Proper nutrient management is essential to increase yield, quality and profi t. This study was conducted to estimate the N concentrations of chinese cabbage (Brassica campestris L. ssp. pekinensis ‘Norangbom’) plug seedlings using visible and near infrared spectroscopy for nondestructive N detection. Chinese cabbage seeds were sown and raised in three 200-cell plug trays fi lled with growing mixture in a plant growth chamber with three different levels (40%, 80%, and 100%) of required N. Refl ectance for leaves of chinese cabbage seedlings was measured with a spectrophotometer 15 days after the experiment started. Refl ectance was measured in the 400 to 2500 nm wavelength range at 1.1-nm increments. The leaves were dried afterwards to measure their water content and were analyzed for their actual N contents. The experiment was repeated twice (group I and II). Correlation coeffi cient spectrum, standard deviation spectrum, stepwise multiple linear regression (SMLR), and partial least squares (PLS) regression were used to determine wavelengths for N prediction models. Performances of SMLR and PLS were similar. For the validation data set (group II), SMLR produced an r of 0.846 and PLS yielded r of 0.840. The most signifi cant wavelength 710 nm, which was identifi ed by all methods, was correlated to chlorophyll. Water content positively correlated with N concentration (r = 0.76). Wavelengths of 1467, 1910, and 1938 nm selected by SMLR from both groups also showed that water had a strong effect on N prediction. Wavelengths near 2136 nm indicated that protein had potential use in N prediction. Wavelengths near 550 and 840 nm could also contribute to N prediction. Chinese cabbage is the most important leafy vegetable in Korea. An estimated 1,242 million plants of its plug seedlings are transplanted yearly (Kim and Lee, 2000). Chinese cabbage benefi ts from fertilizers with high N content. Inappropriate nutrient management may lead to undesirable effects on total or marketable yield, environmental contamination and profi t. Application of too little N causes reduced yields, shortens storage life, and delays maturity, while excess N may cause rapid growth leading to coarse, loose heads, cracking, tipburn, poor processing, and low storage quality (Peet, 2004). This situation creates a need for monitoring the N status of crops. Visible and near infrared spectroscopy (VIS–NIR) have been largely used to detect nutrient status for crops, due to their suitability for rapid and nondestructive determination of nutrient concentrations (Miller and Thomas, 2003). Two main N sources generally exist in green leaves, i.e., chlorophyll and proteins. Chlorophyll contains 5% to 10% N. Chlorophyll exhibits strong absorption in the visible region arising from conjugated carbon–carbon single and double bonds of the porphyrin ring and the magnesium (Mg) ion. The infrared spectra of chlorophyll show strong absorption due to C–H bonds in the phytol tail of the molecule (Katz et al., 1966). Chlorophyll absorbs light at wavelengths of 430, 450, 650, and 660 nm and refl ects light at 550 nm which makes leaves look green (Farabee, 2001). Proteins, which are the primary nitrogenous constitute in leaves, typically hold 70% to 80% of total N. Spectral bands for N related to protein absorptions at 2054 and 2172 nm are due to N in the molecular structure, in particular to C–N and N–H bonds (Kokaly, 2001). Water, taking up to 90% of the total weight, is the major component of chinese cabbage. It is a good absorber of middle-infrared energy, so the greater the water inside leaves, the lower the middle-infrared refl ectance. Experiments by Curcio and Petty (1951) showed that fi ve prominent absorption bands for pure water were at 760, 970, 1190, 1450, and 1940 nm. When the water content of the plant decreases to 50%, the refl ectance at any portion of the visible, nearand middle-infrared regions will largely increase. Many studies have shown a strong relationship between N concentration and leaf refl ectance spectra. Thomas and Oerther (1972) found a non-linear relationship between refl ectance at 550 nm and leaf N content of sweet pepper leaves with a correlation coeffi cient of –0.93. Card et al. (1988) found that N in dried and ground foliage leaves could be determined accurately from refl ectance with a laboratory spectrometer (r = 0.90). Katayama et al. (1996) found that the correlation coeffi cient (r) between absorption spectra and starch of sweet potato was 0.949. Lee et al. (1999) found that SPAD (Soil and Plant Analyzer Development, Minolta Inc.) readings, which were based on 659 and 940 nm by transmittance, were well correlated with N content in corn ear leaves (r = 0.96). Carter and Knapp (2001) reported that wavelengths near 700 nm were crucial for estimating leaf chlorophyll stress. Bell et al. (2004) found that VIS–NIR was effective for turfgrass N estimation (r = 0.76). Some wavelengths of 448, 669, 719, 1377, 1773, and 2231 nm were identifi ed by Min and Lee (2005) for citrus leaves as signifi cant wavelengths for N detection. They also reported that VIS–NIR has potential as a rapid method for citrus leaf N prediction (r = 0.85). The success of calibration model development and wavelength selection largely relies on statistical solutions. The partial least squares (PLS) procedure, used as a powerful tool in chemometrics, works by extracting successive linear combinations of the predictors, which optimally explain response variation and predictor variation. PLS has been described as a two-step method where the fi rst step reduces data matrix dimensions and the second step identifi es latent structure models in the data matrix (Helland, 2001; Lingjaerde and Christophersen, 2000). In contrast to principal component regression (PCR), which chooses factors that explain the maximum variance in predictor variables without considering the response variables, the PLS method balances the two objectives, seeking the factors that explain both response and predictor variations (SAS, 1990). The predicted residual sum of squares (PRESS) statistic in PLS measures how well the regression equation fi ts the data set. An optimal number of factors is generally obtained when PRESS is minimized (Sundberg, 1999), and a smaller PRESS value indicates a better model prediction. However, selecting the number of factors where the absolute minimum PRESS exists may not be the best choice. By using the CVTEST cross-validation option in SAS PLS, a statistical comparison can be performed to test the signifi cance of differences in the PRESS value for each number of factors, thus determining how many factors should be selected for a calibration model. The overall objective of this research was to explore the feasibility of using near-infrared FebruaryBook 1 162 12/14/05 10:55:35 AM 163 HORTSCIENCE VOL. 41(1) FEBRUARY 2006 spectroscopy for nondestructive N detection of chinese cabbage seedlings. More specifi cally, the objectives were to investigate characteristics of refl ectance spectra for the leaves of chinese cabbage seedlings, to fi nd the optimal number of factors (or wavelengths) that could best describe chinese cabbage leaf properties, and to develop a calibration model for predicting N concentrations of unknown chinese cabbage samples using diffuse refl ectance spectroscopy in the visible (VIS) and near infrared (NIR) regions for better N management. Materials and Methods Chinese cabbage (Brassica campestris L. ssp. pekinensis ‘Norangbom’) seeds were sown and raised in 200-cell plug trays (Bumnong Co., Ltd., Korea) fi lled with growing mixture (BM2, Berger Peat Moss, Canada) in a plant growth chamber. Two metal halide lamps (MT400DL/BH; EYE Lighting International of North America, Inc.) were used to provide illumination with photoperiod of 12 h·d after seeds were germinated. Photosynthetic photon fl ux on the plug trays was 250 ± 12 μmol·m·s. Air temperature was controlled to 25/13 °C for light/dark periods. Nutrient solutions with three different levels (Normal, N80, and N40) of N requirements were prepared (Table 1). The normal nutrient solution for optimal chinese cabbage growth suggested by the National Horticultural Research Institute in Korea, which is composed of 8.0N–2.4P–2.4K–4.8Ca–1.6Mg (mg·L), was formulated by Lee et al. (2000). Normal, N80 and N40 treatments were the nutrient solutions containing 100%, 80%, and 40% of the N requirements, which was in both ammonium and nitrate forms. N80 and N40 treatments were used to create N defi ciency conditions for this experiment. A conductivity meter (model YSI31; YSI Inc., Yellow Springs, Ohio) was used to measure electric conductivity of the nutrient solution. Electric conductivities in the treatments were shown as 1.1 mS·cm for Normal, 1.2 mS·cm for N80, and 1.0 mS·cm for N40. The pH of the nutrient solutions measured by a pH meter (model SA720; ORION Research Inc.) showed pH 6.2 for normal, pH 6.2 for N80, and pH 6.3 for N40. The nutrient solution was supplied once every 2 d during 10 d after germination and then once everyday by sub-irrigation. Typically it would take about 15 d for the chinese cabbage seedlings to develop three to four true leaves after germination in a growth chamber, which would be considered to be optimal for transplanting to a fi eld in Korea. Therefore, the third true leaf showing the biggest leaf in a plant was sampled and prepared for refl ectance measurement 15 d after germination started. The experiment to estimate the N concentration of leaves for chinese cabbage plug seedlings was replicated once in this study. Forty seedlings were randomly selected from each N treatment for refl ectance measurement, which yielded 120 samples (= 3 N treatments × 40 samples per N treatment) for each experiment. The data set from the fi rst experiment was labeled as group I, and the data set from the replicated experiment was labeled as group II. After refl ectance measurements, water conte
[1]
Won Suk Lee,et al.
ASSESSING NITROGEN STRESS IN CORN VARIETIES OF VARYING COLOR
,
1999
.
[2]
W. S. Lee,et al.
DETERMINATION OF SIGNIFICANT WAVELENGTHS AND PREDICTION OF NITROGEN CONTENT FOR CITRUS
,
2005
.
[3]
J. Curcio,et al.
Near infrared absorption spectrum of liquid water
,
1951
.
[4]
John B. Solie,et al.
Optical sensing of turfgrass chlorophyll content and tissue nitrogen
,
2004
.
[5]
Raymond F. Kokaly,et al.
Investigating a Physical Basis for Spectroscopic Estimates of Leaf Nitrogen Concentration
,
2001
.
[6]
Seiji Tamiya,et al.
Prediction of starch, moisture, and sugar in sweetpotato by near infrared transmittance
,
1996
.
[7]
G. Carter,et al.
Leaf optical properties in higher plants: linking spectral characteristics to stress and chlorophyll concentration.
,
2001,
American journal of botany.
[8]
I. Helland.
Some theoretical aspects of partial least squares regression
,
2001
.
[9]
G. Miller,et al.
Using Near Infrared Reflectance Spectroscopy to Evaluate Phosphorus, Potassium, Calcium, and Magnesium Concentrations in Bermudagrass
,
2003
.