was no difference between REP and NIR/G or NIR/R. However, REP resulted in higher correlation with NIR/R and NIR/G (simple ratio) than with NDVI or GRN NDVI (normalized index). AREA2 had a good linear relationship with both REP (average r2= 0.81) and NDVI (average r2= 0.84). 16 Stillwater 222, 20082009 GreenSeeker There was a significant influence of N rate on GSNDVI at Feekes 4, 5, 7 and 10 (Table 6). The NDVI values for this location ranged from 0.35 to 0.61, 0.50 to 0.77, 0.27 to 0.60, and 0.27 to 0.60 at Feekes 4, 5, 7, and 10 respectively. Compared with 2007 2008, GSNDVI behaved differently, since the highest GSNDVI was recorded at Feekes 5 in each N rate and dramatically decreased at Feekes 7. As recorded in 20072008, the highest CV for GSNDVI was recorded at Feekes 10. Nitrogen rate and GSNDVI had a significant linear relationship at all growth stages and a quadratic relationship was detected only at Feekes 4, which was similar in 20072008 (Figure 9). SPAD Meter There were significant influences of N rate on SPAD meter readings at Feekes 4, 5, 7 and 10 (Table 7). SPAD values for this location ranged from 39 to 48, 38 to 43, 35 to 44 and 33 to 42 at Feekes 4, 5, 7, and 10 respectively. As recorded in 20072008, the CV for the SPAD readings dramatically decreased as plant growth advanced, especially between Feekes 4 and Feekes 5 (CV= 8.5, and 5.1). A linear relationship was found at all growth stages and a quadratic relationship was found after Feekes10, which was similar in 20072008 year (Fig. 10). Spectral Indices There were significant influences of N rate on rededge position (REP) at all growth stages (Table 8). The REP shifted to longer wavelengths with the increase in N rates. As recorded in 20072008, linear correlation between REP and N rates increased with advancing plant growth (r2= 0.73, 0.92, 0.97, and 0.99 at Feekes 4, 5, 7 and 10 respectively). A quadratic relationship was detected at Feekes 4 and 5 (Fig. 11). AREA2 increased with advancing plant growth as well as N rate (Table 9). The ranges were 46, 17 57, 68 and 78 at Feekes 4, 5, 7, and 10, respectively. The plant N response would be bigger as plant growth advances. At Feekes 7 and 10, AREA2 had the same response to different N rates as GSNDVI and REP (Fig. 12). Relationships Between Indices Table 10a, b, c and d shows that at Feekes 5 and 10, there was higher correlation between REP and SPAD (r2=0.53 and 0.66) but at Feekes 4 and 7, there was higher correlation between GSNDVI and SPAD (r=0.45 and 0.76 at Feekes 4 and 7) than between REP and SPAD (r2=0.32 and 0.62 at Feekes 4 and 7). This indicates that REP was more sensitive to chlorophyll content than NDVI at Feekes 5 and 10 in this year, but it was all growth stages in the previous year. Correlation coefficients between GSNDVI and REP were relatively high and the correlation stayed similar with advancing plant growth (r2=0.79, 0.85, 0.77, and 0.85 at Feekes 4, 5, 7, and 10 respectively). The same relationships were found between spectrometer based NDVI. Compared with 20072008, there was differing correlation between REP and NIR/G or NIR/R. Slightly higher correlation was found between REP and NIR/R than between REP and NIR/G. Compared with 20072008, only at Feekes 10, REP resulted higher correlation with NIR/R and NIR/G (simple ratio) than with NDVI or GRN NDVI (normalized index). Through all growth stages, AREA2 had the highest correlation with NDVI. Summary for Experiment 222 in Stillwater GSNDVI is sensitive to different plant N responses as well as different plant growth stages but the sensitivity to differing plant N response tends to be higher in REP, especially as the plant grows. SPAD is sensitive to different N rates at the early growth stages but not after Feekes 5. The accurate metric detection of REP might be challenging in the field because the change of wavelength is between 1 and 10 nm. Especially at early growth stages, the range of REP was only about 4 nm between 0 kg N/ha and 135 kg N/ha treated plant. GSNDVI and REP had a good linear relationship at all growth stages (average r2=0.79). SPAD and REP had a linear relationship (average r2=0.65) but it was not superior compared to GSNDVI and REP. GSNDVI and SPAD 18 linear relationship was average r2=0.62 and it indicated that REP was more sensitive to chlorophyll concentration than GSNDVI. Perkins, 20072008 At Feekes 7 in 2008, spectrometer data were excluded due to measurement error. At Feekes 5 in 2009, all data were excluded due to measurement error. GreenSeeker There were significant influences of N rate on GSNDVI at Feekes 4, 5, 7 and 10 (Table 11). The NDVI values for this location ranged from 0.43 to 0.56, 0.44 to 0.77, 0.53 to 0.86, and 0.46 to 0.71 at Feekes 4, 5, 7, and 10 respectively. Compared with Stillwater, the CV for GSNDVI decreased with advancing plant growth and the lowest CV across different N rates was recorded at Feekes 10. Nitrogen rate and GSNDVI had a significant linear relationship and a non quadratic relationship was recorded at all growth stages. SPAD Meter There were significant influences of N rate on SPAD meter readings at Feekes 4, 7 and 10 (Table 12). SPAD readings for this location ranged from 46 to 54, 41 to 49, 31 to 41 and 34 to 48 at Feekes 4, 5, 7, and 10 respectively. The CV for the SPAD readings stayed constant at all growth stages and its decrease was not detected like that at Stillwater. Spectral Indices At all growth stages, the range of REP stayed small, 3 to 6 nm, across different N rates (Table 13). It resulted in no significant difference of N rate on REP especially at the early growth stages which are important. The CV for REP stayed constant at all growth stages. On the other hand, N rates had significant differences on AREA2 (Table 14). At all growth stages, linear relationships were detected but no quadratic relationship was found between AREA 2 and N rates. This means that AREA2 had similar behavior with 19 GSNDVI. Feekes 10 had the greatest linear relationship (r2=0.96) between AREA2 and N rate. This indicates that AREA2 is sensitive to detect plant N response at different N rates. Relationships Between Indices As recorded at Stillwater, the relationship between SPAD and REP (average r2 =0.29) tends to be higher than between SPAD and NDVI (average r2=0.27). (Table 15a, b, c, and d) At all growth stages, GSNDVI and REP had a high linear relationship (average r2=0.76). AREA2 tended to have higher correlation with REP (average r2=0.86) than with GSNDVI (average r2=0.74). Perkins, 20082009 GreenSeeker There were significant influences of N rate on GSNDVI at Feekes 4, 5, 7 and 10 (Table 16). The GSNDVI values for this location ranged from 0.24 to 0.55, 0.30 to 0.78, and 0.33 to 0.63 at Feekes 4, 7, and 10 respectively. As recorded in 20072008, the CV for NDVI decreased with advancing plant growth and the lowest CV was recorded at Feekes 10. N rate and GSNDVI had a significant linear relationship at all growth stages and a quadratic relationship was recorded at Feekes 7 and 10. SPAD Meter There were significant influences of N rate on SPAD Meter readings (SPAD) at Feekes 4, 7 and 10 (Table 17). SPAD readings for this location ranged from 40 to 59, 26 to 44, and 27 to 41 at Feekes 4, 7, and 10 respectively. The CV for SPAD readings increased with advancing plant growth which was not observed at Stillwater. As recorded in 20072008, linear relationships and non quadratic relationships were found at all growth stages. 20 Spectral Indices There were significant influences of N rate on REP at all growth stages (Table 18). The REP shifted to longer wavelengths with increases in N rate. At Feekes 4, REP range was 3 but it increased to 8 at Feekes 7 and 10. REP had significant linear relationships with N rates at all growth stages. A quadratic relationship was detected only at Feekes 7. The AREA2 tended to increase as plant growth advanced as well as with N rate increase (Table 19). As recorded in 20072008, Feekes 10 had the best linear relationship between N rate and AREA 2 (r2=0.90, 0.89 and 0.99 at Feekes 4, 7, and 10 respectively). This means that AREA 2 was sensitive to plant growth, which ultimately could mean good sensitivity for plant biomass. Relationships Between Indices Table 20a, b, and c showed that at all growth stages, GSNDVI and REP were highly correlated (r2=0.66, 0.61, and 0.81 at Feekes 4, 7 and 10). The relationship between REP and SPAD had higher correlation (r2=0.74, 0.81 and 0.74 at Feekes 4, 7 and 10) than the relationship between GSNDVI and SPAD (r2= 0.90, 0.55 and 0.53 at Feekes 4, 7 and 10). GSNDVI tended to have a higher linear relationship with AREA2 (average r2=0.82) than REP (average r2=0.7). There were no significant differences between NIR/G and NIR/R with REP. Summary of N&P Study in Perkins GDNDVI was sensitive to different plant N response as well as different growth stages. SPAD was sensitive to different plant N responses at the early growth stages but tended to decrease as plant grows. REP was sensitive to different N rate responses as well as growth stages but the REP range from high N rates to low N rates was really small especially at the early growth stages (Range = 4 nm and 8 nm at Feekes 4 and 10). This same result was found in Stillwater. GSNDVI and REP had a good linear relationship at all growth stages (average r2=0.72). SPAD and REP had a linear relationship (average r2=0.53) which was stronger than SPAD and GSNDVI (average r2=0.47). Because the SPAD value is computed from a 21 chlorophyll meter it is highly correlated with chlorophyll concentration, and as such defines REP as more sensitive to chlorophyll concentration than NDVI. 22 DISCUSSION AND CONCLUSIONS In this study, there are three things to be discussed in terms of REP as a potential index to detect in plant N status. The first thing that needs to be discussed is the advantage and disadvantage of REP in field based research. Many studies showed that REP was highly correlated with chlorophyll content at leaf and plant canopy level (Chappelle et all.1991; Cho and Skidmore 2006). It was also shown in our research that REP was highly correlated with Chlorophyll SPAD meter readings. The position and the shape of first derivative spectrum gives more opportunities to differentiate plant N response (Cheng et al., 2005; Kupfer and et all., 1990). As described in Figure 13, the shape of the first derivative reflectance is sharper in low N treated plants compared with high N treated plants, which have two maximum points with round shape. At the same plant growth stage, not only can we differentiate plant N status by “the shape” of the first derivative reflectance but also we can distinguish by “the position” of the maximum point of the first derivative reflectance. The basic point is that you can manipulate two outcomes, “the position and the shape” of the first derivative reflectance, from rededge bands which does not happen with NDVI or simple ratio. Filella and Penuelas (1994) showed that the area of the first derivative reflectance has strong correlation with plant biomass. Another advantage is that under high biomass, NDVI became saturated but rededge position could give more accurate estimates of biomass (Filella and Penuelas,1994; Mutanga and Skidmore, 2004). The results show that GSNDVI is sensitive to different plant N responses as well as different plant growth stages but the sensitivity to different plant N response tends to be higher in REP, especially as the plant grows. However there are also disadvantages. Red edge seems very sensitive to noise. Rededge could be obtained by using a hyperspectometer but because of high contents of information per pixel, the analysis of derivative system requires the right techniques and time (Ruffin and King,1999). To eliminate noise, a smoothing method is applied but it might lose or intensively modify original data (Shafri and Yusof, 2009). Figure 13 23 shows the first derivative method of original data. It does not clearly show the maximum points both at low and high N rates. Figure 14 shows the first derivative which is filtered by 10 nm. It shows a smoother line than Figure 15 and could detect two peaks in the high N rate; however it still showed noise at the low N rate. Figure 16 shows one more weighed filtering at 10 nm. Of course the line gets smoother but loses one of two peaks at the high N rate and modifies the shape of the first derivative reflectance compared with Figure 15. In Figure 17, the first derivative was computed through a trend fitting method using TableCurve 2D v5.01. A peak of the first derivative was clear but it loses a lot of information especially the shape of the first derivative reflectance at the high N rate compared with original data. Engineers developed active sensors like the GreenSeeker or transmittance sensors like the SPAD meter for field application. The purpose of the active sensor development is to eliminate noise which is caused by solar radiation. Solar intensity changed dramatically through the day and it limited the application of sensing technology in fields. For rededge detection, a passive sensor would be required. It means that the energy source depends on nonstable solar radiation and may limit the practical application in fields. Especially for the N rate recommendation tool, target users are farmers who do not concern themselves with taking data at noon or at night. They take data whenever they can and it could be the negative points of passive sensors because they create noise. It might have the potential to determine rededge position by using optical sensor using a linear method. However, it requires more spectral bands compared to NDVI and it will increase the cost for the field application. With that, in this wheat study, the range of rededge position was narrow, not more than 15 nm, between nonN treated plants and high N treated plants. So even if rededge is more sensitive to chlorophyll content or biomass, if it requires passive measurements, it is speculative whether the sensitivity is worth it to look at agronomic fields. It is also unclear if they can overcome noise and clearly detect narrow wavebands, whether the technology allows computing rededge for on the go sensing and optimum N recommendations. From the agronomic application aspect, hyperspectrometers cost more than band spectrometers which are used in the field today. It needs to be asked if today’s engineer could develop a sensor capable of detecting rededge position clearly with the farmers’ budget in mind. To explain these questions, futher engineering investigation is required. 24 Second, the relationship between NDVI and REP needs to be discussed. Researchers have been discussing that the REP is the alternative index due to the high sensitivity under dense green biomass. In Oklahoma, farmers apply mid season N at Feekes 4 and 5. So they need to make a decision for the N rate before Feekes 5. At that time, the ground is not fully covered by biomass (Figure 18 and 19). So NDVI saturation should not be a problem. But with the decrease in biomass on the ground, other problems arise. With a decrease in plant cover on the ground, the soil is more exposed. It might increase the noise for NDVI. In Figure 20, the relationship between REP and GSNDVI is described at Feekes 4. Under the high N rate which was expected to have more biomass on the ground (Figure 18 and 19), the correlation between REP and GSNDVI was high but it decreased as N rate decreased. The decrease in correlation could be explained by the reduction of plant biomass which ultimately increases soil exposed area and it influences differently on the value of REP and NDVI. Some research noted that there is less influence of soil background on REP. Therefore, NDVI might be more influenced by soil background and it ends up behaving differently than REP. This statement also could explain Figure 21. Conventional tillage had a better linear correlation between REP and GSNDVI compared with no till. Due to remaining residues in no till, the soil background was less homogenous compared with conventional tillage and it could reduce the relationship between REP and GSNDVI. To verify the question, further research is required by taking biomass under the different soil backgrounds. But from an agronomic view, farmers put out preplant N and good biomass should cover the soil. It means that it is not like the behavior between REP and NDVI in the check or 0 kg N/ha. Therefore correlation above r2=0.7 is estimated at Feekes 4 in practical field. In this research, there was no biomass data, but at Feekes 4 and 5, REP and NDVI were strongly correlated (average r= 0.88) (Figure 22). The third aspect to be discussed is whether REP may provide for the determination of optimum N rates in winter wheat. In the OSU N algorithm, N rate is determined based on the following factors: plant N response index (RI) and inseason estimated yield (INSEY), which is computed by NDVI divided by the number of days with growing degree days (GDD>0) from planting to sensing, estimated yield potential without additional N (YP0) and with additional N (YPN) computed from the INSEY formula, grain N uptake, and nitrogen use efficiency. OSU recognized the increasing nitrogen use 25 efficiency more than 15 % using this algorithm. NDVI has been used because of its strong relationship with biomass. Research has shown REP is strongly correlated with chlorophyll content. However, N concentration which is ultimately correlated with chlorophyll concentration, was weakly correlated with yield (Holtz et all., 2008). Therefore, it is essential to investigate the ability of REP to predict biomass. Also, to utilize the REP as an index, REP has to increase the sensitivity of estimation of plant biomass significantly, especially at Feekes 4 to 5, because it is an important time from agronomic perspective to determine N rates. Under the current OSU N algorithm, the 0.01 NDVI difference could change the N rate between 1 kg/ha and 5.6 kg/ha in winter wheat. If REP could detect plant biomass more accurately, it is important how sensitive it is. Even if REP could detect 0.01 NDVI differences and change the N rates to, for example, 1 kg N/ha to 1.5 kg N/ha, farmers will put down additional 2 kg N/ha anyway. They do not require that much precision. In conclusion, REP behaves very similar to NDVI. Because of economical inputs needed to implement REP into sensor based technology which enables us to apply this information in agronomic fields, at this time, more evidence is needed that suggest REP is significantly better than NDVI. The data suggest that this new sensor may not be worth the investment in winter wheat. Further research is needed to evaluate REP with biomass at the early plant growth stage. 26 REFERENCES 1. Boelman, N.T., M. Stieglitz, H.M. Rueth, M. Sommerkorn, K.L. Griffin, G.R. Shaver, and J.A. Gamon. 2003. Response of NDVI, biomass, and ecosystem gas exchange to longterm warming and fertilization in wet sedge tundra. Oecologia. 135:414–421. 2. Boman., R.K., R.L. Westerman, W.R. Raun. And M.E. Jojola. 1995. Time of nitrogen application: Effects on winter wheat and residual soil nitrate. Soil.Sci.Soc. Am.J. 59:1364`369. 3. Boochs., F. and G. Kupfer. 1990. Shape of the red edge as vitality indicator for plants. Int. J. Remote Sens. 11:17411753. 4. Chappelle, E.W., M.S. Kim, and J.E. McMurtrey III. 1991. The red edge shift: An explanation of its relationship to stress and the concentration of chlorophyll a. Proceedings of the 11th Annual International Geoscience and Remote Sens. Symposium, Espoo, Finland; UNITED STATES; 36 June 1991. 22872290. 5. Chen, Z., and Elvidge, C. D., 1993, Description of derivativebased high spectralresolution (AVIRIS) green vegetation index. Proc. SPIE. 1937: 43 54. 6. Cheng., Y., C. Hu, H. Dai and Y. Lei. 2005. Spetral red edge parameters for winter wheat under different nitrogen support levels. Proceeding of SPIEThe international society for optical engineering. 5884:16. 7. Cho, M.A. and A.K. Skidmore. 2006. A new technique for extracting the res edge position from hyperspectral data: The linear extrapolation method. Remote Sensing of Environment. 101:181193. 8. Clevers, J. G. P. W. 1994. Plant vitality and yield indicators. Imaging spectrometryA tool for environmental observation. 4:193219. 9. Clevers, J. G. P. W., S. M. de Jong, , G. F. Epema, , F.V. der Meer, , W. H. Bakker, A. K. Skidmore and E.A. Addink, 2001. MERIS and the rededge 27 10. position. International journal of applied earth observation and geoinformation (JAG). 3: 313320. 11. Clevers, J. G. P. W. and R. Jongschaap. 2006. Imaging spectrometry for agriculture applications. Imaging spectrometry: Basic principal and prospective application. 157197. Dordrecht, Netherlands. Springer. 12. Collins, W. 1978. Remote sensing of crop trpe and maturity. Photogr. Engineer. Remote Sensing. 44:4355. 13. Elmore, A.J., J.F. Mustard, S.J. Manning, and D.B. Lobell. 2000. Quantifying vegetation change in semiarid environments precision and accuracy of spectral mixture analysis and the normalized difference vegetation index. Remote Sens. Environ. 73:87102. 14. Daberkow, S.D. and W.D McBride. 2003. Farm and operator characteristics affecting the awareness and adoption of precision agriculture technologies in the U.S. Precision Agriculture. In: Precision Agriculture 2001: Proceedings of the 3rd European Conference on precision Agriculture. 4:163177. 15. Dawson., T.P., and P.J. Curran. 1998. A new technique for interpolating the reflectance red edge position. Int. J. Remote Sens. 19:21332139. 16. Filella, I. and J. Penuelas. 1994. The red edge positon and shape as indicators of plant chlorophyll content, biomass and hydric status. 1994. INT. J. Remote Sens. 15: 14591470. 17. Fox R.H., W.P. Piekielek, and K.M. Macneal. 1994. Using a Chlorophyll Meter to Predict Nitrogen Fertilizer Needs of Winter Wheat. Soil. Sci. Plant. Anal. 25:171 181. 18. Gitelson. A.A., Y.J. Kaufman, R. Stark and D.C. Rundquist. Novel algorithms for remote estimation of vegetation fraction. 2002. Remote Sens. Environ. 80:76–87. 19. Guyot, G., & Baret, F. (1988). Utilisation de la haute résolution spectrale pour suivre l'état des couverts végétaux. Proceedings of the 4th International colloquium on spectral signatures of objects in remote sensing. ESA SP287, Assois, France. 279−286. 20. Hatfield, J. L., G. Asrar, and E. T. Kanemasu. 1984. Intercepted photosynthetically active radiation estimated by spectral reflectance. Remote Sens. Environ. 14: 6575. 21. Holtz, S., K. Grima, B. Tubana, D.B. Arnall, D. Edmond, Y. Kanke, J.B. Solie, and W.R. Raun. 2008. Above ground nitrogen accumulation as a function of time in corn and winter wheat. J. Plant Nutr. ??? 28 22. Kollenkark, J. C., Vanderbilt, V. C., Daughtry, C. S. T., and Bauer, M. E. 1982. Influence of solar illumination angle on soybean canopy reflectance. Appl. Opt. 21:11791184. 23. Kupfer, F.B.G., K. Dockter and W. Kuhbauch. 1990. Shape of the red edge as vitality indicator for plants. INT. J. Remote Sens. 11:17411753. 24. Middleton, E.M., P.K. E. Campbell, J.E. McMurtrey, L.A. Corp, L.M. Butcher, and E.W. Chappelle. 2002. “Red edge” optical properties of corn leaves from different nitrogen regimes. In: IEEE International Geoscience and Remote Sensing Symposium, 2002. IGRASS 2002. 4:22082210. 25. Mutanga, O. and Skidmore, A.K. 2004. Hyperspectral band depth analysis for a better estimation of grass biomass (Cenchrus ciliaris) measured under controlled laboratory conditions. International Journal of Applied earth Observation and Geoinformation. 5:8796. 26. Mutanga, O. and Skidmore, A.K. 2004. Narrow band vegetation indices overcome the saturation problem in biomass estimation. Int. J. Remote Sens. 25:39994014. 27. Mutanga, O. and Skidmore, A.K. 2007. Red edge shift and biochemical content in grass canopies. ISPRS Journal of Photogrammetry and Remote Sensing. 62:3442. 28. Myneni, R.B., F.G. Hall, P.J. Sellers, and A.L. Marshak. 1995. The interpretation of spectral vegetation indexes. Geosci. Remote Sens. 33: 481486. 29. Raun, W.R., and G.V. Johnson. 1999. Improving nitrogen use efficiency for cereal production. Agron. J. 91:357–363. 30. Raun, W.R., J.B. Solie, G.V. Johnson, M.L. Stone, R.W. Mullen, K.W. Freeman, W.E. Thomason, and E.V. Lukina. 2002. Improving nitrogen use efficiency in cereal grain production with optical sensing and variable rate application. Agron. J. 94:815–820. 31. Raun, W.R., J.B. Solie, M.L. Stone, K.L. Martin, K.W. Freeman, R.W. Mullen, H. Zhang, J.S. Schepers, and G.V. Johnson. 2005. Optical sensorbased algorithm for crop nitrogen fertilization. Commun. Soil Sci. Plant Anal. 36:2759–2781. 32. Rondeaux, G., M. Steven, and F. Baret. 1996. Optimization of soiladjusted vegetation indices. Remote Sens. Environ. 55: 95–107. 33. Ruffin,C., and King, R. L. 1999. The analysis of hyperspectral data using SavitzkyGolayfiltering Theoretical basis In: IEEE International Geoscience and 29 Remote Sensing Symposium, 1999. CD ROMIEEE Catalog Number 99CH36293. 34. Shafri, H.Z.M., M.A.M Shalleh and A. Ghiyamat. 2006. Hyperspectral remote sensing of vegetation using red edge position. Am. J. Appl. Sci3. 6:18641871 35. Shafri, H.Z.M., and M.R.M. Yusof. 2009. Trends and issues in noise reduction for hyperspectral vegetation reflectance spectra. European J. Sci. Res. 23:404410. 36. Solie, J.B., W.R. Raun, R.W. Whitney, M.L. Stone and J.D. Ringer. 1996. Optical sensor based field element size and sensing strategy for nitrogen application. Trans. ASAE 39(6):19831992. 37. Waldren, R.P and A.D. Flowerday. 1979. Growth stages and distribution of dry matter, N, P, and K in winter wheat. Agron. J. 71:391397. 30 APPPENDIX 31 Table 1. Analysis of variance for GreenSeeker NDVI at different N rates and different growth stages, Stillwater, OK, 20072008. Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 3 ns ns ns ns Treatment 3 * * * * Error 28 0.06219 0.06211 0.05793 0.0842 Linear * * * * Quodratic * ns ns ns TRT, (NP K kg/ha) Treatment Means 067 45 0.543 0.505 0.495 0.362 456745 0.631 0.595 0.589 0.457 906745 0.762 0.717 0.713 0.609 1356745 0.777 0.775 0.794 0.696 CV, % 7. 4 7. 7 7. 4 10 .9 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 32 Table 2. Analysis of variance for SPAD readings at different N rates and different growth stages, Stillwater, OK, 20072008. Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 3 ns ns ns Ns Treatment 3 * * * * Error 28 637.6 108.7 45.3 143.7 Linear * * * * Quodratic ns ns * * TRT, (NP K kg/ha) Treatment Means 067 45 31.78 38.55 29.25 34.61 456745 35.13 40.01 30.21 33.46 906745 44.58 44.66 33.86 34.36 1356745 47.81 46.88 37.31 37.67 CV, % 12 .7 4. 9 4. 1 6. 8 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 33 Table 3. Analysis of variance for rededge position (REP) at different N rates and different growth stages, Stillwater, OK, 20072008. Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 3 ns * ns  Treatment 3 * * *  Error 28 21.96 21.22 7.83  Linear * * *  Quodratic ns ns *  TRT, (NP K kg/ha) Treatment Means 067 45 714.28 716.16 715.72  456745 715.51 716.44 716.38  906745 718.10 718.9 718.47  1356745 718.61 720.46 721.11  CV, % 0.1 3 0.1 3 0.0 8  *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 34 Table 4. Analysis of variance for AREA2 at different N rates and different growth stages, Stillwater, OK, 20072008. Source of Variation Df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 3 ns ns *  Treatment 3 * * *  Error 28 7256 3309 5053  Linear * * *  Quodratic ns ns ns  TRT, (NP K kg/ha) Treatment Means 067 45 77.29 83.17 101.64  456745 109.33 97.74 118.46  906745 137.23 127.13 156.49  1356745 153.45 138.93 193.32  CV, % 14 .3 10 .3 10 .1  *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 35 Table 5a. Correlation matrices and simple correlation coefficients, Feekes growth stage 4, Stillwater, OK, 20072008. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 4 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRN NDVI 1 GSNDVI 2 SPAD 0.69* 3 NDVI1 0.69* 0.52* 4 NDVI2 0.69* 0.52* 0.98* 5 REP 0.74* 0.74* 0.76* 0.76* 6 Area1 0.83* 0.76* 0.55* 0.55* 0.71* 7 Area2 0.85* 0.74* 0.53* 0.53* 0.67* 0.98* 8 NIR/G 0.49* 0.46* 0.98* 0.98* 0.71* 0.44* 0.46* 9 NIR/R 0.66* 0.53* 0.98* 0.98* 0.77* 0.50* 0.53* 0.98* 10 GRN NDVI 0.61* 0.45* 0.98* 0.98* 0.69* 0.45* 0.46* 0.98* 0.98* 36 Table 5b. Correlation matrices and simple correlation coefficients, Feekes growth stage 5, Stillwater, OK, 20072008. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 5 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRN NDVI 1 GSNDVI 2 SPAD 0.76* 3 NDVI1 0.92* 0.74* 4 NDVI2 0.92* 0.74* 0.98* 5 REP 0.77* 0.83* 0.81* 0.81* 6 Area1 0.90* 0.74* 0.94* 0.94* 0.83* 7 Area2 0.92* 0.74* 0.96* 0.96* 0.86* 0.98* 8 NIR/G 0.90* 0.79* 0.98* 0.98* 0.88* 0.96* 0.96* 9 NIR/R 0.90* 0.79* 0.98* 0.98* 0.88* 0.94* 0.96* 0.98* 10 GRN NDVI 0.92* 0.76* 0.98* 0.98* 0.83* 0.94* 0.96* 0.98* 0.98* 37 Table 5c. Correlation matrices and simple correlation coefficients, Feekes growth stage 7, Stillwater, OK, 20072008. Feekes 7 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRN NDVI 1 GSNDVI 2 SPAD 0.66* 3 NDVI1 0.83* 0.64* 4 NDVI2 0.83* 0.66* 0.98* 5 REP 0.81* 0.81* 0.83* 0.85* 6 Area1 0.66* 0.62* 0.59* 0.59* 0.77* 7 Area2 0.77* 0.71* 0.76* 0.76* 0.90* 0.90* 8 NIR/G 0.83* 0.77* 0.92* 0.92* 0.94* 0.90* 9 NIR/R 0.81* 0.71* 0.96* 0.96* 0.88* 0.69* 0.83* 0.98* 10 GRN NDVI 0.85* 0.76* 0.96* 0.96* 0.92* 0.58* 0.72* 0.98* 0.96*  *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. 38 Table 5d. Correlation matrices and simple correlation coefficients, Feekes growth stage 10, Stillwater, OK, 20072008. Feekes 10 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRN NDVI 1 GSNDVI 2 SPAD .26* 3 NDVI1 4 NDVI2 5 REP 6 Area1 7 Area2 8 NIR/G 9 NIR/R 10 GRN NDVI *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. 39 Table 6. Analysis of variance for GreenSeeker NDVI at different N rates and different growth stages, Stillwater, OK, 20082009. Source of Variation Df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 3 * ns ns ns Treatment 3 * * * * Error 28 0.034 0.062 0.038 0.065 Linear * * * * Quodratic * ns ns ns TRT, (NP K kg/ha) Treatment Means 067 45 0.358 0.505 0.266 0.273 456745 0.519 0.595 0.402 0.419 906745 0.599 0.717 0.487 0.470 1356745 0.612 0.775 0.596 0.604 CV, % 7. 1 7. 7 8. 9 11 .6 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 40 Table 7. Analysis of variance for SPAD readings at different N rates and different growth stages, Stillwater, OK, 20082009 Source of Variation Df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Means Squares Replication 3 ns ns ns ns Treatment 3 * * * * Error 28 356.7 112.43 104.34 180.9 Linear * * * * Quodratic ns ns ns * TRT, (NP K kg/ha) Treatment Means 067 45 39.18 37.55 35.43 32.91 456745 44.10 39.80 38.73 33.17 906745 45.6 40.88 42.05 37.43 1356745 48.35 42.95 44.03 42.46 CV, % 8. 5 5. 3 5. 1 7. 4 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 41 Table 8. Analysis of variance for REP at different N rates and different growth stages, Stillwater, OK, 20089. Source of Variation Df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 3 ns ns ns ns Treatment 3 * * * * Error 28 12.009 4.323 24.76 19.54 Linear * * * * Quodratic * * ns ns TRT, (NP K kg/ha) Treatment Means 067 45 714.64 716.26 714.45 714.15 456745 716.27 717.32 716.22 715.97 906745 717.68 718.27 717.69 717.52 1356745 717.64 718.58 718.48 719.65 CV, % 0. 1 0.0 6 0.1 4 0.1 2 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 42 Table 9. Analysis of variance for AREA2 at different N rates and different growth stages, Stillwater, OK, 20082009. Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 3 * * ns ns Treatment 3 * * * * Error 28 3629 5145 4122 7089 Linear * * * * Quodratic ns * ns ns TRT, (NP K kg/ha) Treatment Means 067 45 46.46 78.48 98.64 85.81 456745 67.77 122.81 115.36 119.18 906745 84.99 154.76 149.7 147.14 1356745 92.37 135.91 166.54 164.09 CV, % 16 .6 11 .7 30 .6 13 .1 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 43 Table 10a. Correlation matrices and simple correlation coefficients, Feekes growth stage 4, Stillwater, OK, 20082009. Feekes 4 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRNNDVI 1 GSNDVI 2 SPAD 0.45* 3 NDVI1 0.94* 0.45* 4 NDVI2 0.94* 0.46* 0.98* 5 REP 0.79* 0.32* 0.83* 0.83* 6 Area1 0.34* 0.03* 0.36* 0.38* 0.46* 7 Area2 0.50* 0.10* 0.52* 0.55* 0.46* 0.96* 8 NIR/G 0.90* 0.50* 0.96* 0.96* 0.76* 0.26* 0.42* 9 NIR/R 0.92* 0.48* 0.98* 0.98* 0.81* 0.40* 0.56* 0.96* 10 GRN NDVI 0.92* 0.52* 0.98* 0.96* 0.77* 0.25* 0.42* 0.98* 0.96* *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. 44 Table 10b. Correlation matrices and simple correlation coefficients, Feekes growth stage 5, Stillwater, OK, 20082009. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 5 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRNNDVI 1 GSNDVI 2 SPAD 0.49* 3 NDVI1 0.81* 0.41* 4 NDVI2 0.81* 0.40* 0.98* 5 REP 0.85* 0.53* 0.83* 0.83* 6 Area1 0.56* 0.26* 0.69* 0.69* 0.61* 7 Area2 0.67* 0.29* 0.85* 0.85* 0.72* 0.96* 8 NIR/G 0.77* 0.35* 0.98* 0.98* 0.76* 0.66* 0.83* 9 NIR/R 0.81* 0.40* 0.98* 0.98* 0.83* 0.66* 0.81* 0.98* 10 GRN NDVI 0.77* 0.35* 0.98* 0.98* 0.76* 0.67* 0.83* 0.98* 0.98* 45 Table 10c. Correlation matrices and simple correlation coefficients, Feekes growth stage 7, Stillwater, OK, 20082009. Feekes 7 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRNNDVI 1 GSNDVI 2 SPAD 0.76* 3 NDVI1 0.50* 0.32* 4 NDVI2 0.50* 0.34* 0.98* 5 REP 0.77* 0.62* 0.52* 0.53* 6 Area1 0.41* 0.25* 0.71* 0.69* 0.59* 7 Area2 0.42* 0.27* 0.76* 0.74* 0.59* 0.96* 8 NIR/G 0.44* 0.31* 0.81* 0.81* 0.56* 0.81* 0.86* 9 NIR/R 0.45* 0.35* 0.76* 0.76* 0.55* 0.83* 0.86* 0.98* 10 GRN NDVI 0.45* 0.32* 0.85* 0.85* 0.58* 0.81* 0.85* 0.98* 0.92* *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. 46 Table 10d. Correlation matrices and simple correlation coefficients, Feekes growth stage 10, Stillwater, OK, 20082009. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 10 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRNNDVI 1 GSNDVI 2 SPAD 0.55* 3 NDVI1 0.71* 0.44* 4 NDVI2 0.71* 0.44* 0.98* 5 REP 0.85* 0.66* 0.85* 0.85* 6 Area1 0.74* 0.50* 0.94* 0.94* 0.88* 7 Area2 0.74* 0.50* 0.94* 0.94* 0.88* 0.98* 8 NIR/G 0.67* 0.52* 0.98* 0.98* 0.85* 0.92* 0.92* 9 NIR/R 0.71* 0.53* 0.96* 0.96* 0.86* 0.92* 0.92* 0.98* 10 GRN NDVI 0.67* 0.45* 0.98* 0.98* 0.83* 0.92* 0.92* 0.98* 0.98*  47 Table 11. Analysis of variance for GreenSeeker NDVI at different N rates and different growth stages, Perkins, OK, 20072008. Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 2 ns ns ns ns Treatment 3 * * * * Error 8 0.029 0.031 0.03 0.016 Linear * * * * Quodratic ns ns ns ns TRT, (NP K kg/ha) Treatment Means 06 70 0.429 0.436 0.525 0.455 56670 0.419 0.576 0.618 0.553 112670 0.537 0.697 0.805 0.698 168670 0.559 0.771 0.856 0.708 CV, % 14.3 11.5 10.1 8.6 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 48 Table 12. Analysis of variance for SPAD readings at different N rates and different growth stages, Perkins, OK, 20082009. Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 2 ns ns Ns Ns Treatment 3 * * * Ns Error 8 40.43 48.59 28.22 36.69 Linear * * * Ns Quodratic ns ns Ns Ns TRT, (NP K kg/ha) Treatment Means 06 70 45.76 40.66 30.93 33.67 56670 50.2 41.33 34.63 34.83 112670 53.9 44.93 40.9 34.57 168670 53.5 48.8 40.93 37.57 CV, % 5.1 6.47 5.855 7.85 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 49 Table 13. Analysis of variance for REP at different N rates and different growth stages, Perkins, OK, 20072008. Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 2 Ns ns  ns Treatment 3 Ns Ns  * Error 8 6.61 9.08  9.4 Linear Ns *  * Quodratic Ns ns  ns TRT, (NP K kg/ha) Treatment Means 06 70 717.44 718.16  714.32 56670 717.17 718.2  716.42 112670 719.44 719.95  719.33 168670 719.55 721.19  719.95 CV, % 0.1 5 0.1 7  0.1 7 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 50 Table 14. Analysis of variance AREA2 at different N rates and different growth stages, Perkins, OK, 20072008 Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 2 ns *  ns Treatment 3 * ns  * Error 8 2337 3862  1630 Linear * *  * Quodratic ns ns  ns TRT, (NP K kg/ha) Treatment Means 06 70 70.3 119.7  108.9 56670 60.9 118.8  132.7 112670 111.9 158.4  177.6 168670 107.6 180.8  190.4 CV, % 22 .5 17 .6  10 .8 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 51 Table 15a. Correlation matrices and simple correlation coefficients, Feekes growth stage 4, Perkins, OK, 20072008. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 4 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRN NDVI 1 GSNDVI 2 SPAD 0.22* 3 NDVI1 0.88* 0.14 4 NDVI2 0.88* 0.14 0.98* 5 REP 0.88* 0.24* 0.92* 0.92* 6 Area1 0.86* 0.14 0.98* 0.98* 0.88* 7 Area2 0.86* 0.14 0.98* 0.98* 0.88* 0.98* 8 NIR/G 0.88* 0.15 0.98* 0.98* 0.88* 0.98* 0.98* 9 NIR/R 0.86* 0.13 0.98* 0.98* 0.88* 0.98* 0.98* 0.98* 10 GRN NDVI 0.88* 0.17 0.98* 0.98* 0.92* 0.98* 0.98* 0.98* 0.98* 52 Table 15b. Correlation matrices and simple correlation coefficients, Feekes growth stage 5, Perkins, OK, 20072008. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 5 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRN NDVI 1 GSNDVI 2 SPAD 0.35* 3 NDVI1 0.29* 0.11 4 NDVI2 0.29* 0.11 0.98* 5 REP 0.50* 0.34* 0.86* 0.86* 6 Area1 0.44* 0.19 0.96* 0.96* 0.94* 7 Area2 0.44* 0.19 0.96* 0.96* 0.77* 0.98* 8 NIR/G 0.30* 0.10 0.98* 0.98* 0.83* 0.94* 0.94* 9 NIR/R 0.28* 0.07 0.94* 0.88* 0.77* 0.90* 0.90* 0.98* 10 GRN NDVI 0.31* 0.12 0.98* 0.98* 0.88* 0.98* 0.98* 0.98* 0.96* 53 Table 15c. Correlation matrices and simple correlation coefficients, Feekes growth stage 7, Perkins, OK, 20072008. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 7 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRN NDVI 1 GSNDVI 2 SPAD .61* 3 NDVI1 4 NDVI2 5 REP 6 Area1 7 Area2 8 NIR/G 9 NIR/R 10 GRN NDVI  54 Table 15d. Correlation matrices and simple correlation coefficients, Feekes growth stage 10, Perkins, OK, 20072008. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 10 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRNNDVI 1 GSNDVI 2 SPAD 0.28* 3 NDVI1 0.94* 0.19 4 NDVI2 0.94* 0.20* 0.98* 5 REP 0.86* 0.26* 0.94* 0.94* 6 Area1 0.92* 0.21* 0.98* 0.98* 0.94* 7 Area2 0.92* 0.21* 0.98* 0.98* 0.94* 0.98* 8 NIR/G 0.90* 0.23* 0.96* 0.96* 0.92* 0.98* 0.98* 9 NIR/R 0.94* 0.21* 0.96* 0.96* 0.92* 0.98* 0.98* 0.98* 10 GRN NDVI 0.94* 0.22* 0.98* 0.98* 0.96* 0.98* 0.98* 0.98* 0.96* 55 Table 16. Analysis of variance for GreenSeeker NDVI at different N rates and different growth stages, Perkins, OK, 20082009. Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 2 ns  ns ns Treatment 3 *  * * Error 8 0.006  0.012 0.005 Linear *  * * Quodratic ns  * * TRT, (NP K kg/ha) Treatment Means 06 70 0.240  0.294 0.326 56670 0.348  0.572 0.452 112670 0.493  0.750 0.585 168670 0.552  0.778 0.632 CV, % 9. 1  7. 7 5. 5 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 56 Table 17. Analysis of variance for SPAD readings at different N rates and different growth stages, Perkins, OK, 20082009. Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 2 ns  ns ns Treatment 3 *  * * Error 8 34.78  136.66 167.31 Linear *  * * Quodratic ns  ns ns TRT, (NP K kg/ha) Treatment Means 06 70 39.7  26.93 27.13 56670 47.77  27.43 28.87 112670 56.67  43 42.8 168670 58.53  43.83 40.63 CV, % 5. 3  13 .5 15 .1 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 57 Table 18. Analysis of variance for REP at different N rates and different growth stages, Perkins, OK, 20082009. Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 2 ns  ns ns Treatment 3 *  * * Error 8 2.43  8.85 3.15 Linear *  * * Quodratic ns  * Ns TRT, (NP K kg/ha) Treatment Means 06 70 717.98  713.19 712.87 56670 718.48  715.57 711.73 112670 720.74  721.34 719.93 168670 721.02  721.60 720.60 CV, % 0. 1  0.1 7 0. 1 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 58 Table 19. Analysis of variance AREA2 at different N rates and different growth stages, Perkins, OK, 20082009 Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 2 *  ns Ns Treatment 3 *  * * Error 8 211  1711 10437 Linear *  * * Quodratic ns  * Ns TRT, (NP K kg/ha) Treatment Means 06 70 33.66  54.66 47.88 56670 65.44  99.56 118.73 112670 128.67  205.25 202.71 168670 131.43  205.49 301.63 CV, % 7. 3  11 .9 24 .9 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 59 Table 20a. Correlation matrices and simple correlation coefficients, Feekes growth stage 4, Perkins, OK, 20082009. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 4 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRN NDVI 1 GSNDVI 2 SPAD 0.90* 3 NDVI1 0.85* 0.92* 4 NDVI2 0.85* 0.92* 0.98* 5 REP 0.66* 0.74* 0.79* 0.69* 6 Area1 0.79* 0.83* 0.96* 0.96* 0.59* 7 Area2 0.79* 0.83* 0.96* 0.96* 0.59* 0.98* 8 NIR/G 0.79* 0.83* 0.98* 0.98* 0.71* 0.98* 0.98* 9 NIR/R 0.85* 0.90* 0.98* 0.98* 0.62* 0.96* 0.96* 0.98* 10 GRN NDVI 0.79* 0.85* 0.98* 0.98* 0.59* 0.98* 0.98* 0.98* 0.96* 60 Table 20b. Correlation matrices and simple correlation coefficients, Feekes growth stage 7, Perkins, OK, 20082009. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 7 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRN NDVI 1 GSNDVI 2 SPAD 0.53* 3 NDVI1 0.86* 0.69* 4 NDVI2 0.86* 0.69* 0.98* 5 REP 0.81* 0.74* 0.92* 0.96* 6 Area1 0.85* 0.69* 0.98* 0.98* 0.96* 7 Area2 0.85* 0.69* 0.98* 0.98* 0.96* 0.98* 8 NIR/G 0.79* 0.69* 0.96* 0.96* 0.94* 0.98* 0.98* 9 NIR/R 0.85* 0.66* 0.96* 0.96* 0.94* 0.98* 0.98* 0.98* 10 GRN NDVI 0.85* 0.71* 0.98* 0.98* 0.92* 0.98* 0.98* 0.98* 0.98* 61 Table 20c. Correlation matrices and simple correlation coefficients, Feekes growth stage 10, Perkins, OK, 20082009. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 10 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRN NDVI 1 GSNDVI 2 SPAD 0.55* 3 NDVI1 0.88* 0.74* 4 NDVI2 0.86* 0.76* 0.98* 5 REP 0.61* 0.81* 0.79* 0.81* 6 Area1 0.83* 0.49* 0.88* 0.85* 0.53* 7 Area2 0.83* 0.49* 0.88* 0.85* 0.55* 0.98* 8 NIR/G 0.67* 0.67* 0.86* 0.90* 0.79* 0.71* 0.71* 9 NIR/R 0.72* 0.67* 0.90* 0.92* 0.74* 0.76* 0.76* 0.76* 10 GRN NDVI 0.79* 0.76* 0.96* 0.96* 0.85* 0.79* 0.79* 0.94* 0.94* 62 Figure 3. Plot plan for Experiment 222, Stillwater OK. WHEAT FERTILITY EXPERIMENT NO.222 Agronomy Research Station Established 1969 Plot size: 6.1m x 18 m Alley: 5.2 m Total Trial Area: 41.8 m x 159 m TRT Preplant N rate (kg N /ha) Preplant P rate (kg P2O5 / ha) Preplant K rate (kg K2O / ha) 1.* 0 67 45 2.* 45 67 45 3.* 90 67 45 4.* 135 ^ 67 45 5. 90 0 45 6. 90 34 45 7. 90 101 45 8. 90 67 0 9. 90 67 90 10.* 0 0 0 11. 135 ^ 101 90 12. 135 ^ 101 0 13. 90 67 45 (SulPoMag) N applied as 460 (Urea) P applied as 0460 (Triple Super Phosphate) K applied as 0060 (Potash) *  YP plot ^  Split 135 kg N rates to 67 kg N (fall) and 60 kg N (spring) N E S W Location: Stillwater 1, 2 – Harvest Sequence Number 1, 2 – Treatment Number 1 , 2 – Soil Sample Sequence Number OBJECTIVE: To study fertilizer nitrogen, phosphorus, and potassium in winter wheat. In recent years, this study has also been used to develop yield potential models and yield predictions through sensor based technologies. 52 49 48 45 44 41 40 37 36 33 32 29 28 25 24 21 20 17 16 13 12 9 8 5 4 1 13 13 8 5 7 3 2 11 12 6 9 1 10 4 6 2 11 5 3 8 12 10 9 1 7 4 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 Rep 2 Rep 1 Rep 2 Rep 1 51 50 47 46 43 42 39 38 35 34 31 30 27 26 23 22 19 18 15 14 11 10 7 6 3 2 13 13 8 10 7 12 9 5 2 11 1 3 4 6 3 11 1 8 6 9 12 2 5 10 4 7 52 51 50 49 48 47 46 45 44 43 42 41 40 39 38 37 36 35 34 33 32 31 30 29 28 27 Rep 4 Rep 3 Rep 4 Rep 3 63 Figure 4. Plot plan for N & P Study, Perkins, OK. PERKINS N & P STUDY SOUTHWEST AGRONOMY RESEARCH STATION ESTABLISHED 1996 Treatment Preplant N (kg N /ha) Preplant P (kg P2O5 / ha) 1 0 0 2 0 34 3 0 67 4 56 0 5 56 34 6 56 67 7 112 0 8 112 34 9 112 67 10 168 0 11 168 34 12 168 67 N E S W Location: Perkins Plot Size: 3.0 m x 9 m Alleys: 3.0 m 3 4 9 10 15 16 21 22 27 28 33 34 Rep 3 9 5 12 1 8 11 7 4 3 6 2 10 25 26 27 28 29 30 31 32 33 34 35 36 2 5 8 11 14 17 20 23 26 29 32 35 Rep 2 6 11 10 5 1 8 12 9 4 7 2 3 13 14 15 16 17 18 19 20 21 22 23 24 1 6 7 12 13 18 19 24 25 30 31 36 Rep 1 7 10 9 12 1 4 3 6 2 8 11 5 1 2 3 4 5 6 7 8 9 10 11 12 Total Trial Area: 33.5 m x 36.6 m OBJECTIVE: To evaluate nitrogen and phosphorus interactions in winter wheat. 1, 2 – Harvest Sequence Number 1, 2 – Treatment Number 1 , 2 – Soil Sample Sequence Number 64 Figure 5. GreenSeeker NDVI plotted against N rates, Stillwater, OK, 20072008. . 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 45 90 135 NDVI N rate (kg/ha) 20072008 GreenSeeker NDVI Feekes 4 Feekes 5 Feekes 7 Feekes 10 65 Figure 6. SPAD reading plotted against N rates, Stillwater, OK, 20072008. 20 25 30 35 40 45 50 0 45 90 135 SPAD Value N rate (kg/ha) 20072008 SPAD Meter Feeks 4 Feekes 5 Feekes 7 Feekes 10 66 Figure 7. Rededge position (REP) plotted against N rates, Stillwater, OK, 20072008. 711 713 715 717 719 721 0 45 90 135 Wavelength (nm) N rate (kg/ha) 20072008 Rededge position Feekes 4 Feekes 5 Feekes 7 67 Figure 8. AREA2 plotted against N rates, Stillwater, OK, 20072008. 0 50 100 150 200 250 0 45 90 135 Area N rate (kg/ha) 20072008 AREA2 Feekes 4 Feekes 5 Feekes 7 68 Figure 9. REP plotted against N rates, Stillwater, OK, 20082009. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 45 90 135 NDVI N rate (kg/ha) 20082009 GreenSeeker NDVI Feekes 4 Feekes 5 Feekes 7 Feekes10 69 Figure 10. SPAD reading plotted against N rates, Stillwater, 20082009. 20 25 30 35 40 45 50 0 45 90 135 SPAD Value N rate (kg/ha) 20072009 SPAD Meter Feeks 4 Feekes 5 Feekes 7 Feekes 10 70 Figure 11. REP against N rates, Stillwater, OK, 20082009. 711 713 715 717 719 721 0 45 90 135 Wavelength (nm) N rate (kg/ha) 20082009 Rededge position Feekes 4 Feekes 5 Feekes 7 Feekes10 71 Figure 12. AREA2 plotted against N rates, Stillwater, OK, 20082009. 0 50 100 150 200 250 0 45 90 135 Area N rate (kg/ha) 20082009 AREA2 Feekes 4 Feekes 5 Feekes 7 Feekes10 72 Figure 13. Shape of rededge and its position at Feekes 4 in the 0 kg N/ha and 135 kg N/ha plots, Stillwater, OK, 2008. 0.005 1E17 0.005 0.01 0.015 0.02 650 670 690 710 730 750 The first derivative reflectance Wavelength (nm) Rededge position and Shape 0 kg N/ha 135 kg N/ha 73 Figure 14. The first derivative reflectance in the 0 kg N /ha and 135 kg N/ha plots, Stillwater, OK, 2008. 0.005 1E17 0.005 0.01 0.015 0.02 650 670 690 710 730 750 The first derivative reflectance Wavelength (nm) Original data 0 kg N/ha 135 kg N/ha 74 Figure 15. Filtering the first derivative reflectance by averaging over 10 nm widths in the 0 kg N /ha and 135 kg N/ha plots, Stillwater, OK, 2008. 0.005 1E17 0.005 0.01 0.015 0.02 650 670 690 710 730 750 The first derivative reflectance Wavelength (nm) 10 nm filtering 0 kg N/ha 135 kg N/ha 75 Figure 16. Filtering the first derivative reflectance by weight averaging over 10 nm widths in the 0 kg N /ha and 135 kg N/ha plots, Stillwater, OK, 2008. 0.005 1E17 0.005 0.01 0.015 0.02 650 670 690 710 730 750 The first dericative reflectance Wavelength (nm) Weight filtering after 10 nm filtering 0 kg N/ha 135 kg N/ha 76 Figure 17. Filtering the first derivative reflectance by using trend fitting in the 0 kg N /ha and 135 kg N/ha plots, Stillwater, 2008. 0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 640 660 680 700 720 740 760 The first derivative reflectance Wavelength (nm) Filtering by trend fitting 0 kg N/ha 135 kg N/ha Figure 18. Visual image of winter wheat at Feekes 4 plots under conventional tillage 77 4, in the 0 kg N /ha and 135 kg N/ha tillage, Stillwater, OK, 2008. 135 kg N/ha 0 kg N/ha Figure 19. Visual image of winter wheat at Feekes 4 plots under Notill system 78 in the 0 kg N /ha and 112 kg N/ha system, Stillwater, OK, 2008. 112 kg N/ha 0 kg N/ha Figure 20. Linear relationship between different N rates in Stillwater 0 kg N/ha (r² = 0.17) y = 4.5802x + 711.79 45 kg N/ha (r² = 0.80) y = 15.746x + 705.57 712 713 714 715 716 717 718 719 720 0 Rededge postion 0 kg N/ha (r² = 0.56 y = 14.772x + 709.35 45 kg N/ha (r² = 0.29) y = 16.682x + 707.62 712 713 714 715 716 717 718 719 720 0 Rededge postion 79 REP and GreenSeeker NDVI at Feekes 4 Stillwater, OK, 20072009. 90 kg N/ha (r² = 0.73) y = 23.172x + 700.44 135 kg N/ha (r² = 0.87) y = 27.927x + 697.05 0.2 0.4 0.6 0.8 GSNDVI 20072008 Feekes 4 0.56） 90 kg N/ha (r² = 0.67) y = 6.1558x + 714 135 kg N/ha (r² = 0.88) y = 24.344x + 702.34 0.2 0.4 0.6 0.8 GSNDVI 20082009 Feekes 4 at 1 check 0 kg N/ha 45 kg N/ha 90 kg N/ha 135 kg N/ha 1 check 0 kg N/ha 45 kg N/ha 90 kg N/ha 135 kg N/ha 80 Figure 21. Linear relationship between GSNDVI and REP under the conventional tillage or notill system at all growth stages, Stillwater and Perkins, OK, 20082009. Conventional (r² = 0.66) y = 13.895x + 708.5 Notill (r² = 0.42) y = 14.612x + 710.91 710 712 714 716 718 720 722 724 726 728 0 0.2 0.4 0.6 0.8 1 Rededge position GSNDVI 20072008 Conventional Tillage Notill System Conventional (r² = 0.70) y = 9.7454x + 711.88 Notill (r² = 0.34) y = 12.423x + 711.6 710 712 714 716 718 720 722 724 726 728 0 0.2 0.4 0.6 0.8 1 Rededge position GSNDVI 20082009 Conventional Tillage Notill System 81 Figure 22. Linear relationship between REP and GreenSeeker NDVI at Feekes 4 and Feekes 5, Stillwater, OK, 20082009. Feekes 5 (r² = 0.76) y = 10.463x + 711.12 Feekes 4 (r² = 0.81) y = 7.9209x + 712.48 712 713 714 715 716 717 718 719 720 721 722 723 0 0.2 0.4 0.6 0.8 1 Rededge position GSNDVI 20082009 Feekes 4 Feekes 5 Feekes 4 (r² = 0.80) y = 15.567x + 706.26 Feekes 5 (r² = 0.77) y = 13.801x + 709.19 712 713 714 715 716 717 718 719 720 721 722 723 0 0.2 0.4 0.6 0.8 1 Rededge position GSNDVI 20072008 Feekes 4 Feekes 5 82 Feekes 4 (r² = 0.45) y = 27.031x + 30.224 Feekes 5 (r² = 0.48) y = 15.317x + 30.41 20 25 30 35 40 45 50 55 60 0 0.2 0.4 0.6 0.8 1 SPAD GSNDVI 20082009 Feekes 4 Feekes 5 Figure 23. Linear relationship between SPAD and GreenSeeker NDVI at Feekes 4 and Feekes 5, Stillwater, OK, 20082009. Feekes 4 (r² = 0.67) y = 58.428x + 0.6167 Feekes 5 (r² = 0.76) y = 25.966x + 26.023 20 25 30 35 40 45 50 55 60 0 0.2 0.4 0.6 0.8 1 SPAD GSNDVI 20072008 Feekes 4 Feekes 5 83 Figure 24. Linear relationship between SPAD and REP at Feekes 4 and Feekes 5 in Stillwater OK, 20082009. Feekes 4 (r² = 0.74） y = 3.4823x  2455.8 Feekes 5 (r² = 0.83) y = 1.7273x  1197.7 20 25 30 35 40 45 50 55 60 712 714 716 718 720 722 724 SPAD Rededge position 20072008 Feekes 4 Feekes 5 Feekes 4 (r² = 0.33) y = 1.9405x  1346.5 Feekes 5 (r² = 0.53） y = 1.8523x  1288.9 20 25 30 35 40 45 50 55 60 712 714 716 718 720 722 SPAD Rededge position 20082009 Feekes 4 Feekes 5 VITA Yumiko Kanke Candidate for the Degree of Plant and Soil Science Master of Science/Arts Thesis: RED EDGE AS A POTENTIAL INDEX FOR DETECTING DIFFERENCES IN PLANT NITROGEN STATUS IN WINTER WHEAT Major Field: Agronomy Biographical: Education: Received Bachelor of Science degree in Plant and Soil Science from the Oklahoma State University, Stillwater, Oklahoma in May 2008. Completed the requirements for the Master of Science degree from Soil Science at Oklahoma State University, Stillwater, Oklahoma in December 2009. Experience: Employed by Wheat Breeding program in Oklahoma State University as a field’s assistant May 2006 to August 2006; employed by Oklahoma State University Soil Fertility testing lab as a lab assistant August 2006 to May 2008; employed by Oklahoma State University, department of plant and soil science as a graduate research assistant, May to present. Professional Memberships: American Society of Agronomy, Soil Society of America, Phi Kappa Phi, Phi Beta Delta Honor Society for International Scholars. ADVISER’S APPROVAL: Dr. William R. Raun Name: Yumiko Kanke Date of Degree: December, 2009 Institution: Oklahoma State University Location: Stillwater, Oklahoma Title of Study: RED EDGE AS A POTENTIAL INDEX FOR DETECTING DIFFERENCES IN PLANT NITROGEN STATUS IN WINTER WHEAT Pages in Study: 83 Candidate for the Degree of Master of Science Major Field: Agronomy Scope and Method of Study: To determine whether the rededge position has potential to be better indices and ultimately more accurate fertilizer N recommendations for winter wheat compared to NDVI, especially at Feekes 4 to 5 which is crucial time to make decision for N rate. Findings and Conclusions: NDVI was sensitive to plant N response as well as different plant growth stages but the sensitivity tended to decrease as N rate increased. REP sensitivity to plant N response increased as N rate increased and with advancing plant growth. NDVI and REP were linearly correlated at all growth stages (r2=0.79). REP and SPAD meter readings were also highly correlated (r2= 0.62) as were NDVI and SPAD (r2=0.56). REP sensitivity was expected to be higher than NDVI sensitivity at high plant biomass or N concentration, but this was not obvious for winter wheat. Concerning the technical problems of measuring REP in the field, further engineering will be required to evaluate REP as an alternative N index in agronomic fields.
Click tabs to swap between content that is broken into logical sections.
Rating  
Title  Red edge as a potential index for detecting differences in plant nitrogen status in winter wheat 
Date  20091201 
Author  Kanke, Yumiko 
Keywords  NDVI, Rededge position, winter wheat 
Department  Plant & Soil Science 
Document Type  
Full Text Type  Open Access 
Abstract  Alternative indices have been proposed to better detect N status in cereals. Normalized difference vegetation index (NDVI) is a key input for the OSU Nrate algorithm that is now in commercial use. However, under certain conditions NDVI has low sensitivity; therefore rededge position (REP) has been tested as a new index. The objective of this paper was to determine whether the rededge index could be useful for detecting differences in N status for winter wheat compared to NDVI. Field research was conducted at various locations in Oklahoma. Indirect measurements were collected by using a GreenSeeker sensor, a passive spectrometer, and the SPAD meter to measure plant N status in 4 different N treatments. NDVI was sensitive to plant N response as well as different plant growth stages but the sensitivity tended to decrease as N rate increased. REP sensitivity to plant N response increased as N rate increased and with advancing plant growth. NDVI and REP were linearly correlated at all growth stages (r2=0.79). REP and SPAD meter readings were also highly correlated (r2 =0.62) as were NDVI and SPAD (r2=0.56). REP sensitivity was expected to be higher than NDVI sensitivity at high plant biomass or N concentration, but this was not obvious for winter wheat. Concerning the technical problems of measuring REP in the field, further engineering will be required to evaluate REP as an alternative N index in agronomic fields. 
Note  Thesis 
Rights  © Oklahoma Agricultural and Mechanical Board of Regents 
Transcript  RED EDGE AS A POTENTIAL INDEX FOR DETECTING DIFFERENCES IN PLANT NITROGEN STATUS IN WINTER WHEAT By YUMIKO KANKE Bachelor of Plant and Soil Science Oklahoma State University Stillwater, Oklahoma 2009 Submitted to the Faculty of the Graduate College of the Oklahoma State University in partial fulfillment of the requirements for the Degree of MASTER OF SCIENCE December, 2009 ii RED EDGE AS A POTENTIAL INDEX FOR DETECTING DIFFERENCES IN PLANT NITROGEN STATUS IN WINTER WHEAT Thesis Approved: Dr. William Raun Thesis Adviser Dr. John Solie Dr. Randy Taylor Dr. A. Gordon Emslie Dean of the Graduate College iii ACKNOWLEDGMENTS First of all I would like to thank Dr. William R. Raun who gave me many opportunities in Oklahoma, and also abroad. Dr. Raun was the person who taught me how science should be and what wisdom is. I also appreciate time with Dr. John Solie, Dr. Marvin Stone and Dr. Randy Taylor who gave thoughtful advice and support for my academic career especially from an engineering prospective. I thank soil fertility friends; Jacob Vossenkemper, Jerry May, Guilherme Torres, Emily Rutto, Birehane Desta, and Jonathan Kelly who are members of the soil fertility team. iv TABLE OF CONTENTS Chapter Page I. ABSTRACT ......................................................................................................... 1 II. INTRODUCTION ................................................................................................ 2 III. REVIEW OF LITERATURE ................................................................................. 3 IV. OBJECTIVES AND HYPOTHES ......................................................................... 7 V. MATERIALS AND METHODS ............................................................................ 8 Location .................................................................................................... 8 Data Measurements.................................................................................. 9 Spectral Calculation .................................................................................10 VI. RESULTS ..........................................................................................................14 Stillwater, 20072008 ...............................................................................14 Stillwater, 20082009 ...............................................................................16 Summary for Stillwater .............................................................................17 Perkins, 20072008 ..................................................................................18 Perkins, 20082009 ..................................................................................19 Summary for Perkins ...............................................................................20 VII. DISCUSSION AND CONCLUSIONS .................................................................22 VIII. REFERENCES ..................................................................................................26 IX. APPPENDIX ......................................................................................................30 v LIST OF TABLES Table Page 1. Analysis of variance for GreenSeeker NDVI at different N rates and different growth stages, Stillwater, OK, 20072008. ..................................................................31 2. Analysis of variance for SPAD readings at different N rates and different growth stages, Stillwater, OK, 20072008. ..................................................................32 3. Analysis of variance for rededge position (REP) at different N rates and different growth stages, Stillwater, OK, 20072008. ......................................................33 4. Analysis of variance for AREA2 at different N rates and different growth stages, Stillwater, OK, 20072008. ..............................................................................34 5a. Correlation matrices and simple correlation coefficients, Feekes growth stage 4, Stillwater, OK, 20072008. ..............................................................................35 5b. Correlation matrices and simple correlation coefficients, Feekes growth stage 5, Stillwater, OK, 20072008.….....................................................................….... 36 5c. Correlation matrices and simple correlation coefficients, Feekes growth stage 7, Stillwater, OK, 20072008.….....................................................................….... 37 5d. Correlation matrices and simple correlation coefficients, Feekes growth stage 10, Stillwater, OK, 20072008.….....................................................................….... 38 6. Analysis of variance for GreenSeeker NDVI at different N rates and different growth stages, Stillwater, OK, 20082009 ...................................................................39 7. Analysis of variance for SPAD readings at different N rates and different growth stages, Stillwater, OK, 20082009 ...................................................................40 8. Analysis of variance for REP at different N rates and different growth stages, Stillwater, OK, 20082009. ..............................................................................41 9. Analysis of variance for AREA2 at different N rates and different growth stages Stillwater, OK, 20082009. ..............................................................................42 10a. Correlation matrices and simple correlation coefficients, Feekes growth stage 4, Stillwater, OK, 20082009. ..............................................................................43 10b. Correlation matrices and simple correlation coefficients, Feekes growth stage 5, Stillwater, OK, 20082009. ..............................................................................44 10c. Correlation matrices and simple correlation coefficients, Feekes growth stage 7, Stillwater, OK, 20082009. ..............................................................................45 10d. Correlation matrices and simple correlation coefficients, Feekes growth stage 10, Stillwater, OK, 20082009. ..............................................................................46 11. Analysis of variance for GreenSeeker NDVI at different N rates and different growth stages, Perkins, OK, 20072008......................................................................47 vi 12. Analysis of variance for SPAD readings at different N rates and different growth stages, Perkins, OK, 20072008......................................................................48 13. Analysis of variance for REP at different N rates and different growth stages,Perkins, OK, 20072008. ............................................................................................... 49 14. Analysis of variance for AREA2 at different N rates and different growth stages, Perkins, OK, 20072008. .................................................................................50 15a. Correlation matrices and simple correlation coefficients, Feekes growth stage 4, Perkins, OK, 20072008. .................................................................................51 15b. Correlation matrices and simple correlation coefficients, Feekes growth stage 5, Perkins, OK, 20072008. .................................................................................52 15c. Correlation matrices and simple correlation coefficients, Feekes growth stage 7, Perkins, OK, 20072008. .................................................................................53 15c. Correlation matrices and simple correlation coefficients, Feekes growth stage 10, Perkins, OK, 20072008. .................................................................................54 16. Analysis of variance for GreenSeeker NDVI at different N rates and different growth stages, Perkins, OK, 20082009. .....................................................................55 17. Analysis of variance for SPAD readings at different N rates and different growth stages, Perkins, OK, 20082009. .....................................................................56 18. Analysis of variance for REP at different N rates and different growth stages, Perkins, OK, 20082009. .................................................................................57 19. Analysis of variance for AREA2 at different N rates and different growth stages, Perkins, OK, 20082009. .................................................................................58 20a. Correlation matrices and simple correlation coefficients, Feekes growth stage 4, Perkins, OK, 20082009. .................................................................................59 20b. Correlation matrices and simple correlation coefficients, Feekes growth stage 7, Perkins, OK, 20082009. .................................................................................60 20c. Correlation matrices and simple correlation coefficients, Feekes growth stage 10, Perkins, OK, 20082009……………………………………………………………..61 vii LIST OF FIGURES Page 1. The rededge position determined the linear method. ................................................10 2. The visula description of Area 1 and 2 .......................................................................13 3. Plot plan for Experiment 222, Stillwater, OK. ............................................................62 4. Plot plan for N & P Study, Perkins, OK. .....................................................................63 5. GreenSeeker NDVI plotted against N rates, Stillwater, OK 20072008 ......................64 6. SPAD reading plotted against N rates, Stillwater, OK 20072008. .............................65 7. REP plotted against N rates, Stillwater, OK 20072008. ............................................66 8. AREA2 plotted against N rates, Stillwater, OK 20072008. ........................................67 9. GreenSeeker NDVI plotted against N rates, Stillwater, OK 20082009. .....................68 10. SPAD reading plotted against N rates, Stillwater, OK 20082009. ...........................69 11. REP plotted against N rates, Stillwater, OK 20082009. ..........................................70 12. AREA2 plotted against N rates, Stillwater, OK 20082009. ......................................71 13. Shape of rededge and its position at Feekes 4 in the 0 kg N/ha and 135 kg N/ha plots, Stillwater, OK, 2008 ...............................................................................72 14.The first derivative reflectance in the 0 kg N/ha and 135 kg N/ha plots, Stillwater, OK, 2008. ...............................................................................................................73 15. Filtering the first derivative reflectance by averaging over 10 nm widths in the 0 kg N/ha and 135 kg N/ha plots, Stillwater, OK, 2008. ...........................................74 16. Filtering the first derivative reflectance by weight averaging over 10 nm widths in the 0 kg N/ha and 135 kg N/ha plots, Stillwater, OK, 2008. ...................................75 17. Filtering the first derivative reflectance by using trend fitting in the 0 kg N/ha and 135 kg N/ha plots, Stillwater, OK, 2008. .................................................................76 18. Visual image of winter wheat at Feekes 4, 0 kg N/ha and 135 kg N/ha under conventional tillage, Stillwater, OK, 2008. .......................................................77 19. Visual image of winter wheat at Feekes 4, 0 kg N/ha and 168 kg N/ha under notill system, Perkins, OK, 2008. .............................................................................78 20. Linear relationship between REP and GreenSeeker NDVI at Feekes 4 at different N rates, Stillwater, OK, 20072009......................................................................79 21.Linear relationship between GSNDVI and REP under the conbentional tillage or notill system at all growth stages, Stillwater and Perkins, OK, 2008200…………… 80 22. Linear relationship between REP and GreenSeeker NDVI at Feekes 4 and Feekes 5, Stillwater, OK, 20072009………………………………………………………….. 81 23. Linear relationship between SPAD and GreenSeeker NDVI at Feekes 4 and Feekes 5, Stillwater, OK, 2007009………………………………………………………. 82 24. Linear relationship between SPAD and REP at Feekes 4 and Feekes 5, Stillwater, OK, 20072009………………………………………………………………………. 83 1 ABSTRACT Alternative indices have been proposed to better detect N status in cereals. Normalized difference vegetation index (NDVI) is a key input for the OSU Nrate algorithm that is now in commercial use. However, under certain conditions NDVI has low sensitivity; therefore rededge position (REP) has been tested as a new index. The objective of this paper was to determine whether the rededge index could be useful for detecting differences in N status for winter wheat compared to NDVI. Field research was conducted at various locations in Oklahoma. Indirect measurements were collected by using a GreenSeeker sensor, a passive spectrometer, and the SPAD meter to measure plant N status in 4 different N treatments. NDVI was sensitive to plant N response as well as different plant growth stages but the sensitivity tended to decrease as N rate increased. REP sensitivity to plant N response increased as N rate increased and with advancing plant growth. NDVI and REP were linearly correlated at all growth stages (r2=0.79). REP and SPAD meter readings were also highly correlated (r2= 0.62) as were NDVI and SPAD (r2=0.56). REP sensitivity was expected to be higher than NDVI sensitivity at high plant biomass or N concentration, but this was not obvious for winter wheat. Concerning the technical problems of measuring REP in the field, further engineering will be required to evaluate REP as an alternative N index in agronomic fields. 2 INTRODUCTION Nitrogen (N) is one of the major limiting mineral nutrients for plant growth. Due to the continuous increase in fertilizer costs and growing environmental concerns associated with fertilizer use, application of N fertilizer according to plant need becomes increasingly popular due to its potential for increasing NUE and reducing input costs. To determine the optimum N rate based on plant need, optical sensing technologies have been developed to detect N status in plants. Normalized difference vegetative index (NDVI) computed from optical sensing is one of the most widely used indices for the evaluation of plant N status. There are some drawbacks to NDVI, however. It has been reported to have low sensitivity at high chlorophyll content or abundant biomass. Gitelson et al. (2002) listed several possible reasons for low senstivity of NDVI. Decreased NIR reflectance was associated with changes in leaf orientation from one growth stage to the next, reduction in chlorophyll content at senescence, and increasing soil moisture. This also causes the poor estimation of biomass once soil is covered by vegetation (Clevers and Jongschaap, 2001). To overcome these limitations, wavebands called “rededge” were employed as new spectra used to evaluate plant N conditions. Rededge wavebands are between RED and NIR and correspond to wavelengths between 680 nm to 740 nm. These bands were shown to have greater sensitivity at higher chlorophyll content, which was detected as greener biomass. Oklahoma State University (OSU) has been developing algorithms for N fertilization for various crops since the early 1990’s. The algorithms are based on the use of an optical, active light, handheld GreenSeekerTM sensor which detects the fraction of light being reflected off of the plant. The OSU algorithm uses GreenSeekerTM NDVI values as the key input for calculation of the optimum midseason N fertilization rate. Due to reported limitations of NDVI at high chlorophyll concentrations or plant biomass, it was necessary to evaluate the potential of red edge for detecting N differences.. 3 REVIEW OF LITERATURE Raun and Solie (1999) estimated worldwide cereal NUE to be approximately 33 %. Efficient N fertilization based on seasonspecific plant need is one of the methods used to increase NUE. Optical, remote sensing technologies for assessing plant N status midseason have promoted more efficient utilization of N. The optical realtime active light GreenSeekerTM sensor developed by NTech Industries, Inc., in conjunction with OSU, allows for the determination of plant responses to N fertilization. NDVI is used to estimate the Response Index (RI) and yield potential (YP) which facilitates more informed decision making associated with midseason N rate application (Raun et al., 2005). This technology has shown to deliver up to US$25 per hectare (US$10 per acre) in fertilizer savings to winter wheat producers. NDVI is highly correlated with chlorophyll content which is directly associated with photosynthetic capacity (Myneni et al., 1995). Also, NDVI is correlated with above ground biomass as well as N uptake in grain (Solie et al., 1996; Boelman et al., 2003). However, NDVI has been reported to have low sensitivity at high chlorophyll content or abundant biomass due to variation in leaf orientation, as well as the effect of soil background (Hatfield et al., 1984; Rondeaux et al., 1996; Elmore et all., 2000). To overcome these limitations, rededge has been proposed as an alternative index to sense plant conditions. Rededge is referred to as the region defined by 680 nm to 740 nm, which is between RED and NIR wavelengths. The main difference between NDVI and rededge is that NDVI reflectance stays horizontal even though the intensity changes and it is the ratio of reflectance, but rededge encumbers the study of inflection positions on the slope of reflectance, or so called rededge position (REP). Due to the slope changing and is expanded, rededge can give more information by analyzing the shape and the maximum point of the first derivative reflectance on the rededge waveband. 4 Several methods have been developed to find the REP. One is detecting the inflection point which is the maximum slope between RED and NIR (Clevers and Jongschaap, 2006). It uses the first derivative analysis to detect REP (Chen and Elvidge, 1993). Another method is the linear method. This method was introduced by Guyot and Baret (1988). The estimation of reflected maximum inflection point (Rre) is calculated by using reflectance of 670 nm and 780nm. Then the wavelength of rededge position is calculated as 700+40((Rre R700)/(R740R700)). Shafri et al (2006) reported this linear model as an estimator for REP and easier but has more soil background noise. With the linear method, the overestimation of REP 10 nm wide compared with the first derivative method was found. (Dwason and Curran, 1998) However, Dwason and Curran also reported that both methods were correlated at different chlorophyll content and the correlation coefficient of REP determined by different methods was high (R2>0.99). Mutanga and Skidmore (2007) drew attention to the double points for the rededge especially in the high N treated plant using the first derivative system. If double REP exists, the linear method is not appropriate to detect the rededge position. Cho and Skidmore (2006) also developed another method where REP is determined by the intersection of the farred and NIR lines on the first derivative reflectance. REP determined by this method increased the linear relationship with N concentration compared with the first derivative method or linear method. Clevers and Jongschaap (2006) summarized sensitivity analysis for REP. REP is influenced by chlorophyll content, leaf mesophyll structure, and LAI. On the other hand, leaf orientation, solar angle and soil background had a small influence on REP. Also by combining plant growth models with REP, they increased the estimation of yield in sugarcane. Deviation of rededgeshifts and rededgeshape associated with chlorophyll content was studied by Collin (1978), who discovered rededge shifts associated with different growth stages. Maximum derivative rededge parameters were found to be highly correlated with C/N ratio in corn leaves (Middletion et al., 2002). Rededge has the potential for accurate determinations of variation in plant biomass and chlorophyll content in winter wheat (Boochs and Kupfer, 1990). Also, the rededge index enables the user to pick up the variation of red (703nm) amplitude corresponding to different N rate responses in two wheat varieties (Boochs and Kupfer, 1990). Mutanga and Skidmore (2004) investigated band depth analysis methodology which used red edge 5 bands for better biomass estimation at high canopy density. This research showed high correlation between biomass and band depth (0.81), band depth ratio (0.83), and normalized band depth index (0.85) but the coefficient between biomass and NDVI was low (0.31) . Following previous research that LAI or plant biomass and chlorophyll concentration was related to REP, Mutanga and Skidmore (2007) showed that foliar N concentration and REP were correlated. It was discovered that REP which was established by calculating the inflection point between 680 nm and 750 nm, was highly correlated with nitrogen concentration (r =0.89). From previous research, it was required to test whether red edge has the potential to better detect N response. However, there are also some drawbacks with agronomic applications. In general, complicated methodology and the high cost of spectrometers or sensing tools restrict their use among crop producers (Daberkow and McBride, 2003). Spectrometer indices are widely available, but their use is restricted due to their high sensitivity to sunlight intensity (Kollenkark et al., 1982). REP has the potential to evaluate plant N status, but technologies employing REP tend to be costprohibitive. GreenSeeker is a commercially available, optical activelight sensor (independent of incident light) and is user friendly. However, the cost of the GreenSeeker handheld sensor is approximately US$4000, and is still considerably high for most farmers to use. To manufacture an activelight sensor employing REP would be more expensive because it requires much narrower and sensitive light bands compared to those used in the GreenSeeker sensor. For example, a hyperspectrometer must precisely detect within ± 5 nm. Research scientists at the Department of Biosystems and Agricultural Engineering at OSU have estimated the cost of such a sensor at approximately $8600, which simply employs the ratio of reflectances. From an agronomic perspective, it is crucial that timing of N application as well as the rate of N fertilizer be considered. Good application comes from a good decision. The best time to make a decision for midseason N application in winter wheat is at Feekes 4. It was reported that when midseason N application was made between Feekes 3 to 4, there was no yield loss (Boman et al., 1995). After Feekes 4, tissue damage and lower forage yields were detected from having applied foliar N. Rapid N uptake occurs between Feekes 2 to 4 and by Feekes 7, wheat takes up more than a third of the total accumulated (Waldren and Flowerday, 1979). Therefore, in winter wheat, it is essential to determine the N rate for midseason application at or before 6 Feekes 4. At this time, the wheat plant does not completely cover the ground, so NDVI is still sensitive. It is, thus, essential to investigate how NDVI and REP behave differently for early season growth of winter wheat. As an N rate recommendation tool, the SPAD meter is also commercially available. This device emits light at 650 and 940 nm. The transmittance ratio is then used for estimation of chlorophyll content. It has also been used for determining optimum N fertilization rate in wheat (Fox R.H. et al., 1994). Because it is one of the decision making tools for mid season N rate, it is essential to evaluate how it behaves differently with REP. In summary, REP should be an indicator of N status in plants. However, little research has been conducted to compare REP with the indices computed from commercially available tools, which determine midseason N rate application. Considering all the factors mentioned above it was considered prudent to investigate rededge and investigate its potential for detecting plant N status and ultimately identify optimum midseason N fertilization compared to NDVI in winter wheat. 7 OBJECTIVES AND HYPOTHES The objective of this paper was to determine whether the rededge index has the potential to be a useful index for detecting difference in N status for winter wheat compared to NDVI or commercially available instruments such as the SPAD meter. • Determine whether the REP has different behavior compared with NDVI measured from GreenSeeker at different growth stages and different N rates. • Determine whether the REP has different behavior compared with SPAD values measured from Minolta SPAD 502 chlorophyll meter at different growth stages and different N rates. • Determine whether REP could be a better index for detecting differences in plant N status. 8 MATERIALS AND METHODS A spectrometer, a chlorophyll meter (SPAD502), and a GreenSeeker handheld sensor were used to collect data in winter wheat. Measurements were taken at different growth stages (Feekes growth stages 4, 5, 7, and 10) for two cropping seasons. Location Data were collected from longterm winter wheat experimental plots located at Stillwater (Experiment # 222; Fig.3) and Perkins (N & P Study; Fig.4), Oklahoma. Experiment # 222 was established in 1969 under conventional tillage on a Kirkland silt loam (fine, mixed, superactive, thermic Udertic Paleustoll). The N & P study was initiated in 1996, also under conventional tillage on a Teller sandy loam (fineloamy, mixed, thermic Udic Argiustoll). These experiments are longterm NPK trials consisting of thirteen treatments (Experiment # 222) with four replications, and twelve treatments (N & P study) with three replications, respectively. Both were arranged in a randomized complete block design (RCBD). Five (treatments 14, and 10) and four (treatments 3,6,9 and 12) treatments were used in Experiment # 222 and in the N & P study, respectively. 9 Data Measurements Three instruments were used to obtain data for this study: the Minolta SPAD 502 meter, an Ocean Optics USB4000 spectrometer, and the GreenSeeker NTech handheld optical sensor. All of the readings were taken from a 1 m2 area in each treatment. SPAD Meter The Minolta SPAD 502 chlorophyll meter determines the relative amount of chlorophyll by measuring light transmitted or absorbed by plant leaves. The SPAD 502 is a compact meter that measures chlorophyll using optical density differences at two wavelengths (650 nm and 940 nm) with a measurement area of 2 mm x 3 mm. Twenty SPAD readings were randomly taken from winter wheat plant leaves within the 1 m2 sampling area, and subsequently averaged. Spectral Measurements The Ocean Optics USB4000 spectrometer operates with Spectrasuite (crossplatform Spectroscopy software) to measure reflectance. This spectrometer can detect reflectance from 2001100 nm at a high resolution (optical resolution of 1.5 nm full width half maximum). Reflectance of the plant canopy was computed by (the reflected light from the surface of the plant canopy minus black measurement to eliminate noise)/(incident light minus black measurement). Incident light was taken measuring reflectance of a 1m2 white board composed of Barium sulfate. Black measurement was taken measuring reflectance by covering the sensor with a cap and fabric material. Greenseeker® Sensor The GreenSeeker handheld optical sensor is an active sensor that measures reflectance in both red (671±10 nm) and near infrared (NIR; 780±10 nm) wavebands at a distance of 0.6 to 1.0 m from the canopy. It then calculates NDVI using the equation: NIR d NDVI NIR d Re Re  ρ ρ ρ ρ + = Where: ρ NIR = fraction of emitted NIR radiation from the sensed area ρRED = fraction of emitted red radiation from the sensed area Two readings per plot were these was calculated. One reading con Spectral Calculation Spectrometer readings indicies. For the REP, two methods were techniques and the linear method. because with this method, there is potential to compute REP the derivative method, spectrometer reflectance from 650 nm to 750 nm were and transported into Table Curve 2.D formula . By using the formula, the maximum point of the REP. For the linear method Figure 1. Red 10 taken with the GreenSeeker sensor, and an average of consists of 10 observations per second. were used to compute NDVI, REP, and simple ratio , applied: Derivative method by curve fitting In this study, the linear method was more focused using an active sensor. D. software and interpolated a using curve fitting first derivative was recorded as . (Figure 1), the interpretation by Clevers (1994) w Rededge position determine by the linear method , sists : For collected , was used. 11 2 ρ670 ρ780 ρ + re = (  ) (  ) 700 40* 740 700 700 ρ ρ ρ ρ λ re = + re Where: re ρ = Reflectance at estimated rededge ρ 670 = Reflectance at 670 nm ρ 700 = Reflectance at 700 nm ρ 740 = Reflectance at 740 nm ρ 780 = Reflectance at 780 nm ρ 780 = Reflectance at 780 nm re λ = Waverength of rededge position For the NIR, RED and GREEN spectral reflection, the following wave bands were selected and average reflectances were computed as the point of reflectance. NIR1 ρ = Average reflectance between 750775 nm NIR2 ρ =Average reflectance between 780805 nm ρ RED = Average reflectance between 740765 nm GREEN ρ = Average reflectance between 540565 nm Indices formulas are listed as following. 12 1. NIR RED NDVI NIR RED ρ ρ ρ ρ + = 1 1  1 2. NIR RED NDVI NIR RED ρ ρ ρ ρ + = 2 2  2 3. NIR GREEN GRNNDVI NIR GREEN ρ ρ ρ ρ + = 2 2  2 4. 2 2 IGREEN GRNRATIO NIR ρ ρ = 5. GREEN REDRATIO NIR ρ ρ 2 = Also Rededgearea was computed. Rededgearea was determined as the area between continuum line from 550 nm to 750 nm and reflectance. The reflectance at 550 nm and 750 nm was used to determine slope and intercept of the continuum line. Rededge area was determined by subtracting reflectance area from the area under the continuum line. Continuum line (λ ) y =a*λ+b ‘a’ and ‘b’ is computed from following formulas ρ 550 =a*550+b ρ 750 =a*750+b Area1 = *( ) 2 ( ) *( ) ( 1) ( ) ( 1) ( ) 750 ( ) ( 1) ( ) 550 750 550 i i i i i i y λ λ ρ ρ λ λ λ − + − − + + Σ + Σ Area2 = *( ) 2 ( ) *( ) ( 1) ( ) ( 1) ( ) 750 ( ) ( 1) ( ) 650 750 650 i i i i i i y λ λ ρ ρ λ λ λ − + − − + + Σ + Σ 13 0.2 0.3 0.4 0.5 0.6 0.7 0.8 500 550 600 650 700 750 800 Reflectance Wavelength (nm) Continuum Line Continuum Line Area1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 500 550 600 650 700 750 800 R e f l e c t a n c e Wavelength (nm) Continuum Line Area2 Figure 2. Visual description of Area1 and 2 14 RESULTS The following details results by location and year. In 2007, spectrometer measurements at Feekes 10 were excluded due to measurement errors. Stillwater 222, 20072008 GreenSeeker There were significat influences of N rate on GreenSeeker NDVI (GSNDVI) at Feekes 4, 5, 7 and 10 (Table 1). The NDVI values for this location ranged from 0.54 to 0.78, 0.50 to 0.77, 0.50 to 0.79, and 0.36 to 0.70 at Feekes 4, 5, 7, and 10 respectively. At Feekes 4, the highest GSNDVI was recorded. In general, as plant growth progressed GSNDVI decreased, especially at Feekes 10. The CV (coefficient of variation) for NDVI stayed constant and ranged from 7.4 to 7.7 between Feekes 4 to Feekes 7 but increased to 10.9 at Feekes 10. Plant N response was obvious at mature growth stages and the GreenSeeker sensor could detect these differences. Nitrogen rate and GSNDVI had significant linear relationships at all growth stages. Quadratic relationships were found only at Feekes 4 (Fig. 5). SPAD Meter There were significant influences of N rate on SPAD meter readings at Feekes growth stages 4, 5, 7 and 10. The SPAD readings for this location ranged from 32 to 48, 39 to 47, 30 to 37 and 33 to 38 at Feekes 4, 5, 7, and 10 respectively (Table 2). The CV for the SPAD readings dramatically decreased across different N rates with advancing plant growth, especially between Feekes 4 and Feekes 5 (12.7, and 4.1). It indicates that as plant growth progressed, small variability is detected by SPAD meter in different N rates and the sensitivity of the SPAD meter decreased with plant growth. Nitrogen rate 15 and SPAD readings had a significant linear relationship at all growth stages. Figure 6 shows a quadratic relationship after Feekes 7 and 10 at low N rates (0 and 45 kg N/ha). Spectral Indices There were significant influences of N rate on rededge position (REP) at all growth stages (Table 3). The range of REP increased as plant growth progressed (4.33, 3.88 and 5.39 nm at Feekes 4, 5, and 7 respectively). This trend was the same for GSNDVI. REP shifted to longer wavelengths with an increase in N rate (Fig. 7). At high N rates, shifts of REP to longer wavelength were clearly detected as growth stage increased. This means that the linear relationship between REP and N rates increased with advancing plant growth (r2=0.87, 0.98 and 0.99 at Feekes 4, 5, and 7). It indicates that REP has sensitivity to high N rates at mature growth stages, which was expected with high chlorophyll concentration and biomass. With GSNDVI and SPAD readings, N rate and REP had a significant linear relationship at all growth stages. A quadratic relationship was detected after Feekes 5. There were significant influences of N rate on rededgearea 2 (AREA2) as well as linear relationships at all growth stages (Table 4). AREA2 behaved similarly as REP increased with advancing plant growth as well as N rates (Fig. 8). Relationships Between Indices Overall, each index was highly correlated with all others at all growth stages (Tables 5a, b, c, and d). There was slightly higher correlation between REP and SPAD (r2=0.74, 0.83, and 0.81 at Feekes 4, 5, and 7 respectively) compared with correlation between GSNDVI and SPAD (r2=0.69, 0.76, and 0.66 at Feekes 4, 5, and 7 respectively) at all growth stages. It possibly indicates that REP has higher sensitivity to chlorophyll concentration than NDVI. Correlation between GSNDVI and REP was also relatively high and the correlation increased with advancing plant growth (r2=0.74, 0.77, and 0.81 at Feekes 4, 5, and 7 respectively). Similar relationships were found between the spectrometer based NDVI. There was no difference between REP and NIR/G or NIR/R. However, REP resulted in higher correlation with NIR/R and NIR/G (simple ratio) than with NDVI or GRN NDVI (normalized index). AREA2 had a good linear relationship with both REP (average r2= 0.81) and NDVI (average r2= 0.84). 16 Stillwater 222, 20082009 GreenSeeker There was a significant influence of N rate on GSNDVI at Feekes 4, 5, 7 and 10 (Table 6). The NDVI values for this location ranged from 0.35 to 0.61, 0.50 to 0.77, 0.27 to 0.60, and 0.27 to 0.60 at Feekes 4, 5, 7, and 10 respectively. Compared with 2007 2008, GSNDVI behaved differently, since the highest GSNDVI was recorded at Feekes 5 in each N rate and dramatically decreased at Feekes 7. As recorded in 20072008, the highest CV for GSNDVI was recorded at Feekes 10. Nitrogen rate and GSNDVI had a significant linear relationship at all growth stages and a quadratic relationship was detected only at Feekes 4, which was similar in 20072008 (Figure 9). SPAD Meter There were significant influences of N rate on SPAD meter readings at Feekes 4, 5, 7 and 10 (Table 7). SPAD values for this location ranged from 39 to 48, 38 to 43, 35 to 44 and 33 to 42 at Feekes 4, 5, 7, and 10 respectively. As recorded in 20072008, the CV for the SPAD readings dramatically decreased as plant growth advanced, especially between Feekes 4 and Feekes 5 (CV= 8.5, and 5.1). A linear relationship was found at all growth stages and a quadratic relationship was found after Feekes10, which was similar in 20072008 year (Fig. 10). Spectral Indices There were significant influences of N rate on rededge position (REP) at all growth stages (Table 8). The REP shifted to longer wavelengths with the increase in N rates. As recorded in 20072008, linear correlation between REP and N rates increased with advancing plant growth (r2= 0.73, 0.92, 0.97, and 0.99 at Feekes 4, 5, 7 and 10 respectively). A quadratic relationship was detected at Feekes 4 and 5 (Fig. 11). AREA2 increased with advancing plant growth as well as N rate (Table 9). The ranges were 46, 17 57, 68 and 78 at Feekes 4, 5, 7, and 10, respectively. The plant N response would be bigger as plant growth advances. At Feekes 7 and 10, AREA2 had the same response to different N rates as GSNDVI and REP (Fig. 12). Relationships Between Indices Table 10a, b, c and d shows that at Feekes 5 and 10, there was higher correlation between REP and SPAD (r2=0.53 and 0.66) but at Feekes 4 and 7, there was higher correlation between GSNDVI and SPAD (r=0.45 and 0.76 at Feekes 4 and 7) than between REP and SPAD (r2=0.32 and 0.62 at Feekes 4 and 7). This indicates that REP was more sensitive to chlorophyll content than NDVI at Feekes 5 and 10 in this year, but it was all growth stages in the previous year. Correlation coefficients between GSNDVI and REP were relatively high and the correlation stayed similar with advancing plant growth (r2=0.79, 0.85, 0.77, and 0.85 at Feekes 4, 5, 7, and 10 respectively). The same relationships were found between spectrometer based NDVI. Compared with 20072008, there was differing correlation between REP and NIR/G or NIR/R. Slightly higher correlation was found between REP and NIR/R than between REP and NIR/G. Compared with 20072008, only at Feekes 10, REP resulted higher correlation with NIR/R and NIR/G (simple ratio) than with NDVI or GRN NDVI (normalized index). Through all growth stages, AREA2 had the highest correlation with NDVI. Summary for Experiment 222 in Stillwater GSNDVI is sensitive to different plant N responses as well as different plant growth stages but the sensitivity to differing plant N response tends to be higher in REP, especially as the plant grows. SPAD is sensitive to different N rates at the early growth stages but not after Feekes 5. The accurate metric detection of REP might be challenging in the field because the change of wavelength is between 1 and 10 nm. Especially at early growth stages, the range of REP was only about 4 nm between 0 kg N/ha and 135 kg N/ha treated plant. GSNDVI and REP had a good linear relationship at all growth stages (average r2=0.79). SPAD and REP had a linear relationship (average r2=0.65) but it was not superior compared to GSNDVI and REP. GSNDVI and SPAD 18 linear relationship was average r2=0.62 and it indicated that REP was more sensitive to chlorophyll concentration than GSNDVI. Perkins, 20072008 At Feekes 7 in 2008, spectrometer data were excluded due to measurement error. At Feekes 5 in 2009, all data were excluded due to measurement error. GreenSeeker There were significant influences of N rate on GSNDVI at Feekes 4, 5, 7 and 10 (Table 11). The NDVI values for this location ranged from 0.43 to 0.56, 0.44 to 0.77, 0.53 to 0.86, and 0.46 to 0.71 at Feekes 4, 5, 7, and 10 respectively. Compared with Stillwater, the CV for GSNDVI decreased with advancing plant growth and the lowest CV across different N rates was recorded at Feekes 10. Nitrogen rate and GSNDVI had a significant linear relationship and a non quadratic relationship was recorded at all growth stages. SPAD Meter There were significant influences of N rate on SPAD meter readings at Feekes 4, 7 and 10 (Table 12). SPAD readings for this location ranged from 46 to 54, 41 to 49, 31 to 41 and 34 to 48 at Feekes 4, 5, 7, and 10 respectively. The CV for the SPAD readings stayed constant at all growth stages and its decrease was not detected like that at Stillwater. Spectral Indices At all growth stages, the range of REP stayed small, 3 to 6 nm, across different N rates (Table 13). It resulted in no significant difference of N rate on REP especially at the early growth stages which are important. The CV for REP stayed constant at all growth stages. On the other hand, N rates had significant differences on AREA2 (Table 14). At all growth stages, linear relationships were detected but no quadratic relationship was found between AREA 2 and N rates. This means that AREA2 had similar behavior with 19 GSNDVI. Feekes 10 had the greatest linear relationship (r2=0.96) between AREA2 and N rate. This indicates that AREA2 is sensitive to detect plant N response at different N rates. Relationships Between Indices As recorded at Stillwater, the relationship between SPAD and REP (average r2 =0.29) tends to be higher than between SPAD and NDVI (average r2=0.27). (Table 15a, b, c, and d) At all growth stages, GSNDVI and REP had a high linear relationship (average r2=0.76). AREA2 tended to have higher correlation with REP (average r2=0.86) than with GSNDVI (average r2=0.74). Perkins, 20082009 GreenSeeker There were significant influences of N rate on GSNDVI at Feekes 4, 5, 7 and 10 (Table 16). The GSNDVI values for this location ranged from 0.24 to 0.55, 0.30 to 0.78, and 0.33 to 0.63 at Feekes 4, 7, and 10 respectively. As recorded in 20072008, the CV for NDVI decreased with advancing plant growth and the lowest CV was recorded at Feekes 10. N rate and GSNDVI had a significant linear relationship at all growth stages and a quadratic relationship was recorded at Feekes 7 and 10. SPAD Meter There were significant influences of N rate on SPAD Meter readings (SPAD) at Feekes 4, 7 and 10 (Table 17). SPAD readings for this location ranged from 40 to 59, 26 to 44, and 27 to 41 at Feekes 4, 7, and 10 respectively. The CV for SPAD readings increased with advancing plant growth which was not observed at Stillwater. As recorded in 20072008, linear relationships and non quadratic relationships were found at all growth stages. 20 Spectral Indices There were significant influences of N rate on REP at all growth stages (Table 18). The REP shifted to longer wavelengths with increases in N rate. At Feekes 4, REP range was 3 but it increased to 8 at Feekes 7 and 10. REP had significant linear relationships with N rates at all growth stages. A quadratic relationship was detected only at Feekes 7. The AREA2 tended to increase as plant growth advanced as well as with N rate increase (Table 19). As recorded in 20072008, Feekes 10 had the best linear relationship between N rate and AREA 2 (r2=0.90, 0.89 and 0.99 at Feekes 4, 7, and 10 respectively). This means that AREA 2 was sensitive to plant growth, which ultimately could mean good sensitivity for plant biomass. Relationships Between Indices Table 20a, b, and c showed that at all growth stages, GSNDVI and REP were highly correlated (r2=0.66, 0.61, and 0.81 at Feekes 4, 7 and 10). The relationship between REP and SPAD had higher correlation (r2=0.74, 0.81 and 0.74 at Feekes 4, 7 and 10) than the relationship between GSNDVI and SPAD (r2= 0.90, 0.55 and 0.53 at Feekes 4, 7 and 10). GSNDVI tended to have a higher linear relationship with AREA2 (average r2=0.82) than REP (average r2=0.7). There were no significant differences between NIR/G and NIR/R with REP. Summary of N&P Study in Perkins GDNDVI was sensitive to different plant N response as well as different growth stages. SPAD was sensitive to different plant N responses at the early growth stages but tended to decrease as plant grows. REP was sensitive to different N rate responses as well as growth stages but the REP range from high N rates to low N rates was really small especially at the early growth stages (Range = 4 nm and 8 nm at Feekes 4 and 10). This same result was found in Stillwater. GSNDVI and REP had a good linear relationship at all growth stages (average r2=0.72). SPAD and REP had a linear relationship (average r2=0.53) which was stronger than SPAD and GSNDVI (average r2=0.47). Because the SPAD value is computed from a 21 chlorophyll meter it is highly correlated with chlorophyll concentration, and as such defines REP as more sensitive to chlorophyll concentration than NDVI. 22 DISCUSSION AND CONCLUSIONS In this study, there are three things to be discussed in terms of REP as a potential index to detect in plant N status. The first thing that needs to be discussed is the advantage and disadvantage of REP in field based research. Many studies showed that REP was highly correlated with chlorophyll content at leaf and plant canopy level (Chappelle et all.1991; Cho and Skidmore 2006). It was also shown in our research that REP was highly correlated with Chlorophyll SPAD meter readings. The position and the shape of first derivative spectrum gives more opportunities to differentiate plant N response (Cheng et al., 2005; Kupfer and et all., 1990). As described in Figure 13, the shape of the first derivative reflectance is sharper in low N treated plants compared with high N treated plants, which have two maximum points with round shape. At the same plant growth stage, not only can we differentiate plant N status by “the shape” of the first derivative reflectance but also we can distinguish by “the position” of the maximum point of the first derivative reflectance. The basic point is that you can manipulate two outcomes, “the position and the shape” of the first derivative reflectance, from rededge bands which does not happen with NDVI or simple ratio. Filella and Penuelas (1994) showed that the area of the first derivative reflectance has strong correlation with plant biomass. Another advantage is that under high biomass, NDVI became saturated but rededge position could give more accurate estimates of biomass (Filella and Penuelas,1994; Mutanga and Skidmore, 2004). The results show that GSNDVI is sensitive to different plant N responses as well as different plant growth stages but the sensitivity to different plant N response tends to be higher in REP, especially as the plant grows. However there are also disadvantages. Red edge seems very sensitive to noise. Rededge could be obtained by using a hyperspectometer but because of high contents of information per pixel, the analysis of derivative system requires the right techniques and time (Ruffin and King,1999). To eliminate noise, a smoothing method is applied but it might lose or intensively modify original data (Shafri and Yusof, 2009). Figure 13 23 shows the first derivative method of original data. It does not clearly show the maximum points both at low and high N rates. Figure 14 shows the first derivative which is filtered by 10 nm. It shows a smoother line than Figure 15 and could detect two peaks in the high N rate; however it still showed noise at the low N rate. Figure 16 shows one more weighed filtering at 10 nm. Of course the line gets smoother but loses one of two peaks at the high N rate and modifies the shape of the first derivative reflectance compared with Figure 15. In Figure 17, the first derivative was computed through a trend fitting method using TableCurve 2D v5.01. A peak of the first derivative was clear but it loses a lot of information especially the shape of the first derivative reflectance at the high N rate compared with original data. Engineers developed active sensors like the GreenSeeker or transmittance sensors like the SPAD meter for field application. The purpose of the active sensor development is to eliminate noise which is caused by solar radiation. Solar intensity changed dramatically through the day and it limited the application of sensing technology in fields. For rededge detection, a passive sensor would be required. It means that the energy source depends on nonstable solar radiation and may limit the practical application in fields. Especially for the N rate recommendation tool, target users are farmers who do not concern themselves with taking data at noon or at night. They take data whenever they can and it could be the negative points of passive sensors because they create noise. It might have the potential to determine rededge position by using optical sensor using a linear method. However, it requires more spectral bands compared to NDVI and it will increase the cost for the field application. With that, in this wheat study, the range of rededge position was narrow, not more than 15 nm, between nonN treated plants and high N treated plants. So even if rededge is more sensitive to chlorophyll content or biomass, if it requires passive measurements, it is speculative whether the sensitivity is worth it to look at agronomic fields. It is also unclear if they can overcome noise and clearly detect narrow wavebands, whether the technology allows computing rededge for on the go sensing and optimum N recommendations. From the agronomic application aspect, hyperspectrometers cost more than band spectrometers which are used in the field today. It needs to be asked if today’s engineer could develop a sensor capable of detecting rededge position clearly with the farmers’ budget in mind. To explain these questions, futher engineering investigation is required. 24 Second, the relationship between NDVI and REP needs to be discussed. Researchers have been discussing that the REP is the alternative index due to the high sensitivity under dense green biomass. In Oklahoma, farmers apply mid season N at Feekes 4 and 5. So they need to make a decision for the N rate before Feekes 5. At that time, the ground is not fully covered by biomass (Figure 18 and 19). So NDVI saturation should not be a problem. But with the decrease in biomass on the ground, other problems arise. With a decrease in plant cover on the ground, the soil is more exposed. It might increase the noise for NDVI. In Figure 20, the relationship between REP and GSNDVI is described at Feekes 4. Under the high N rate which was expected to have more biomass on the ground (Figure 18 and 19), the correlation between REP and GSNDVI was high but it decreased as N rate decreased. The decrease in correlation could be explained by the reduction of plant biomass which ultimately increases soil exposed area and it influences differently on the value of REP and NDVI. Some research noted that there is less influence of soil background on REP. Therefore, NDVI might be more influenced by soil background and it ends up behaving differently than REP. This statement also could explain Figure 21. Conventional tillage had a better linear correlation between REP and GSNDVI compared with no till. Due to remaining residues in no till, the soil background was less homogenous compared with conventional tillage and it could reduce the relationship between REP and GSNDVI. To verify the question, further research is required by taking biomass under the different soil backgrounds. But from an agronomic view, farmers put out preplant N and good biomass should cover the soil. It means that it is not like the behavior between REP and NDVI in the check or 0 kg N/ha. Therefore correlation above r2=0.7 is estimated at Feekes 4 in practical field. In this research, there was no biomass data, but at Feekes 4 and 5, REP and NDVI were strongly correlated (average r= 0.88) (Figure 22). The third aspect to be discussed is whether REP may provide for the determination of optimum N rates in winter wheat. In the OSU N algorithm, N rate is determined based on the following factors: plant N response index (RI) and inseason estimated yield (INSEY), which is computed by NDVI divided by the number of days with growing degree days (GDD>0) from planting to sensing, estimated yield potential without additional N (YP0) and with additional N (YPN) computed from the INSEY formula, grain N uptake, and nitrogen use efficiency. OSU recognized the increasing nitrogen use 25 efficiency more than 15 % using this algorithm. NDVI has been used because of its strong relationship with biomass. Research has shown REP is strongly correlated with chlorophyll content. However, N concentration which is ultimately correlated with chlorophyll concentration, was weakly correlated with yield (Holtz et all., 2008). Therefore, it is essential to investigate the ability of REP to predict biomass. Also, to utilize the REP as an index, REP has to increase the sensitivity of estimation of plant biomass significantly, especially at Feekes 4 to 5, because it is an important time from agronomic perspective to determine N rates. Under the current OSU N algorithm, the 0.01 NDVI difference could change the N rate between 1 kg/ha and 5.6 kg/ha in winter wheat. If REP could detect plant biomass more accurately, it is important how sensitive it is. Even if REP could detect 0.01 NDVI differences and change the N rates to, for example, 1 kg N/ha to 1.5 kg N/ha, farmers will put down additional 2 kg N/ha anyway. They do not require that much precision. In conclusion, REP behaves very similar to NDVI. Because of economical inputs needed to implement REP into sensor based technology which enables us to apply this information in agronomic fields, at this time, more evidence is needed that suggest REP is significantly better than NDVI. The data suggest that this new sensor may not be worth the investment in winter wheat. Further research is needed to evaluate REP with biomass at the early plant growth stage. 26 REFERENCES 1. Boelman, N.T., M. Stieglitz, H.M. Rueth, M. Sommerkorn, K.L. Griffin, G.R. Shaver, and J.A. Gamon. 2003. Response of NDVI, biomass, and ecosystem gas exchange to longterm warming and fertilization in wet sedge tundra. Oecologia. 135:414–421. 2. Boman., R.K., R.L. Westerman, W.R. Raun. And M.E. Jojola. 1995. Time of nitrogen application: Effects on winter wheat and residual soil nitrate. Soil.Sci.Soc. Am.J. 59:1364`369. 3. Boochs., F. and G. Kupfer. 1990. Shape of the red edge as vitality indicator for plants. Int. J. Remote Sens. 11:17411753. 4. Chappelle, E.W., M.S. Kim, and J.E. McMurtrey III. 1991. The red edge shift: An explanation of its relationship to stress and the concentration of chlorophyll a. Proceedings of the 11th Annual International Geoscience and Remote Sens. Symposium, Espoo, Finland; UNITED STATES; 36 June 1991. 22872290. 5. Chen, Z., and Elvidge, C. D., 1993, Description of derivativebased high spectralresolution (AVIRIS) green vegetation index. Proc. SPIE. 1937: 43 54. 6. Cheng., Y., C. Hu, H. Dai and Y. Lei. 2005. Spetral red edge parameters for winter wheat under different nitrogen support levels. Proceeding of SPIEThe international society for optical engineering. 5884:16. 7. Cho, M.A. and A.K. Skidmore. 2006. A new technique for extracting the res edge position from hyperspectral data: The linear extrapolation method. Remote Sensing of Environment. 101:181193. 8. Clevers, J. G. P. W. 1994. Plant vitality and yield indicators. Imaging spectrometryA tool for environmental observation. 4:193219. 9. Clevers, J. G. P. W., S. M. de Jong, , G. F. Epema, , F.V. der Meer, , W. H. Bakker, A. K. Skidmore and E.A. Addink, 2001. MERIS and the rededge 27 10. position. International journal of applied earth observation and geoinformation (JAG). 3: 313320. 11. Clevers, J. G. P. W. and R. Jongschaap. 2006. Imaging spectrometry for agriculture applications. Imaging spectrometry: Basic principal and prospective application. 157197. Dordrecht, Netherlands. Springer. 12. Collins, W. 1978. Remote sensing of crop trpe and maturity. Photogr. Engineer. Remote Sensing. 44:4355. 13. Elmore, A.J., J.F. Mustard, S.J. Manning, and D.B. Lobell. 2000. Quantifying vegetation change in semiarid environments precision and accuracy of spectral mixture analysis and the normalized difference vegetation index. Remote Sens. Environ. 73:87102. 14. Daberkow, S.D. and W.D McBride. 2003. Farm and operator characteristics affecting the awareness and adoption of precision agriculture technologies in the U.S. Precision Agriculture. In: Precision Agriculture 2001: Proceedings of the 3rd European Conference on precision Agriculture. 4:163177. 15. Dawson., T.P., and P.J. Curran. 1998. A new technique for interpolating the reflectance red edge position. Int. J. Remote Sens. 19:21332139. 16. Filella, I. and J. Penuelas. 1994. The red edge positon and shape as indicators of plant chlorophyll content, biomass and hydric status. 1994. INT. J. Remote Sens. 15: 14591470. 17. Fox R.H., W.P. Piekielek, and K.M. Macneal. 1994. Using a Chlorophyll Meter to Predict Nitrogen Fertilizer Needs of Winter Wheat. Soil. Sci. Plant. Anal. 25:171 181. 18. Gitelson. A.A., Y.J. Kaufman, R. Stark and D.C. Rundquist. Novel algorithms for remote estimation of vegetation fraction. 2002. Remote Sens. Environ. 80:76–87. 19. Guyot, G., & Baret, F. (1988). Utilisation de la haute résolution spectrale pour suivre l'état des couverts végétaux. Proceedings of the 4th International colloquium on spectral signatures of objects in remote sensing. ESA SP287, Assois, France. 279−286. 20. Hatfield, J. L., G. Asrar, and E. T. Kanemasu. 1984. Intercepted photosynthetically active radiation estimated by spectral reflectance. Remote Sens. Environ. 14: 6575. 21. Holtz, S., K. Grima, B. Tubana, D.B. Arnall, D. Edmond, Y. Kanke, J.B. Solie, and W.R. Raun. 2008. Above ground nitrogen accumulation as a function of time in corn and winter wheat. J. Plant Nutr. ??? 28 22. Kollenkark, J. C., Vanderbilt, V. C., Daughtry, C. S. T., and Bauer, M. E. 1982. Influence of solar illumination angle on soybean canopy reflectance. Appl. Opt. 21:11791184. 23. Kupfer, F.B.G., K. Dockter and W. Kuhbauch. 1990. Shape of the red edge as vitality indicator for plants. INT. J. Remote Sens. 11:17411753. 24. Middleton, E.M., P.K. E. Campbell, J.E. McMurtrey, L.A. Corp, L.M. Butcher, and E.W. Chappelle. 2002. “Red edge” optical properties of corn leaves from different nitrogen regimes. In: IEEE International Geoscience and Remote Sensing Symposium, 2002. IGRASS 2002. 4:22082210. 25. Mutanga, O. and Skidmore, A.K. 2004. Hyperspectral band depth analysis for a better estimation of grass biomass (Cenchrus ciliaris) measured under controlled laboratory conditions. International Journal of Applied earth Observation and Geoinformation. 5:8796. 26. Mutanga, O. and Skidmore, A.K. 2004. Narrow band vegetation indices overcome the saturation problem in biomass estimation. Int. J. Remote Sens. 25:39994014. 27. Mutanga, O. and Skidmore, A.K. 2007. Red edge shift and biochemical content in grass canopies. ISPRS Journal of Photogrammetry and Remote Sensing. 62:3442. 28. Myneni, R.B., F.G. Hall, P.J. Sellers, and A.L. Marshak. 1995. The interpretation of spectral vegetation indexes. Geosci. Remote Sens. 33: 481486. 29. Raun, W.R., and G.V. Johnson. 1999. Improving nitrogen use efficiency for cereal production. Agron. J. 91:357–363. 30. Raun, W.R., J.B. Solie, G.V. Johnson, M.L. Stone, R.W. Mullen, K.W. Freeman, W.E. Thomason, and E.V. Lukina. 2002. Improving nitrogen use efficiency in cereal grain production with optical sensing and variable rate application. Agron. J. 94:815–820. 31. Raun, W.R., J.B. Solie, M.L. Stone, K.L. Martin, K.W. Freeman, R.W. Mullen, H. Zhang, J.S. Schepers, and G.V. Johnson. 2005. Optical sensorbased algorithm for crop nitrogen fertilization. Commun. Soil Sci. Plant Anal. 36:2759–2781. 32. Rondeaux, G., M. Steven, and F. Baret. 1996. Optimization of soiladjusted vegetation indices. Remote Sens. Environ. 55: 95–107. 33. Ruffin,C., and King, R. L. 1999. The analysis of hyperspectral data using SavitzkyGolayfiltering Theoretical basis In: IEEE International Geoscience and 29 Remote Sensing Symposium, 1999. CD ROMIEEE Catalog Number 99CH36293. 34. Shafri, H.Z.M., M.A.M Shalleh and A. Ghiyamat. 2006. Hyperspectral remote sensing of vegetation using red edge position. Am. J. Appl. Sci3. 6:18641871 35. Shafri, H.Z.M., and M.R.M. Yusof. 2009. Trends and issues in noise reduction for hyperspectral vegetation reflectance spectra. European J. Sci. Res. 23:404410. 36. Solie, J.B., W.R. Raun, R.W. Whitney, M.L. Stone and J.D. Ringer. 1996. Optical sensor based field element size and sensing strategy for nitrogen application. Trans. ASAE 39(6):19831992. 37. Waldren, R.P and A.D. Flowerday. 1979. Growth stages and distribution of dry matter, N, P, and K in winter wheat. Agron. J. 71:391397. 30 APPPENDIX 31 Table 1. Analysis of variance for GreenSeeker NDVI at different N rates and different growth stages, Stillwater, OK, 20072008. Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 3 ns ns ns ns Treatment 3 * * * * Error 28 0.06219 0.06211 0.05793 0.0842 Linear * * * * Quodratic * ns ns ns TRT, (NP K kg/ha) Treatment Means 067 45 0.543 0.505 0.495 0.362 456745 0.631 0.595 0.589 0.457 906745 0.762 0.717 0.713 0.609 1356745 0.777 0.775 0.794 0.696 CV, % 7. 4 7. 7 7. 4 10 .9 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 32 Table 2. Analysis of variance for SPAD readings at different N rates and different growth stages, Stillwater, OK, 20072008. Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 3 ns ns ns Ns Treatment 3 * * * * Error 28 637.6 108.7 45.3 143.7 Linear * * * * Quodratic ns ns * * TRT, (NP K kg/ha) Treatment Means 067 45 31.78 38.55 29.25 34.61 456745 35.13 40.01 30.21 33.46 906745 44.58 44.66 33.86 34.36 1356745 47.81 46.88 37.31 37.67 CV, % 12 .7 4. 9 4. 1 6. 8 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 33 Table 3. Analysis of variance for rededge position (REP) at different N rates and different growth stages, Stillwater, OK, 20072008. Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 3 ns * ns  Treatment 3 * * *  Error 28 21.96 21.22 7.83  Linear * * *  Quodratic ns ns *  TRT, (NP K kg/ha) Treatment Means 067 45 714.28 716.16 715.72  456745 715.51 716.44 716.38  906745 718.10 718.9 718.47  1356745 718.61 720.46 721.11  CV, % 0.1 3 0.1 3 0.0 8  *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 34 Table 4. Analysis of variance for AREA2 at different N rates and different growth stages, Stillwater, OK, 20072008. Source of Variation Df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 3 ns ns *  Treatment 3 * * *  Error 28 7256 3309 5053  Linear * * *  Quodratic ns ns ns  TRT, (NP K kg/ha) Treatment Means 067 45 77.29 83.17 101.64  456745 109.33 97.74 118.46  906745 137.23 127.13 156.49  1356745 153.45 138.93 193.32  CV, % 14 .3 10 .3 10 .1  *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 35 Table 5a. Correlation matrices and simple correlation coefficients, Feekes growth stage 4, Stillwater, OK, 20072008. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 4 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRN NDVI 1 GSNDVI 2 SPAD 0.69* 3 NDVI1 0.69* 0.52* 4 NDVI2 0.69* 0.52* 0.98* 5 REP 0.74* 0.74* 0.76* 0.76* 6 Area1 0.83* 0.76* 0.55* 0.55* 0.71* 7 Area2 0.85* 0.74* 0.53* 0.53* 0.67* 0.98* 8 NIR/G 0.49* 0.46* 0.98* 0.98* 0.71* 0.44* 0.46* 9 NIR/R 0.66* 0.53* 0.98* 0.98* 0.77* 0.50* 0.53* 0.98* 10 GRN NDVI 0.61* 0.45* 0.98* 0.98* 0.69* 0.45* 0.46* 0.98* 0.98* 36 Table 5b. Correlation matrices and simple correlation coefficients, Feekes growth stage 5, Stillwater, OK, 20072008. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 5 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRN NDVI 1 GSNDVI 2 SPAD 0.76* 3 NDVI1 0.92* 0.74* 4 NDVI2 0.92* 0.74* 0.98* 5 REP 0.77* 0.83* 0.81* 0.81* 6 Area1 0.90* 0.74* 0.94* 0.94* 0.83* 7 Area2 0.92* 0.74* 0.96* 0.96* 0.86* 0.98* 8 NIR/G 0.90* 0.79* 0.98* 0.98* 0.88* 0.96* 0.96* 9 NIR/R 0.90* 0.79* 0.98* 0.98* 0.88* 0.94* 0.96* 0.98* 10 GRN NDVI 0.92* 0.76* 0.98* 0.98* 0.83* 0.94* 0.96* 0.98* 0.98* 37 Table 5c. Correlation matrices and simple correlation coefficients, Feekes growth stage 7, Stillwater, OK, 20072008. Feekes 7 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRN NDVI 1 GSNDVI 2 SPAD 0.66* 3 NDVI1 0.83* 0.64* 4 NDVI2 0.83* 0.66* 0.98* 5 REP 0.81* 0.81* 0.83* 0.85* 6 Area1 0.66* 0.62* 0.59* 0.59* 0.77* 7 Area2 0.77* 0.71* 0.76* 0.76* 0.90* 0.90* 8 NIR/G 0.83* 0.77* 0.92* 0.92* 0.94* 0.90* 9 NIR/R 0.81* 0.71* 0.96* 0.96* 0.88* 0.69* 0.83* 0.98* 10 GRN NDVI 0.85* 0.76* 0.96* 0.96* 0.92* 0.58* 0.72* 0.98* 0.96*  *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. 38 Table 5d. Correlation matrices and simple correlation coefficients, Feekes growth stage 10, Stillwater, OK, 20072008. Feekes 10 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRN NDVI 1 GSNDVI 2 SPAD .26* 3 NDVI1 4 NDVI2 5 REP 6 Area1 7 Area2 8 NIR/G 9 NIR/R 10 GRN NDVI *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. 39 Table 6. Analysis of variance for GreenSeeker NDVI at different N rates and different growth stages, Stillwater, OK, 20082009. Source of Variation Df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 3 * ns ns ns Treatment 3 * * * * Error 28 0.034 0.062 0.038 0.065 Linear * * * * Quodratic * ns ns ns TRT, (NP K kg/ha) Treatment Means 067 45 0.358 0.505 0.266 0.273 456745 0.519 0.595 0.402 0.419 906745 0.599 0.717 0.487 0.470 1356745 0.612 0.775 0.596 0.604 CV, % 7. 1 7. 7 8. 9 11 .6 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 40 Table 7. Analysis of variance for SPAD readings at different N rates and different growth stages, Stillwater, OK, 20082009 Source of Variation Df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Means Squares Replication 3 ns ns ns ns Treatment 3 * * * * Error 28 356.7 112.43 104.34 180.9 Linear * * * * Quodratic ns ns ns * TRT, (NP K kg/ha) Treatment Means 067 45 39.18 37.55 35.43 32.91 456745 44.10 39.80 38.73 33.17 906745 45.6 40.88 42.05 37.43 1356745 48.35 42.95 44.03 42.46 CV, % 8. 5 5. 3 5. 1 7. 4 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 41 Table 8. Analysis of variance for REP at different N rates and different growth stages, Stillwater, OK, 20089. Source of Variation Df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 3 ns ns ns ns Treatment 3 * * * * Error 28 12.009 4.323 24.76 19.54 Linear * * * * Quodratic * * ns ns TRT, (NP K kg/ha) Treatment Means 067 45 714.64 716.26 714.45 714.15 456745 716.27 717.32 716.22 715.97 906745 717.68 718.27 717.69 717.52 1356745 717.64 718.58 718.48 719.65 CV, % 0. 1 0.0 6 0.1 4 0.1 2 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 42 Table 9. Analysis of variance for AREA2 at different N rates and different growth stages, Stillwater, OK, 20082009. Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 3 * * ns ns Treatment 3 * * * * Error 28 3629 5145 4122 7089 Linear * * * * Quodratic ns * ns ns TRT, (NP K kg/ha) Treatment Means 067 45 46.46 78.48 98.64 85.81 456745 67.77 122.81 115.36 119.18 906745 84.99 154.76 149.7 147.14 1356745 92.37 135.91 166.54 164.09 CV, % 16 .6 11 .7 30 .6 13 .1 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 43 Table 10a. Correlation matrices and simple correlation coefficients, Feekes growth stage 4, Stillwater, OK, 20082009. Feekes 4 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRNNDVI 1 GSNDVI 2 SPAD 0.45* 3 NDVI1 0.94* 0.45* 4 NDVI2 0.94* 0.46* 0.98* 5 REP 0.79* 0.32* 0.83* 0.83* 6 Area1 0.34* 0.03* 0.36* 0.38* 0.46* 7 Area2 0.50* 0.10* 0.52* 0.55* 0.46* 0.96* 8 NIR/G 0.90* 0.50* 0.96* 0.96* 0.76* 0.26* 0.42* 9 NIR/R 0.92* 0.48* 0.98* 0.98* 0.81* 0.40* 0.56* 0.96* 10 GRN NDVI 0.92* 0.52* 0.98* 0.96* 0.77* 0.25* 0.42* 0.98* 0.96* *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. 44 Table 10b. Correlation matrices and simple correlation coefficients, Feekes growth stage 5, Stillwater, OK, 20082009. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 5 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRNNDVI 1 GSNDVI 2 SPAD 0.49* 3 NDVI1 0.81* 0.41* 4 NDVI2 0.81* 0.40* 0.98* 5 REP 0.85* 0.53* 0.83* 0.83* 6 Area1 0.56* 0.26* 0.69* 0.69* 0.61* 7 Area2 0.67* 0.29* 0.85* 0.85* 0.72* 0.96* 8 NIR/G 0.77* 0.35* 0.98* 0.98* 0.76* 0.66* 0.83* 9 NIR/R 0.81* 0.40* 0.98* 0.98* 0.83* 0.66* 0.81* 0.98* 10 GRN NDVI 0.77* 0.35* 0.98* 0.98* 0.76* 0.67* 0.83* 0.98* 0.98* 45 Table 10c. Correlation matrices and simple correlation coefficients, Feekes growth stage 7, Stillwater, OK, 20082009. Feekes 7 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRNNDVI 1 GSNDVI 2 SPAD 0.76* 3 NDVI1 0.50* 0.32* 4 NDVI2 0.50* 0.34* 0.98* 5 REP 0.77* 0.62* 0.52* 0.53* 6 Area1 0.41* 0.25* 0.71* 0.69* 0.59* 7 Area2 0.42* 0.27* 0.76* 0.74* 0.59* 0.96* 8 NIR/G 0.44* 0.31* 0.81* 0.81* 0.56* 0.81* 0.86* 9 NIR/R 0.45* 0.35* 0.76* 0.76* 0.55* 0.83* 0.86* 0.98* 10 GRN NDVI 0.45* 0.32* 0.85* 0.85* 0.58* 0.81* 0.85* 0.98* 0.92* *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. 46 Table 10d. Correlation matrices and simple correlation coefficients, Feekes growth stage 10, Stillwater, OK, 20082009. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 10 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRNNDVI 1 GSNDVI 2 SPAD 0.55* 3 NDVI1 0.71* 0.44* 4 NDVI2 0.71* 0.44* 0.98* 5 REP 0.85* 0.66* 0.85* 0.85* 6 Area1 0.74* 0.50* 0.94* 0.94* 0.88* 7 Area2 0.74* 0.50* 0.94* 0.94* 0.88* 0.98* 8 NIR/G 0.67* 0.52* 0.98* 0.98* 0.85* 0.92* 0.92* 9 NIR/R 0.71* 0.53* 0.96* 0.96* 0.86* 0.92* 0.92* 0.98* 10 GRN NDVI 0.67* 0.45* 0.98* 0.98* 0.83* 0.92* 0.92* 0.98* 0.98*  47 Table 11. Analysis of variance for GreenSeeker NDVI at different N rates and different growth stages, Perkins, OK, 20072008. Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 2 ns ns ns ns Treatment 3 * * * * Error 8 0.029 0.031 0.03 0.016 Linear * * * * Quodratic ns ns ns ns TRT, (NP K kg/ha) Treatment Means 06 70 0.429 0.436 0.525 0.455 56670 0.419 0.576 0.618 0.553 112670 0.537 0.697 0.805 0.698 168670 0.559 0.771 0.856 0.708 CV, % 14.3 11.5 10.1 8.6 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 48 Table 12. Analysis of variance for SPAD readings at different N rates and different growth stages, Perkins, OK, 20082009. Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 2 ns ns Ns Ns Treatment 3 * * * Ns Error 8 40.43 48.59 28.22 36.69 Linear * * * Ns Quodratic ns ns Ns Ns TRT, (NP K kg/ha) Treatment Means 06 70 45.76 40.66 30.93 33.67 56670 50.2 41.33 34.63 34.83 112670 53.9 44.93 40.9 34.57 168670 53.5 48.8 40.93 37.57 CV, % 5.1 6.47 5.855 7.85 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 49 Table 13. Analysis of variance for REP at different N rates and different growth stages, Perkins, OK, 20072008. Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 2 Ns ns  ns Treatment 3 Ns Ns  * Error 8 6.61 9.08  9.4 Linear Ns *  * Quodratic Ns ns  ns TRT, (NP K kg/ha) Treatment Means 06 70 717.44 718.16  714.32 56670 717.17 718.2  716.42 112670 719.44 719.95  719.33 168670 719.55 721.19  719.95 CV, % 0.1 5 0.1 7  0.1 7 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 50 Table 14. Analysis of variance AREA2 at different N rates and different growth stages, Perkins, OK, 20072008 Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 2 ns *  ns Treatment 3 * ns  * Error 8 2337 3862  1630 Linear * *  * Quodratic ns ns  ns TRT, (NP K kg/ha) Treatment Means 06 70 70.3 119.7  108.9 56670 60.9 118.8  132.7 112670 111.9 158.4  177.6 168670 107.6 180.8  190.4 CV, % 22 .5 17 .6  10 .8 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 51 Table 15a. Correlation matrices and simple correlation coefficients, Feekes growth stage 4, Perkins, OK, 20072008. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 4 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRN NDVI 1 GSNDVI 2 SPAD 0.22* 3 NDVI1 0.88* 0.14 4 NDVI2 0.88* 0.14 0.98* 5 REP 0.88* 0.24* 0.92* 0.92* 6 Area1 0.86* 0.14 0.98* 0.98* 0.88* 7 Area2 0.86* 0.14 0.98* 0.98* 0.88* 0.98* 8 NIR/G 0.88* 0.15 0.98* 0.98* 0.88* 0.98* 0.98* 9 NIR/R 0.86* 0.13 0.98* 0.98* 0.88* 0.98* 0.98* 0.98* 10 GRN NDVI 0.88* 0.17 0.98* 0.98* 0.92* 0.98* 0.98* 0.98* 0.98* 52 Table 15b. Correlation matrices and simple correlation coefficients, Feekes growth stage 5, Perkins, OK, 20072008. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 5 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRN NDVI 1 GSNDVI 2 SPAD 0.35* 3 NDVI1 0.29* 0.11 4 NDVI2 0.29* 0.11 0.98* 5 REP 0.50* 0.34* 0.86* 0.86* 6 Area1 0.44* 0.19 0.96* 0.96* 0.94* 7 Area2 0.44* 0.19 0.96* 0.96* 0.77* 0.98* 8 NIR/G 0.30* 0.10 0.98* 0.98* 0.83* 0.94* 0.94* 9 NIR/R 0.28* 0.07 0.94* 0.88* 0.77* 0.90* 0.90* 0.98* 10 GRN NDVI 0.31* 0.12 0.98* 0.98* 0.88* 0.98* 0.98* 0.98* 0.96* 53 Table 15c. Correlation matrices and simple correlation coefficients, Feekes growth stage 7, Perkins, OK, 20072008. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 7 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRN NDVI 1 GSNDVI 2 SPAD .61* 3 NDVI1 4 NDVI2 5 REP 6 Area1 7 Area2 8 NIR/G 9 NIR/R 10 GRN NDVI  54 Table 15d. Correlation matrices and simple correlation coefficients, Feekes growth stage 10, Perkins, OK, 20072008. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 10 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRNNDVI 1 GSNDVI 2 SPAD 0.28* 3 NDVI1 0.94* 0.19 4 NDVI2 0.94* 0.20* 0.98* 5 REP 0.86* 0.26* 0.94* 0.94* 6 Area1 0.92* 0.21* 0.98* 0.98* 0.94* 7 Area2 0.92* 0.21* 0.98* 0.98* 0.94* 0.98* 8 NIR/G 0.90* 0.23* 0.96* 0.96* 0.92* 0.98* 0.98* 9 NIR/R 0.94* 0.21* 0.96* 0.96* 0.92* 0.98* 0.98* 0.98* 10 GRN NDVI 0.94* 0.22* 0.98* 0.98* 0.96* 0.98* 0.98* 0.98* 0.96* 55 Table 16. Analysis of variance for GreenSeeker NDVI at different N rates and different growth stages, Perkins, OK, 20082009. Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 2 ns  ns ns Treatment 3 *  * * Error 8 0.006  0.012 0.005 Linear *  * * Quodratic ns  * * TRT, (NP K kg/ha) Treatment Means 06 70 0.240  0.294 0.326 56670 0.348  0.572 0.452 112670 0.493  0.750 0.585 168670 0.552  0.778 0.632 CV, % 9. 1  7. 7 5. 5 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 56 Table 17. Analysis of variance for SPAD readings at different N rates and different growth stages, Perkins, OK, 20082009. Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 2 ns  ns ns Treatment 3 *  * * Error 8 34.78  136.66 167.31 Linear *  * * Quodratic ns  ns ns TRT, (NP K kg/ha) Treatment Means 06 70 39.7  26.93 27.13 56670 47.77  27.43 28.87 112670 56.67  43 42.8 168670 58.53  43.83 40.63 CV, % 5. 3  13 .5 15 .1 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 57 Table 18. Analysis of variance for REP at different N rates and different growth stages, Perkins, OK, 20082009. Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 2 ns  ns ns Treatment 3 *  * * Error 8 2.43  8.85 3.15 Linear *  * * Quodratic ns  * Ns TRT, (NP K kg/ha) Treatment Means 06 70 717.98  713.19 712.87 56670 718.48  715.57 711.73 112670 720.74  721.34 719.93 168670 721.02  721.60 720.60 CV, % 0. 1  0.1 7 0. 1 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 58 Table 19. Analysis of variance AREA2 at different N rates and different growth stages, Perkins, OK, 20082009 Source of Variation df Feekes 4 Feekes 5 Feekes 7 Feekes 10 Mean Squares Replication 2 *  ns Ns Treatment 3 *  * * Error 8 211  1711 10437 Linear *  * * Quodratic ns  * Ns TRT, (NP K kg/ha) Treatment Means 06 70 33.66  54.66 47.88 56670 65.44  99.56 118.73 112670 128.67  205.25 202.71 168670 131.43  205.49 301.63 CV, % 7. 3  11 .9 24 .9 *Significant at the 0.05 probability level; df = degrees of freedom; ns =not significant; CV=coefficient of variation 59 Table 20a. Correlation matrices and simple correlation coefficients, Feekes growth stage 4, Perkins, OK, 20082009. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 4 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRN NDVI 1 GSNDVI 2 SPAD 0.90* 3 NDVI1 0.85* 0.92* 4 NDVI2 0.85* 0.92* 0.98* 5 REP 0.66* 0.74* 0.79* 0.69* 6 Area1 0.79* 0.83* 0.96* 0.96* 0.59* 7 Area2 0.79* 0.83* 0.96* 0.96* 0.59* 0.98* 8 NIR/G 0.79* 0.83* 0.98* 0.98* 0.71* 0.98* 0.98* 9 NIR/R 0.85* 0.90* 0.98* 0.98* 0.62* 0.96* 0.96* 0.98* 10 GRN NDVI 0.79* 0.85* 0.98* 0.98* 0.59* 0.98* 0.98* 0.98* 0.96* 60 Table 20b. Correlation matrices and simple correlation coefficients, Feekes growth stage 7, Perkins, OK, 20082009. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 7 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRN NDVI 1 GSNDVI 2 SPAD 0.53* 3 NDVI1 0.86* 0.69* 4 NDVI2 0.86* 0.69* 0.98* 5 REP 0.81* 0.74* 0.92* 0.96* 6 Area1 0.85* 0.69* 0.98* 0.98* 0.96* 7 Area2 0.85* 0.69* 0.98* 0.98* 0.96* 0.98* 8 NIR/G 0.79* 0.69* 0.96* 0.96* 0.94* 0.98* 0.98* 9 NIR/R 0.85* 0.66* 0.96* 0.96* 0.94* 0.98* 0.98* 0.98* 10 GRN NDVI 0.85* 0.71* 0.98* 0.98* 0.92* 0.98* 0.98* 0.98* 0.98* 61 Table 20c. Correlation matrices and simple correlation coefficients, Feekes growth stage 10, Perkins, OK, 20082009. *Significant at the 0.05 probability level. GSNDVI=GreenSeeker NDVI, SPAD=SPAD meter reading, NDVI1= Spectrometer based NDVI where Rnir=750775nm and Rred=740765 nm, NDVI2= Spectrometer based NDVI where Rnir=780805 nm, REP=Rededge position, Area1=Rededge area between 550 nm to 750 nm, Area2=Rededge area between 650 nm to 750 nm. NIR/G=simple ratio where Rnir=780 805nm and Rgreen=540565 nm. NIR/R=simple ratio where Rnir=780805 nm and Rred=740765 nm. GRN NDVI= Spectrometer based NDVI where Rnir=750775nm and Rgreen=540565 nm. Feekes 10 Variables GSNDVI SPAD NDVI1 NDVI2 REP Area1 Area2 NIR/G NIR/R GRN NDVI 1 GSNDVI 2 SPAD 0.55* 3 NDVI1 0.88* 0.74* 4 NDVI2 0.86* 0.76* 0.98* 5 REP 0.61* 0.81* 0.79* 0.81* 6 Area1 0.83* 0.49* 0.88* 0.85* 0.53* 7 Area2 0.83* 0.49* 0.88* 0.85* 0.55* 0.98* 8 NIR/G 0.67* 0.67* 0.86* 0.90* 0.79* 0.71* 0.71* 9 NIR/R 0.72* 0.67* 0.90* 0.92* 0.74* 0.76* 0.76* 0.76* 10 GRN NDVI 0.79* 0.76* 0.96* 0.96* 0.85* 0.79* 0.79* 0.94* 0.94* 62 Figure 3. Plot plan for Experiment 222, Stillwater OK. WHEAT FERTILITY EXPERIMENT NO.222 Agronomy Research Station Established 1969 Plot size: 6.1m x 18 m Alley: 5.2 m Total Trial Area: 41.8 m x 159 m TRT Preplant N rate (kg N /ha) Preplant P rate (kg P2O5 / ha) Preplant K rate (kg K2O / ha) 1.* 0 67 45 2.* 45 67 45 3.* 90 67 45 4.* 135 ^ 67 45 5. 90 0 45 6. 90 34 45 7. 90 101 45 8. 90 67 0 9. 90 67 90 10.* 0 0 0 11. 135 ^ 101 90 12. 135 ^ 101 0 13. 90 67 45 (SulPoMag) N applied as 460 (Urea) P applied as 0460 (Triple Super Phosphate) K applied as 0060 (Potash) *  YP plot ^  Split 135 kg N rates to 67 kg N (fall) and 60 kg N (spring) N E S W Location: Stillwater 1, 2 – Harvest Sequence Number 1, 2 – Treatment Number 1 , 2 – Soil Sample Sequence Number OBJECTIVE: To study fertilizer nitrogen, phosphorus, and potassium in winter wheat. In recent years, this study has also been used to develop yield potential models and yield predictions through sensor based technologies. 52 49 48 45 44 41 40 37 36 33 32 29 28 25 24 21 20 17 16 13 12 9 8 5 4 1 13 13 8 5 7 3 2 11 12 6 9 1 10 4 6 2 11 5 3 8 12 10 9 1 7 4 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 Rep 2 Rep 1 Rep 2 Rep 1 51 50 47 46 43 42 39 38 35 34 31 30 27 26 23 22 19 18 15 14 11 10 7 6 3 2 13 13 8 10 7 12 9 5 2 11 1 3 4 6 3 11 1 8 6 9 12 2 5 10 4 7 52 51 50 49 48 47 46 45 44 43 42 41 40 39 38 37 36 35 34 33 32 31 30 29 28 27 Rep 4 Rep 3 Rep 4 Rep 3 63 Figure 4. Plot plan for N & P Study, Perkins, OK. PERKINS N & P STUDY SOUTHWEST AGRONOMY RESEARCH STATION ESTABLISHED 1996 Treatment Preplant N (kg N /ha) Preplant P (kg P2O5 / ha) 1 0 0 2 0 34 3 0 67 4 56 0 5 56 34 6 56 67 7 112 0 8 112 34 9 112 67 10 168 0 11 168 34 12 168 67 N E S W Location: Perkins Plot Size: 3.0 m x 9 m Alleys: 3.0 m 3 4 9 10 15 16 21 22 27 28 33 34 Rep 3 9 5 12 1 8 11 7 4 3 6 2 10 25 26 27 28 29 30 31 32 33 34 35 36 2 5 8 11 14 17 20 23 26 29 32 35 Rep 2 6 11 10 5 1 8 12 9 4 7 2 3 13 14 15 16 17 18 19 20 21 22 23 24 1 6 7 12 13 18 19 24 25 30 31 36 Rep 1 7 10 9 12 1 4 3 6 2 8 11 5 1 2 3 4 5 6 7 8 9 10 11 12 Total Trial Area: 33.5 m x 36.6 m OBJECTIVE: To evaluate nitrogen and phosphorus interactions in winter wheat. 1, 2 – Harvest Sequence Number 1, 2 – Treatment Number 1 , 2 – Soil Sample Sequence Number 64 Figure 5. GreenSeeker NDVI plotted against N rates, Stillwater, OK, 20072008. . 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 45 90 135 NDVI N rate (kg/ha) 20072008 GreenSeeker NDVI Feekes 4 Feekes 5 Feekes 7 Feekes 10 65 Figure 6. SPAD reading plotted against N rates, Stillwater, OK, 20072008. 20 25 30 35 40 45 50 0 45 90 135 SPAD Value N rate (kg/ha) 20072008 SPAD Meter Feeks 4 Feekes 5 Feekes 7 Feekes 10 66 Figure 7. Rededge position (REP) plotted against N rates, Stillwater, OK, 20072008. 711 713 715 717 719 721 0 45 90 135 Wavelength (nm) N rate (kg/ha) 20072008 Rededge position Feekes 4 Feekes 5 Feekes 7 67 Figure 8. AREA2 plotted against N rates, Stillwater, OK, 20072008. 0 50 100 150 200 250 0 45 90 135 Area N rate (kg/ha) 20072008 AREA2 Feekes 4 Feekes 5 Feekes 7 68 Figure 9. REP plotted against N rates, Stillwater, OK, 20082009. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 45 90 135 NDVI N rate (kg/ha) 20082009 GreenSeeker NDVI Feekes 4 Feekes 5 Feekes 7 Feekes10 69 Figure 10. SPAD reading plotted against N rates, Stillwater, 20082009. 20 25 30 35 40 45 50 0 45 90 135 SPAD Value N rate (kg/ha) 20072009 SPAD Meter Feeks 4 Feekes 5 Feekes 7 Feekes 10 70 Figure 11. REP against N rates, Stillwater, OK, 20082009. 711 713 715 717 719 721 0 45 90 135 Wavelength (nm) N rate (kg/ha) 20082009 Rededge position Feekes 4 Feekes 5 Feekes 7 Feekes10 71 Figure 12. AREA2 plotted against N rates, Stillwater, OK, 20082009. 0 50 100 150 200 250 0 45 90 135 Area N rate (kg/ha) 20082009 AREA2 Feekes 4 Feekes 5 Feekes 7 Feekes10 72 Figure 13. Shape of rededge and its position at Feekes 4 in the 0 kg N/ha and 135 kg N/ha plots, Stillwater, OK, 2008. 0.005 1E17 0.005 0.01 0.015 0.02 650 670 690 710 730 750 The first derivative reflectance Wavelength (nm) Rededge position and Shape 0 kg N/ha 135 kg N/ha 73 Figure 14. The first derivative reflectance in the 0 kg N /ha and 135 kg N/ha plots, Stillwater, OK, 2008. 0.005 1E17 0.005 0.01 0.015 0.02 650 670 690 710 730 750 The first derivative reflectance Wavelength (nm) Original data 0 kg N/ha 135 kg N/ha 74 Figure 15. Filtering the first derivative reflectance by averaging over 10 nm widths in the 0 kg N /ha and 135 kg N/ha plots, Stillwater, OK, 2008. 0.005 1E17 0.005 0.01 0.015 0.02 650 670 690 710 730 750 The first derivative reflectance Wavelength (nm) 10 nm filtering 0 kg N/ha 135 kg N/ha 75 Figure 16. Filtering the first derivative reflectance by weight averaging over 10 nm widths in the 0 kg N /ha and 135 kg N/ha plots, Stillwater, OK, 2008. 0.005 1E17 0.005 0.01 0.015 0.02 650 670 690 710 730 750 The first dericative reflectance Wavelength (nm) Weight filtering after 10 nm filtering 0 kg N/ha 135 kg N/ha 76 Figure 17. Filtering the first derivative reflectance by using trend fitting in the 0 kg N /ha and 135 kg N/ha plots, Stillwater, 2008. 0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 640 660 680 700 720 740 760 The first derivative reflectance Wavelength (nm) Filtering by trend fitting 0 kg N/ha 135 kg N/ha Figure 18. Visual image of winter wheat at Feekes 4 plots under conventional tillage 77 4, in the 0 kg N /ha and 135 kg N/ha tillage, Stillwater, OK, 2008. 135 kg N/ha 0 kg N/ha Figure 19. Visual image of winter wheat at Feekes 4 plots under Notill system 78 in the 0 kg N /ha and 112 kg N/ha system, Stillwater, OK, 2008. 112 kg N/ha 0 kg N/ha Figure 20. Linear relationship between different N rates in Stillwater 0 kg N/ha (r² = 0.17) y = 4.5802x + 711.79 45 kg N/ha (r² = 0.80) y = 15.746x + 705.57 712 713 714 715 716 717 718 719 720 0 Rededge postion 0 kg N/ha (r² = 0.56 y = 14.772x + 709.35 45 kg N/ha (r² = 0.29) y = 16.682x + 707.62 712 713 714 715 716 717 718 719 720 0 Rededge postion 79 REP and GreenSeeker NDVI at Feekes 4 Stillwater, OK, 20072009. 90 kg N/ha (r² = 0.73) y = 23.172x + 700.44 135 kg N/ha (r² = 0.87) y = 27.927x + 697.05 0.2 0.4 0.6 0.8 GSNDVI 20072008 Feekes 4 0.56） 90 kg N/ha (r² = 0.67) y = 6.1558x + 714 135 kg N/ha (r² = 0.88) y = 24.344x + 702.34 0.2 0.4 0.6 0.8 GSNDVI 20082009 Feekes 4 at 1 check 0 kg N/ha 45 kg N/ha 90 kg N/ha 135 kg N/ha 1 check 0 kg N/ha 45 kg N/ha 90 kg N/ha 135 kg N/ha 80 Figure 21. Linear relationship between GSNDVI and REP under the conventional tillage or notill system at all growth stages, Stillwater and Perkins, OK, 20082009. Conventional (r² = 0.66) y = 13.895x + 708.5 Notill (r² = 0.42) y = 14.612x + 710.91 710 712 714 716 718 720 722 724 726 728 0 0.2 0.4 0.6 0.8 1 Rededge position GSNDVI 20072008 Conventional Tillage Notill System Conventional (r² = 0.70) y = 9.7454x + 711.88 Notill (r² = 0.34) y = 12.423x + 711.6 710 712 714 716 718 720 722 724 726 728 0 0.2 0.4 0.6 0.8 1 Rededge position GSNDVI 20082009 Conventional Tillage Notill System 81 Figure 22. Linear relationship between REP and GreenSeeker NDVI at Feekes 4 and Feekes 5, Stillwater, OK, 20082009. Feekes 5 (r² = 0.76) y = 10.463x + 711.12 Feekes 4 (r² = 0.81) y = 7.9209x + 712.48 712 713 714 715 716 717 718 719 720 721 722 723 0 0.2 0.4 0.6 0.8 1 Rededge position GSNDVI 20082009 Feekes 4 Feekes 5 Feekes 4 (r² = 0.80) y = 15.567x + 706.26 Feekes 5 (r² = 0.77) y = 13.801x + 709.19 712 713 714 715 716 717 718 719 720 721 722 723 0 0.2 0.4 0.6 0.8 1 Rededge position GSNDVI 20072008 Feekes 4 Feekes 5 82 Feekes 4 (r² = 0.45) y = 27.031x + 30.224 Feekes 5 (r² = 0.48) y = 15.317x + 30.41 20 25 30 35 40 45 50 55 60 0 0.2 0.4 0.6 0.8 1 SPAD GSNDVI 20082009 Feekes 4 Feekes 5 Figure 23. Linear relationship between SPAD and GreenSeeker NDVI at Feekes 4 and Feekes 5, Stillwater, OK, 20082009. Feekes 4 (r² = 0.67) y = 58.428x + 0.6167 Feekes 5 (r² = 0.76) y = 25.966x + 26.023 20 25 30 35 40 45 50 55 60 0 0.2 0.4 0.6 0.8 1 SPAD GSNDVI 20072008 Feekes 4 Feekes 5 83 Figure 24. Linear relationship between SPAD and REP at Feekes 4 and Feekes 5 in Stillwater OK, 20082009. Feekes 4 (r² = 0.74） y = 3.4823x  2455.8 Feekes 5 (r² = 0.83) y = 1.7273x  1197.7 20 25 30 35 40 45 50 55 60 712 714 716 718 720 722 724 SPAD Rededge position 20072008 Feekes 4 Feekes 5 Feekes 4 (r² = 0.33) y = 1.9405x  1346.5 Feekes 5 (r² = 0.53） y = 1.8523x  1288.9 20 25 30 35 40 45 50 55 60 712 714 716 718 720 722 SPAD Rededge position 20082009 Feekes 4 Feekes 5 VITA Yumiko Kanke Candidate for the Degree of Plant and Soil Science Master of Science/Arts Thesis: RED EDGE AS A POTENTIAL INDEX FOR DETECTING DIFFERENCES IN PLANT NITROGEN STATUS IN WINTER WHEAT Major Field: Agronomy Biographical: Education: Received Bachelor of Science degree in Plant and Soil Science from the Oklahoma State University, Stillwater, Oklahoma in May 2008. Completed the requirements for the Master of Science degree from Soil Science at Oklahoma State University, Stillwater, Oklahoma in December 2009. Experience: Employed by Wheat Breeding program in Oklahoma State University as a field’s assistant May 2006 to August 2006; employed by Oklahoma State University Soil Fertility testing lab as a lab assistant August 2006 to May 2008; employed by Oklahoma State University, department of plant and soil science as a graduate research assistant, May to present. Professional Memberships: American Society of Agronomy, Soil Society of America, Phi Kappa Phi, Phi Beta Delta Honor Society for International Scholars. ADVISER’S APPROVAL: Dr. William R. Raun Name: Yumiko Kanke Date of Degree: December, 2009 Institution: Oklahoma State University Location: Stillwater, Oklahoma Title of Study: RED EDGE AS A POTENTIAL INDEX FOR DETECTING DIFFERENCES IN PLANT NITROGEN STATUS IN WINTER WHEAT Pages in Study: 83 Candidate for the Degree of Master of Science Major Field: Agronomy Scope and Method of Study: To determine whether the rededge position has potential to be better indices and ultimately more accurate fertilizer N recommendations for winter wheat compared to NDVI, especially at Feekes 4 to 5 which is crucial time to make decision for N rate. Findings and Conclusions: NDVI was sensitive to plant N response as well as different plant growth stages but the sensitivity tended to decrease as N rate increased. REP sensitivity to plant N response increased as N rate increased and with advancing plant growth. NDVI and REP were linearly correlated at all growth stages (r2=0.79). REP and SPAD meter readings were also highly correlated (r2= 0.62) as were NDVI and SPAD (r2=0.56). REP sensitivity was expected to be higher than NDVI sensitivity at high plant biomass or N concentration, but this was not obvious for winter wheat. Concerning the technical problems of measuring REP in the field, further engineering will be required to evaluate REP as an alternative N index in agronomic fields. 



A 

B 

C 

D 

E 

F 

I 

J 

K 

L 

O 

P 

R 

S 

T 

U 

V 

W 


