Use of nondestructive spectroscopy to assess chlorophyll and nitrogen in fresh leaves
CHAPTER 3: The accuracy of nondestructive optical methods for chlorophyll (Chl) assessment based on leaf spectral characteristics depends on the wavelengths used for Chl assessment. To determine the optimum wavelengths for Chl assessment (OWchl) using reflectance spectroscopy, almond (Prunus dulcis (Mill.) D.A. Webb `Nonpareil'), poplar (UCC-1, a hybrid of Populus trichocarpa x P. deltoides, Union Camp, Princeton, NJ) and apple (Malus domestica Borkh `Fuji') trees were grown at different rates of nitrogen (N) fertilization to produce leaves with different Chl concentrations. Spectral reflectance of leaf discs was measured using a spectroradiometer (300 nm to 1100 nm at 1 nm intervals), and total CM concentration in leaf discs was determined. The OWchl for nondestructive Chl assessment by reflectance spectroscopy was identified using three methods (1) the coefficient of determination (R2) from simple linear regression, (2) reflectance sensitivity analysis, and (3) the 1st spectral derivative method. Our results indicated that the 1st derivative method can be used to estimate OWchl in the red edge region and reflectance sensitivity analysis can be used to estimate the OWchl in both the red edge and visible regions. Reflectance sensitivity analysis was more accurate when used for OWchi selection than the r t derivative method because methods used for reflectance sensitivity analysis ensures that differences in the spectral reflectance are caused by differences in Chl concentration, while the 1st derivative method could not distinguish whether spectral differences were the result of differences in Chl concentration or caused by other factors. However, neither the 1" derivative method nor reflectance sensitivity analysis alone could accurately identify the actual OWcnl. R2 was a useful indicator for verifying the accuracy of OWcnl selection. Higher R2 values were usually associated with lower root mean square errors (RMSE) and higher reflectance sensitivity; therefore the wavelengths with the highest R2 and highest reflectance sensitivity were selected as the OWchi. Our results indicate that using simple linear R2 in combination with reflectance sensitivity analysis is the more reliable method for determining OWcnl in plant leaves. CHAPTER 4: One-year-old poplar (UCC-1, a hybrid of Populus trichocarpa x P. deltoides, Union Camp, Princeton, NJ) trees were grown with one of six different rates of nitrogen (N) fertilizer to produce leaves with different chlorophyll (Chl) concentrations (160 to 659µmol.m-2) for identifying optimal wavelengths (OWcnl) and evaluating indices for non-destructive CM assessment using reflectance. Leaves with different Chl concentrations were scanned from 300 to 1100 nm, at 1 nm intervals, using a spectroradiometer to measure spectral reflectance. Concentrations of leaf Chl a, Chl b and Chl a+b were used to determine OWcM usiing wavelength sensitivity analysis and simple linear regressions between reflectance values and Chl concentrations. The coefficient of determination (R2) and root mean square error (RMSE) from the regression analysis and results of the wavelength sensitivity analysis indicated that wavebands from two regions of the spectrum, red edge (700 -730 nm) and visible (540 – 615 m), had larger R2, higher wavelength sensitivity, and smaller RMSE than peaks at other wavebands. These wavebands were determined to be OWcnl for Chl (Chl a, Chl b and Chl a+b) assessment in poplar leaves. The accuracy (R2 and the corresponding RMSE) of different published indices compared in our study differed greatly and was lower than that of the indices developed with the OWcnl we determined using a combination of regression and wavelength sensitivity. The main reason that the published indices were less accurate (smaller R2 with larger RMSE) was because the Chl related wavelengths used in these indices were not close to the OWCI. Our results indicate that identification of OWchi is very important for indices development. When the Chl-related wavelengths (675, 695, 698, 700, 705 or 710 nm) in different proposed indices were replaced by the OWchi we determined in the visible (580, 563 and 574nm) or in red edge (715, 730 and 720nm) for Chl a, Chl b or Chl a+b, all results using modified indices were more accurate than the results using the original indices. The most accurate indices for Chl assessment of poplar leaves were determined to be R750-i000/R0w, (R750-i000-Row)/(R750-i000+Row), 8430-490/Row and (8430-490-Row)/(8430-490+Row), for Chl a, Chl b or Chl a+b, respectively. Although OWCM differ among species, cultivars and Chl (Chl a, Chl b and Chl a+b), our results indicated that wavelength from 700 - 730nm and 540 - 580nm regions of the spectrum can be used as the OWchi for a variety of species. Indices that use a combination of the OWchi with a reference wavelength (RW) at either 430 - 490nm or 750 - 1100nm can be used to eliminate the effect of leaf texture on Chl assessment and more widely applicable across species. These indices can be simplified as RRW/RoW and (RRW –Row)/( RRW+ROW). CHAPTER 5: One-year-old almond (Prunus dulcis (Mill.) D.A. Webb `Nonpareil'), poplar (UCC-1, a hybrid of Populus trichocarpa x P. deltoides, Union Camp, Princeton, NJ) and apple (Malus domestica Borkh `Fuji') trees, 15-year-old purple leaf plum (Prunus cerasifera 'Newport'), 10-year-old purple leaf flowering cherry (Prunus blireiana) grown in the campus landscape of Oregon State University and 8-years-old three apple cultivars (Malus domestica Borkh `Jonagold',`Gala' and `Cameo') grown in orchard conditions were used to assess how (1) optimum wavelengths for chlorophyll assessment (OWchi) are influenced by plant genotype and (2) leaf properties (leaf texture, pigmentation, and water status) influence chlorophyll (Chl) assessment by altering reflectance characteristics of leaves. We determined that the accuracy of using optimum wavelengths for Chl assessment (OWE) in both visible (540 - 580 nm) and red edge (700- 730 nm) regions varied among species and with CM type (e.g. CM a, CM b, total CM). Differences in OWM among species caused by the variation in leaf optical properties impaired the accuracy of indices used for assessing Chl. However, the OWM in the red edge region could be used accurately to measure Chl across all species tested; whereas the OWCM in the visible region could be only used across a wide range of anthocyanins-free species. We determined that variation in reflectance of visible and red edge wavelengths caused by variation of leaf texture or other optical properties could be eliminated by referencing the Chl-sensitive OWcI to a NIR wavelength (750 – 1100 nm) that was sensitive to leaf texture but insensitive to Chl by using the following algorithms: R540-580/R750-1100, R700-750/R750-1100, (R54o-58o-R750-11oo)/(R54o-580+R750-1100), or (R700-740-R750-11oo)/(R700-740+R750 1100). We also determined that the effect of dehydration on spectral reflectance can possibly be eliminated by referencing a Chl-sensitive OWcr,l to a water sensitive wavelength between 1420 nm and 1510 nm. The effect of anthocyanins on the accuracy of Chl assessment differed among species and indices. Indices developed using OWCM from the red edge were more accurate for assessing Chl and were less affected by the existence of anthocyanins, variation of leaf water status and other interferences than indices developed using OWMM from the visible region. Carotenoid concentrations were positively correlated to Chl concentrations and did not influence the accuracy of the indices developed for Chl assessment using OWcl from either the visible or red edge regions. CHAPTER 6: One-year-old almond (Prunus dulcis (Mill.) D.A. Webb `Nonpareil'), poplar (UCC-1, a hybrid of Populus trichocarpa x P. deltoides, Union Camp, Princeton, NJ) and apple (Malus domestica Borkh `Fuji') trees grown with one of six different rates of nitrogen (N) fertilizer were used to assess the influence of meter parameters and sampling technique on the precision and accuracy of commercial hand-held meters (SPAD-502, CCM-200, and CM-1000) used for estimating chlorophyll (CM) and N in fresh leaves. Concentrations of CM and N in leaves were determined and spectroradiometry was used to compare optimum wavelengths (OW) and algorithms for CM and N assessment to those used by hand-held meters. Our results showed that SPAD-502 was more precise and more accurate than CCM-200 and CM-1000 for assessing CM and N in fresh leaves. The precision and accuracy of CM-1000 output was similar to that of CCM-200 when CM-1000 measurements were taken at a constant sampling distance. However, if sampling distance was not constant, the output precision of CM-1000 significantly decreased. The CM-sensitive wavelength used by CM-1000 (700 nm) was more accurate at estimating CM and N concentrations than the wavelengths used by SPAD-502 (650 nm) and CCM-200 (660nm); however the variation in sampling distance, orientation, light intensity, and the inconsistency of light intensity between ambient light sensor and the target leaf made the CM-1000 less accurate than the other two meters. Using spectroradiometery on leaves from the same plants, we found the most accurate wavelengths for CM and N assessment occurred within two regions of the visible spectrum: visible (540-580 nm) and red edge (700-730 nm). However, N assessment using these wavelengths was less accurate than CM assessment. Using transmission or reflectance with one OW in either the visible or red edge region for CM and N assessment was more accurate than or similar to using any of the wavelengths employed by the hand-held meters we tested. When an OW was used with a NIR wavelength (750-1000 nm) in the form of simple ratio or any other algorithm to compensate for leaf texture, we found accuracy of CM and N assessment was greater than the accuracy achieved using only one wavelength. Use of a NIR wavelength was especially important if the accuracy of the estimate obtained using either the CM- or N-sensitive wavelength alone was small (e.g. R2<0.8000 for Chl or R2<0.6000 for N). We also found that OW in both visible and red edge regions varied among species, but this variation did not influence the accuracy of CM and N assessment. Three meter prototypes were developed based on the OW and indices we identified and the factors influencing commercial accuracy we found. Our results indicate that the Prototype-III, using CM-sensitive OW with a CM-insensitive NIR wavelength in combination with the meter design of constant light source and precision sampling distance, was more accurate than all commercial hand-held meters for Chl assessment and than the CM-1000 for N assessment across all species we tested.