Published online 29 June 2007
Published in Soil Sci Soc Am J 71:1398-1405 (2007)
DOI: 10.2136/sssaj2006.330
© 2007 Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
NUTRIENT MANAGEMENT & SOIL & PLANT ANALYSIS
Application of Reflectance Near Infrared Spectroscopy for Animal Slurry Analyses
L. K. Sørensena,*,
P. Sørensenb and
T. S. Birkmosec
a Eurofins–Steins Laboratorium, Ladelundvej 85, 6600 Vejen, Denmark
b Univ. of Aarhus, Faculty of Agricultural Sciences, Dep. of Agroecology, P.O. Box 50, 8830 Tjele, Denmark
c Danish Agricultural Advisory Service, Udkærsvej 15, 8200 Aarhus N, Denmark
* Corresponding author (lks{at}steins.dk).
 |
ABSTRACT
|
|---|
The feasibility of using near infrared spectroscopy (NIRS) for rapid prediction of the composition of cattle and pig slurries was investigated. Samples with a total solids content from <1 to 15% were collected during a 3-yr period and used for calibration and validation. Test samples were sealed in plastic bags and measured in a sample cell moved vertically during measurements. Reflectance data in the range 1200 to 2400 nm were used for calibration based on partial least square regression. Dry matter (DM), N, NH4–N, and P could be determined with r2 values of 0.97, 0.94, 0.92, and 0.87, respectively. The ratios between analyte variation range standard deviation and the root mean square error of prediction (RMSEP) obtained on a calibration independent test set were 6.2 (DM), 4.3 (N), 3.8 (NH4–N), and 3.6 (P). Total C and plant-available N (PAN) could be determined by near infrared spectroscopy with r2 values of 0.94 and 0.89, but the same correlation was obtained by calculation from DM and NH4–N results, respectively. The applicability of NIRS for K, Mg, Ca, Na, S, Cu, and Zn analysis was also investigated. The r2 values were in the range 0.41 to 0.82, with the poorest results for Na, Zn, and K. The corresponding SD/RMSEP ratios were in the range 1.2 to 3.5. We concluded that the applied NIRS methodology is suitable for rapid routine analysis of DM, C, N, NH4–N, P, and PAN in both cattle and pig slurries.
Abbreviations: DM, dry matter • NIR, near infrared • NIRS, near infrared spectroscopy • PAN, plant-available nitrogen • PLS, partial least squares • RMSECV, root mean square error of cross-validation • RMSEP, root mean square error of prediction • SEP, standard error of prediction
|
INTRODUCTION
|
|---|
For optimal use of manure as fertilizer, solely or in combination with mineral fertilizers, the concentration of nutrients must be known to reduce the risk of over- or underfertilization of the crop. In the case of overfertilization, the non-optimal use of resources may result in pollution of surface waters with N and P compounds and it may induce an increased level of NO3 in groundwaters. Since the content of nutrients in animal manure varies, optimal use of the fertilizers requires analysis of each batch before field application. Analysis of manure by classical chemical methods is time consuming and expensive. In addition, it may be very difficult to obtain a representative laboratory sample from a large batch. Repeated sampling and determination of the nutrient composition while emptying the storage tank would, in many cases, be preferable although it would require very short analysis time. The ultimate target may be on-line measurements, but for common use that requires development of simple, robust, and relatively inexpensive techniques.
Near infrared spectroscopy has within the past two decades become a widespread technique for rapid analysis of forage, grains, and food products. The technique has also recently been applied for the determination of DM, NH4–N, and total N in manures from cattle and pigs (Reeves and Van Kessel, 2000a, 2000b; Millmier et al., 2000; Kemsley et al., 2001; Malley et al., 2002; Saeys et al., 2004, 2005a, 2005b; Dolud et al., 2005; Ye et al., 2005; Malley et al., 2005) and poultry (Kemsley et al., 2001; Reeves, 2001a, 2001b; Tasistro et al., 2003; Aiken et al., 2005; Ye et al., 2005). In addition, the possibility of predicting P and K in manures has been investigated by some of the researchers. The variability in results obtained may be ascribed inter alia to the nature of the sample matrices investigated, the heterogeneity of the samples, the concentration ranges of analytes, the ranges of DM, the near infrared (NIR) subregion scanned, the measuring mode (transmission, transflectance, reflectance), and the sample presentation techniques used. The sample presentation techniques reported for direct measurement on manures includes the use of fiber-optic probes on stirred samples (Reeves and Van Kessel, 2000a), bag sampling techniques (Millmier et al., 2000; Reeves and Van Kessel, 2000b; Reeves, 2001b; Ye et al., 2005), liquid sample cells for reflectance or transflectance measurements (Malley et al., 2002, 2005; Tasistro et al., 2003; Saeys et al., 2004; Aiken et al., 2005), remote measuring heads connected to instruments by fiber optics (Saeys et al., 2005a, 2005b), and flow cells (Dolud et al., 2005).
The objective of the present study was to investigate the possibility of developing routine laboratory NIR calibrations suitable for rapid determination of organic nutrients and minerals in both cattle and pig slurries containing 0.5 to 15% DM.
|
MATERIALS AND METHODS
|
|---|
Sampling
Samples of cattle and pig slurries collected from different farms in Denmark were used for calibration and validation. The calibration sample set consisted of 255 samples from cattle and pigs with an equal distribution between animal origins. They were selected by a stratified sampling procedure from a large population of samples from 2004 and 2005 to obtain a relatively even distribution with respect to DM. The validation sample set consisted of 38 cattle and 49 pig samples from a different population. They were selected randomly from incoming laboratory samples in the spring of 2006. The variation ranges of the sample sets are listed in Tables 1 and 2. Samples were stored at –20 ± 2°C until NIR measurement and chemical analysis were initiated.
View this table:
[in this window]
[in a new window]
|
Table 1. Composition of calibration samples (sampled in 2004 and 2005, n = 255 [128 for plant-available N]) as determined by reference methods.
|
|
Sample Preparation
The samples (generally 0.5–1 L) were equilibrated to room temperature (20–24°C) and mixed thoroughly using an Ultra Turrax T50 with a 45-mm stator (Ika-Werke, Staufen, Germany). The optimal mixing time was dependent on sample material but was typically 1 min. The mixed sample was inverted 10 times and immediately a portion was poured into a polyethylene bag (205- by 58-mm flat condition; Foss, Hillerød, Denmark) to a level of 10 cm from the bottom. A plastic funnel was used to facilitate the transfer. Then the upper part of the plastic bag was sealed after removal of most air space. Two sample bags were prepared for each sample.
Spectral Measurements
Just before NIR measurement, the sample bag was inverted 10 times. Then the bag was placed in a high-fat/high-moisture cell with a 10-mm path length (Foss, Hillerød, Denmark) and immediately measured on a NIRSystems 6500 instrument equipped with a vertical transport module (Foss, Hillerød, Denmark). Reflectance measurements were performed in the wavelength region 400 to 2500 nm with 2 nm between collected data points. Two detectors were used: a Si detector for the 400- to 1098-nm region and a PbS detector for the 1100- to 2500-nm region, and the spectra were co-added. Before each sample measurement, 25 reference scans were taken on a ceramic standard supplied with the instrument, and 40 photometric scans were then collected and averaged on each sample. During the scans, the sample was moved four times across the detector area. A single determination was performed under repeatability conditions (measurements were performed by the same operator, using the same instrument, under the same conditions within a short period of time) on each of the two sample bags filled with the sample material.
Spectral Transformations and Calibration
Data were treated with WINISI Version 1.50 software (Foss). Spectral homogeneity of samples was tested before calibration. Samples with a global H based on a Mahalanobis distance (Mark and Tunnell, 1985) of more than 3.5 were deleted as spectral outliers. Calibration was performed by partial least square (PLS) regression on mean spectra with and without scatter correction using multiplicative scatter correction (Geladi et al., 1985) or the standard normal variate transformation combined with detrend (SNVD; Barnes et al., 1989). Additional pretreatment was performed by derivative treatments (Næs et al., 2002, p. 107–114). Wavelengths separated by 8 nm were used in the calibration. Six cross-validation segments were used in the calibration process to determine the optimal number of PLS factors and to estimate the performance of developed calibration models. In cross-validation, calibration models are subsequently developed on parts of the data and tested on other parts. One outlier elimination pass was accepted. A relatively conservative criterion based on a T-value (residual to root mean square error of cross-validation [RMSECV]) of 3.0 was used. The number of PLS factors selected was equal to the fewest factors giving the lowest RMSECV.
Reference Analyses
The total content of N and C was determined by the Dumas principle (Sweeney, 1989), using a Leco CN-2000 instrument (Leco Corp., St. Joseph, MI). Determination of NH4–N was performed by flow analysis according to ISO standard 11732 (International Organization for Standardization, 2005). The elements Na, K, Ca, Mg, Mn, Cu, Zn, and P were measured by inductively coupled plasma (ICP)–optical emission spectrometry after microwave-assisted digestion in HNO3, using an Optima 4300 DV ICP instrument (PerkinElmer, Shelton, CT) and a Mars 5 microwave oven (CEM Corp., Matthews, NC). The DM content was determined from drying at 103°C for 20 h. In all cases, two single determinations were performed on different days.
The content of plant-available N (PAN) in slurry samples was measured by soil incubation experiments. A slurry sample volume equivalent to 12 mg total N was placed on 30 g of soil (dry-weight basis) in a 250-mL polyethylene bottle and then covered with another 30 g of soil (200 mg N kg–1 soil) to simulate an incorporation of slurry in soil. The soil used was a loamy sand soil (mixed, mesic Typic Hapludult) with pH 6.8 containing 17.2 g C kg–1, 1.6 N g kg–1, and 90 g clay kg–1. The soil was sampled from 0 to 20 cm of an arable field and sieved (5 mm) in moist form a few days before starting the experiment. After the manure application, the water content was adjusted to 55% of the water-holding capacity, where the conditions for microbial activity are nearly optimal. The incubation bottles were covered with Parafilm "M" (American National Can, Chicago, IL) with holes for aeration and placed in a dark, temperature-controlled room at 10°C. The soil moisture content was kept nearly constant by controlling the weight of the bottle weekly and adding extra water if necessary. The experimental conditions ensured minimal NH3 loss after the application. Slurry samples were incubated in triplicate, and similar soil samples without manure addition were incubated under the same conditions. After 11 wk, the soil mineral N (NH4, NO3, and NO2) was extracted with 2 M KCl and analyzed by flow analysis using an AutoAnalyzer 3 (Bran+Luebbe GmbH, Norderstedt, Germany). The net release of mineral N from the slurry was calculated from the mineral N in soil treated with slurry minus the mineral N in unamended soil.
Statistics
The standard error of prediction (SEP), which expresses the accuracy of NIR results corrected for the mean difference between NIR and reference methods (bias), was calculated by the following formula:
where xi – yi is the difference between results obtained by the NIR method (xi) and reference method (yi) on sample i,
and N is the total number of samples in the test.
The RMSEP was calculated from the difference between NIR and reference results:
The RMSEP includes SEP and bias in a single term, and the relation between the two terms is given by RMSEP2 
SEP2 + bias2. When the bias is insignificant, the RMSEP tends toward SEP with increasing data number. Generally, RMSEP gives a more realistic estimate of the prediction capability of a calibration than SEP. The SEP and RMSEP terms are calculated on a test set collected independently of the calibration samples.
The RMSECV was calculated by the same formula as for RMSEP. The difference is that the RMSECV was calculated from cross-validation on the calibration set and not on an independent test set.
Samples with global H values above 3.5 were not included in the calculation of the statistical terms. True prediction accuracy (SEPtrue) was estimated using the formula
where SDt is the total standard deviation of the final reference result.
The repeatability standard deviation SDr (i.e., the variability of independent single results obtained by the same operator, using the same apparatus, under the same conditions, on the same test sample, and in a short interval of time) and the intra-laboratory reproducibility standard deviation SDR,intra (i.e., the variability of independent single results obtained on the same test sample in the same laboratory by different operators under different experimental conditions) were determined in accordance with ISO standard 5725–2 (International Organization for Standardization, 1994).
|
RESULTS AND DISCUSSION
|
|---|
Sample Preparation and Presentation
Slurry samples may be quite heterogeneous, consisting of an aqueous phase with dissolved substances together with solid particles including straw materials. Before subsampling for NIR measurements and chemical analyses, the samples were blended with a powerful high-speed device to break up lumps and to reduce particle sizes. The blended samples were more homogeneous but still unstable suspensions of different sized particles. Due to the heterogeneous nature of slurry samples, scanning of a large interacting sample volume may be essential for obtaining good precision and accuracy of measurements. For that purpose, a technique where the test sample is enclosed in a polyethylene bag and moved vertically across the detector area during measurements was selected. This is clearly a laboratory technique that is less suitable for field measurements. For laboratory use, however, the technique results in a relatively clean procedure and the sealed test samples may be stored in the same way as the original sample for a period of time before and after NIR measurement. During measurement, the sample particles will settle to some extent but because of the repeated vertical movement of the sample across the detection area, the settlement is averaged out. Different batches of plastic bags of the same product specification have been used during the test period. No significant batch variation influencing the SEP values was observed.
Other sample presentation strategies were evaluated before the bag sampling technique was selected. One of these included the use of fiber-optic probes, which are immersed in the sample while simultaneously stirring the sample during data acquisition either by a separate device or by the probe. That would be an easy technique for data acquisition but reported results indicate that fiber-optic probes at the present development stage are less suitable when maximum accuracy is important (Reeves and Van Kessel, 2000a; Saeys et al., 2005a, 2000b). That may be ascribed to the shorter wavelength region in which fiber-optics have their optimal performance (Reeves and Van Kessel, 2000a) but biofouling of the sensor surface could also be a problem in routine use (Saeys et al., 2005a). Reported studies indicate that transflectance measurements may improve the predictions (Saeys et al., 2005c). An attempt to decrease the prediction errors by transflectance measurements using probes, however, would not succeed for the sample materials investigated since a relatively short optical pathway is needed, which may easily be clogged by sample particles. Another strategy investigated was a technique where the sample is circulated through a tube system containing an integrated flow cell for NIR measurement (Dolud et al., 2005); however, this technique was found less suitable for relative small laboratory samples and the technique still requires some handling. Furthermore, it would also be sensitive to clogging if the particle sizes are not reduced to a certain level.
Calibration Development
Due to the large variation in DM content and the physical property of the samples, spectral data was collected in reflectance mode. Samples were scanned across the range 400 to 2500 nm, but only data acquired in the 1200- to 2400-nm region were used in calibration development. The 400- to 700-nm visible region and the 700- to 1098-nm NIR region acquired by the Si detector were not used, as they did not improve prediction errors. That may be due to increased complexity of spectral data as photons in these regions may penetrate through samples with low DM content. Even when calibration was restricted to samples with 50 to 130 g kg–1 DM, however, these regions did not add significant spectral information that could not be derived from the 1100- to 2500-nm region. This observation conforms to results obtained in other studies using fiber optics (Reeves and Van Kessel, 2000a) and bag sampling (Reeves and Van Kessel, 2000b). They found the 400- to 1100-nm region less useful than the 1100- to 2500-nm region and they designated the 1800- to 2498-nm region as essential for determining dairy manure nutrients. In our study, the 1100- to 1200- and 2400- to 2500-nm regions measured by the PbS detector were moreover omitted during the calibration development process, as they did not contribute significantly to the prediction accuracy.
Calibration by PLS regression was performed on the combined sets of cattle and pig slurries. Initial experiments showed that it was not possible to obtain significant improvements in prediction errors by developing separate calibrations for cattle and pig slurries. Single calibrations covering both matrices are advantageous in practice because the origin of samples may not always be known by the testing laboratory.
Different pretreatments of spectral data were tested for their ability to remove or reduce disturbing effects not related to the chemical absorption of light. The optimal spectral pretreatment proved to be either first derivative treatment or SNVD, depending on the parameter analyzed (Table 3). The resulting spectral variations of calibration samples derived from these pretreatments are shown in Fig. 1 and 2. Combining different pretreatments did not improve prediction errors except in the case of P. The maximum number of T-outliers removed during calibration was six (2.4%).
View larger version (14K):
[in this window]
[in a new window]
|
Fig. 1. Spectra of calibration samples in the 1100- to 2500-nm region pretreated with first derivative treatment (1,10,10,1).
|
|
View larger version (17K):
[in this window]
[in a new window]
|
Fig. 2. Spectra of calibration samples in the 1100- to 2500-nm region pretreated with standard normal variate transformation combined with detrend.
|
|
Prediction Capability and Intercorrelations
The SD/RMSECV ratios, describing the variation in samples to the size of prediction errors, indicate that NIRS could be a feasible technique for determination of DM, N, NH4–N, C, PAN, and P in slurry samples (Table 3). A commonly used decision criterion is based on a ratio limit of 3. The results obtained on the other elements showing SD/RMSECV ratios around 2 were less obvious for practical use. That was especially true for Zn and the alkali metals Na and K. Inorganic ions and substances can generally not be determined directly by NIR spectroscopy. In some cases, however, they may induce wavelength shifts or they may be intercorrelated to some organic substances in the sample matrix, which makes indirect prediction possible. Plots visualizing the correlation between NIR and reference results are shown in Fig. 3.
View larger version (20K):
[in this window]
[in a new window]
|
Fig. 3. Results obtained by cross validation on calibration sample set. Calibration was performed in the wavelength range 1200 to 2400 nm with pretreatments listed in Table 3 (DM, dry matter; PAN, plant-available nitrogen; NIRS, near infrared spectroscopy).
|
|
To explore the predictive capability of NIR spectroscopy in more detail, the coefficients of determination were compared with the corresponding coefficients obtained from calculation of analyte concentrations from reference results. That may provide valuable information on the ruggedness of the NIR technique. It is mainly O–H, C–H, and N–H bond vibrations that contribute to the NIR spectra. If the analyte is highly correlated to main substances as DM, N, or NH4–N, then the NIR calibration models may predominantly be based on the chemistry of the main substances and not directly on the chemistry of the analyte. If these intercorrelations are not stable to changes in the slurry production and storage processes, it may be difficult to obtain rugged calibrations.
The N content could by predicted by NIRS with an r2 of 0.94 while the correlation between reference data on DM and N could only be expressed by an r2 of 0.22 (Table 4). Thus, the developed NIR calibration was probably based on direct N bond information without significant disturbance from indirect correlations between N and DM. That improves the ruggedness of the developed calibration.
View this table:
[in this window]
[in a new window]
|
Table 4. Coefficients of determination obtained in prediction of analyte concentration from near infrared (NIR) data and reference values.
|
|
The prediction power of NIR spectroscopy in determination of NH4–N was characterized by a relative high r2 value. The NH4–N content could not be related to contents of DM and non-NH4 N (r2 = 0.08). Fortification experiments performed on 15 samples using NH4H2CO3 also demonstrated that N–H bonds present in NH4 were measured by the calibration model (Fig. 4). This is illustrated by the offsetting of the lines presenting the NH4–N concentration profiles of the samples before and after fortification with 2.2 g kg–1 NH4–N added as NH4H2CO3, which was nearly constant with a mean difference of 2.2 g kg–1.
View larger version (11K):
[in this window]
[in a new window]
|
Fig. 4. Ammonium-N results (g kg–1) obtained by near infrared spectroscopy (NIRS) measured before (o) and after ( ) fortification with 2.2 g kg–1 N added as NH4H2CO3.
|
|
The C content could be predicted with a high r2, but as expected there was a high correlation between C and DM contents in samples. A similar situation was the case for PAN, where a high correlation between PAN and NH4–N was observed. The PAN is influenced by the content of NH4–N and by the N mineralization and immobilization processes in soil. The mineralization of organic slurry N is influenced by the composition of organic N compounds while the microbial N immobilization is influenced by the slurry organic matter composition. The net mineralization of organic slurry N after 11-wk incubation in soil varied between –30 and 59% for the samples used. The results obtained showed that this variation could not be detected accurately enough by NIRS to improve the prediction of PAN by direct measurement compared with prediction from NH4–N content. This conforms to reported results from other studies on the use of NIRS for predicting N mineralization of plant materials (Bruun et al., 2005) and dairy manures (Van Kessel and Reeves, 2002).
Prediction of P from NIR spectra was more accurate than calculation of P from reference data on DM and N. That may be due to other more or less complex inner relations between the analyte and matrix components. Fortification experiments with NaH2PO4 proved that free phosphate is not measured directly by the developed calibrations (Fig. 5). This is illustrated by the missing offset of lines presenting the P concentration profiles of the samples before and after fortification with 1.8 g kg–1 P added as NaH2PO4. Fortification with NH4H2PO4, however, showed that the P calibration to some minor extent was related to the N content (Fig. 6). The offsetting of lines was 0.15 g kg–1 P when samples were fortified with 2.0 g kg–1 P.
View larger version (11K):
[in this window]
[in a new window]
|
Fig. 5. Phosphorus results (g kg–1) obtained by near infrared spectroscopy (NIRS) measured before (o) and after () fortification with 1.8 g kg–1 P added as NaH2PO4.
|
|
View larger version (12K):
[in this window]
[in a new window]
|
Fig. 6. Phosphorus results (g kg–1) obtained by near infrared spectroscopy (NIRS) measured before (o) and after () fortification with 2.0 g kg–1 P added as NH4H2PO4.
|
|
Copper was predicted much more accurately by NIRS than by calculation from N and DM contents (Table 4). That may also be due to more or less complex inner relations between the analyte and matrix components.
The r2 values obtained by NIRS on K, Na, Mg, Ca, S, and Zn were in general not higher than the r2 values for the correlations between the analytes and the contents of DM and N based on reference data. In the case of K and Na, it was less. Direct determination by NIRS was slightly more accurate, however, than calculation of results from NIR-predicted contents of DM and N (Table 4).
Validation
The results obtained in the study were within the range of results reported from other studies on similar sample matrices (Table 5). The results reported by Ye et al. (2005) showed generally higher accuracy for minerals but their calibrations were based on a smaller sample population and a larger fraction of samples were removed as outliers.
View this table:
[in this window]
[in a new window]
|
Table 5. Comparison of obtained standard deviation of the concentration range (SD)/root mean square error of cross validation (RMSECV) ratios with corresponding values reported from other studies performed on similar matrices.
|
|
The calibrations (except for PAN) developed on samples from 2004 and 2005 were tested on incoming laboratory samples from spring 2006. Generally, the obtained SEP and SD/RMSEP values obtained on the test set were of the same size as the RMSECV and SD/RMSECV values obtained by cross-validation on the calibration set (Table 6). That shows it is possible to develop useful and rugged NIR calibrations for the determination of DM, N, NH4–N, C, and P in slurry samples.
Precision of Measurements
The SDr values of the NIR measurements including variation from sampling in plastic bags (Table 7) were approximately a factor of 2 less than the SEP values. From this, the accuracy of NIRS could not be improved significantly by performing double measurements instead of single measurements and the sample presentation technique could be characterized as reproducible.
The SDr values of NIR measurements were higher than the SDR,intra values of the reference method for the analytes DM, N, and NH4–N (Table 7). For the other elements, the precision figures were comparable.
Since the imprecision of the reference results was generally lower than the prediction errors of NIRS, the estimated true accuracy of the NIR method was not markedly higher than expressed by the RMSECV and SEP values.
|
CONCLUSIONS
|
|---|
This study shows that it is possible to predict the content of DM, N, NH4–N, and P in slurry samples from cattle and pigs by NIRS, with prediction errors that are several magnitudes lower than the variation in analyte content. The NIR spectra provide selective information for these substances. As in most cases, NIR spectroscopy proved to be less accurate than the classical methods for the determination of DM, N, NH4–N, and P, but the difference may be without practical importance taking into account the field sampling errors. The spectroscopic technique is clearly advantageous because of its simplicity and speed, which makes it suitable for subsampling of field batches.
Carbon and PAN could be predicted with high coefficients of determination but on the samples investigated, NIRS did not improve prediction accuracy compared with calculation from DM and NH4–N, respectively.
The spectroscopic technique seems to be less useful in determination of the elements K, S, Na, Ca, Mg, Cu, and Zn. Although the coefficients of determination may be acceptable for screening purposes, the derived spectral information was in general highly related to the content of DM and N. The only exception appears to be Cu, where NIR measurements did provide an accuracy that could not be obtained by prediction from DM and N contents.
|
ACKNOWLEDGMENTS
|
|---|
We thank the Danish Ministry of Food, Agriculture and Fisheries for funding the project.
|
NOTES
|
|---|
All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Permission for printing and for reprinting the material contained herein has been obtained by the publisher.
Received for publication September 19, 2006.
|
REFERENCES
|
|---|
- Aiken, G.E., D.H. Pote, S.F. Tabler, and T.C. Tabler. 2005. Application of near-infrared reflectance spectroscopy to estimate chemical constituents in broiler litter. Commun. Soil Sci. Plant Anal. 36:2529–2539.[CrossRef][ISI]
- Barnes, R.J., M.S. Dhanoa, and S.J. Lister. 1989. Standard normal variant transformation and de-trending of near infrared diffuse reflectance spectra. Appl. Spectrosc. 43:772–777.[CrossRef][ISI]
- Bruun, S., B. Stenberg, T.A. Breland, J. Gudmundsson, T.M. Henriksen, L.S. Jensen, A. Korsæth, J. Luxhøi, F. Palmason, A. Pedersen, and T. Salo. 2005. Empirical predictions of plant material C and N mineralisation patterns from near infrared spectroscopy, stepwise chemical digestion and C/N ratios. Soil Biol. Biochem. 37:2283–2296.[CrossRef]
- Dolud, M., H. Andree, and T. Hügle. 2005. Rapid analysis of liquid hog manure using near-infrared spectroscopy in flowing conditions. p. 115–122. In S. Cox (ed.) Precision livestock farming ‘05. Wageningen Acad. Publ., Wageningen, the Netherlands.
- Geladi, P., D. MacDougall, and H. Martens. 1985. Linearisation and scatter-correction for near-infrared reflectance spectra of meat. Appl. Spectrosc. 39:491–500.[CrossRef][ISI]
- International Organization for Standardization. 1994. Accuracy (trueness and precision) of measurement methods and results. Part 2: Basic method for the determination of repeatability and reproducibility of a standard measurement method. ISO Standard 5725–2. ISO, Geneva.
- International Organization for Standardisation. 2005. Water quality. Determination of ammonium nitrogen—Method by flow analysis (CFA and FIA) and spectrometric detection. ISO Standard 11732. ISO, Geneva.
- Kemsley, E.K., H.S. Tapp, A.J. Scarlett, S.J. Miles, R. Hammond, and R.H. Wilson. 2001. Comparison of spectroscopic techniques for the determination of Kjeldahl and ammoniacal nitrogen content of farmyard manure. J. Agric. Food Chem. 49:603–609.[CrossRef][ISI][Medline]
- Malley, D.F., C. McClure, P.D. Martin, K. Buckley, and W.P. McCaughey. 2005. Compositional analysis of cattle manure during composting using a field-portable near-infrared spectrometer. Commun. Soil Sci. Plant Anal. 36:455–475.[CrossRef][ISI]
- Malley, D.F., L. Yesmin, and R.G. Eilers. 2002. Rapid analysis of hog manure and manure-amended soils using near-infrared spectroscopy. Soil Sci. Soc. Am. J. 66:1677–1686.[Abstract/Free Full Text]
- Mark, H.L., and D. Tunnell. 1985. Qualitative near-infrared reflectance analysis using Mahalanobis distances. Anal. Chem. 57:1449–1456.
- Millmier, A., J. Lorimor, C. Hirburgh, Jr., C. Fulhage, J. Hattey, and H. Zhang. 2000. Near-infrared sensing of manure nutrients. Trans. ASAE 43:903–908.[ISI]
- Næs, T., T. Isaksson, T. Fearn, and T. Davies. 2002. A user-friendly guide to multivariate calibration and classification. NIR Publ., Chichester, UK.
- Reeves, J.B., III. 2001a. Near- versus mid-infrared diffuse reflectance spectroscopy for determination of minerals in dried poultry manure. Poult. Sci. 80:1437–1443.[Abstract/Free Full Text]
- Reeves, J.B., III. 2001b. Near-infrared diffuse reflectance spectroscopy for the analysis of poultry manures. J. Agric. Food Chem. 49:2193–2197.[CrossRef][ISI][Medline]
- Reeves, J.B., III, and J.S. Van Kessel. 2000a. Determination of ammonium-N, moisture, total C and total N in dairy manures using a near infrared fibre-optic spectrometer. J. Near Infrared Spectrosc. 8:151–160.
- Reeves, J.B., III, and J.S. Van Kessel. 2000b. Near-infrared spectroscopic determination of carbon, total nitrogen, and ammonium-N in dairy manures. J. Dairy Sci. 83:1829–1836.[Abstract]
- Saeys, W., P. Darius, and H. Ramon. 2004. Potential for on-site analysis of hog manure using a visual and near infrared diode array reflectance spectrometer. J. Near Infrared Spectrosc. 12:299–309.
- Saeys, W., A.M. Mouazen, and H. Ramon. 2005a. Potential for onsite and on-line analysis of pig manure using visible and near infrared reflectance spectroscopy. Biosyst. Eng. 91:393–402.[CrossRef]
- Saeys, W., A.M. Mouazen, and H. Ramon. 2005b. Potential for on-site and on-line analysis of hog manure using visual and near-infrared reflectance spectroscopy. p. 131–138. In S. Cox (ed.) Precision livestock farming ‘05. Wageningen Acad. Publ., Wageningen, the Netherlands.
- Saeys, W., J. Xing, J. De Baerdemaeker, and H. Ramon. 2005c. Comparison of transflectance and reflectance to analyse hog manures. J. Near Infrared Spectrosc. 13:99–107.
- Sweeney, R.A. 1989. Generic combustion method for determination of crude protein in feeds: Collaborative study. J. Assoc. Off. Anal. Chem. 72:770–774.[Medline]
- Tasistro, A.S., S. Shaaban, D.E. Kissel, and P.F. Vendrell. 2003. Near-infrared reflectance spectroscopy for the analysis of water and total nitrogen contents in poultry litter. Commun. Soil Sci. Plant Anal. 34:1367–1379.[CrossRef][ISI]
- Van Kessel, J.S., and J.B. Reeves III. 2002. Nitrogen mineralisation potential of dairy manures and its relationship to composition. Biol. Fertil. Soils 36:118–123.[CrossRef]
- Ye, W., J.C. Lorimor, C. Hurburgh, H. Zhang, and J. Hattey. 2005. Application of near-infrared reflectance spectroscopy for determination of nutrient contents in liquid and solid manures. Trans. ASAE 48:1911–1918.[ISI]
TOP
ABSTRACT
INTRODUCTION
