Abstract
Recently, acid detergent analysis has been reported to provide valid data to evaluate decomposition properties and to determine the available nitrogen (AVN) of organic materials, such as compost. However, this methodology requires complex procedures and creates considerable costs. As an alternative, near infrared spectroscopy (NIRS) was evaluated as a simple method to determine acid detergent fiber (ADF), acid detergent lignin (ADL) and acid-detergent-soluble organic matter (ADSOM), in order to evaluate the decomposition properties of cattle and swine manure compost. To establish an easy and accurate method of estimating AVN in cattle and swine manure compost, the accuracies of direct estimations of AVN by NIRS in incubation experiments and indirect estimations by NIRS based on acid-detergent-soluble nitrogen (ADSN) or total nitrogen (TN) were examined. The reflectance spectra of freeze-dried and milled compost samples were determined using a scanning monochromator. Second derivative spectra and multiple regression analysis were used to develop calibration equations for each constituent. The calibration equations for ADF, ADL and ADSOM were “successful” according to commonly applied criteria. Acid-detergent-soluble nitrogen was found to be more suitable than TN for estimating AVN in cattle and swine manure compost. As the accuracies of the estimations of ADSN and TN by NIRS were comparable, the estimation of AVN based on ADSN as determined by NIRS was more accurate than that based on TN determined by NIRS. The direct prediction of AVN through NIRS was not as accurate as the estimation of AVN based on ADSN determined by NIRS. We conclude that NIRS is a practicable alternative to the time-consuming acid detergent analysis of cattle and swine compost, and that ADSN as determined by NIRS is useful for estimating AVN in the compost.
INTRODUCTION
With the increased scale of livestock farming, the management of livestock waste has become a serious problem in Japan (CitationHarada et al. 1993). The Japanese government has enforced environmental laws to encourage the production of compost from livestock waste to reduce both the disposal of untreated livestock waste and the use of chemical fertilizer (CitationIkumo 2005). To use livestock waste compost effectively, it is important to supply sufficient amounts of nutrients to the soil, but high doses of compost might cause groundwater contamination. Therefore, appropriate management plans for the application of compost derived from livestock manure to farmlands are required. As compost serves as a soil amendment as well as an organic fertilizer, suitable evaluation methods of compost quality for both purposes need to be developed.
The aim of applying compost as a soil amendment is the physical, chemical and biological improvement of the soil. The improvement of soils achieved by the application of cattle and swine manure compost needs to be critically evaluated; generally it is greater than in the case of poultry manure compost (CitationFujiwara 2006). Assuming that these differences result from different kinetics of organic matter decomposition, CitationOyanagi et al. (2007) evaluated the decomposition properties of livestock manure compost by acid detergent analysis, which is well established as a method of fiber analysis in forage and food. They demonstrated that this method could be used to characterize readily decomposable organic matter (acid-detergent-soluble organic matter [ADSOM]; CitationOyanagi et al. 2007), organic matter decomposable within 3 months (acid detergent fiber [ADF]) and organic matter decomposable within 3 years (acid detergent lignin [ADL]; CitationOyanagi and Ando 2006). Obviously, this analysis is time consuming, requires complex procedures and creates considerable costs. Near infrared spectroscopy (NIRS) is a simple and rapid analytical method that has been used to determine ADF and the neutral detergent fiber of forage, grain and processed food (CitationKays 2004; CitationOsborne and Fearn 1986; CitationRoberts et al. 2004). Therefore, NIRS has the potential to replace acid detergent analysis of livestock manure compost.
The most important function of compost as an organic fertilizer is as a nitrogen supply to the crop. To determine appropriate application rates of compost to fields, the available nitrogen (AVN), which is defined as the sum of inorganic nitrogen and mineralized nitrogen, has to be known. Because incubation experiments, as the standard method to estimate AVN, require considerable time, the development of alternative methods is desirable. CitationFujiwara and Murakami (2007) developed a simple method of AVN estimation in poultry manure compost using NIRS to analyze uric acid nitrogen, which is readily available for plants (CitationMurakami et al. 2007). Unfortunately, cattle and swine manure do not contain uric acid. CitationVan Kessel and Reeves (2002) studied the nitrogen mineralization of dairy manure using compositional factors and NIRS, but found no simple relationship between nitrogen mineralization and the compositional characteristics. On the contrary, CitationOyanagi et al. (2007) reported a correlation between acid-detergent-soluble nitrogen (ADSN) and the decomposition within 3 months of nitrogen compounds in organic materials, including cattle, swine and poultry manure compost, organic fertilizer and dried garbage. Assuming that the decomposition of nitrogen compounds in 3 months corresponds to the nitrogen mineralization occurring over a single crop cultivation period, they concluded that ADSN might provide a valid estimate of AVN in organic materials. CitationMatsumura and Sato (2005) found a strong correlation between the ADSN of organic materials, such as rapeseed meal, fish meal and cattle manure compost, and the proportion of the total nitrogen (TN) contained in these organic materials that was absorbed by cultivated tomato plants. Taken together, these reports suggested that if a simple method of ADSN analysis was available, it would also be useful for estimating the AVN of cattle and swine manure compost.
We were interested in determining whether the decomposition properties and AVN of cattle and swine manure compost could be evaluated by NIRS. We tested whether the ADSN in manure compost was predictive of AVN, and studied the accuracy of the analysis for each fraction defined by acid detergent analysis, including ADF, ADL, ADSOM and ADSN.
MATERIALS AND METHODS
Compost samples
Cattle and swine manure compost samples (n = 100) were obtained from farms in the Mie, Gifu and Niigata prefectures, Japan. summarizes the number of the samples, composting methods, contents by acid detergent analysis, and TN and AVN. Most samples (95%) of cattle and swine manure compost that had been produced in compost depots or open-type composters contained additives such as rice husks and sawdust, whereas 80% of the swine manure compost produced in enclosed-type composters was free of additives. The average ratio of additives to the manure was 68% (2–80% v/v). The samples were freeze-dried and then milled (Ultra centrifugal mill fitted with 1.0 mm screen; Retsch, Haan, Germany) for chemical and spectral analyses.
Incubation experiments
Although AVN is usually evaluated by incubation experiments lasting 4 weeks at 30°C, CitationOyanagi et al. (2007) proposed that incubation for 12 weeks yields results that correspond more closely to nitrogen mineralization over one crop cultivation period. Therefore, we carried out incubation experiments over 12 weeks.
Brown lowland soil was collected from the plow layer of fields on the experimental farm of the Gifu prefectural agricultural technology center. The chemical properties of the soil were: pH (H2O) 6.25, electrical conductivity (soil : water = 1:5) 3 mS m−1, cation exchange capacity 10.8 cmolc kg−1, Truog-P2O5 0.93 mg g−1, exchangeable K2O 0.14 mg g−1, exchangeable MgO 0.31 mg g−1, exchangeable CaO 2.0 mg g−1, and phosphate absorption coefficient 4.2 mg g−1. A fresh compost sample with a TN content of 30 mg was mixed with 100 g of fresh soil, and was wetted with distilled water to a moisture content of 45% of the maximum water-holding capacity. After incubation at 30°C in the dark for 12 weeks, the inorganic nitrogen content was determined by 2 mol L−1 KCl extraction and the steam distillation method. The net nitrogen mineralization was calculated by subtracting the inorganic nitrogen of an incubated control (soil without compost) from the inorganic nitrogen of an incubated sample.
References analysis
Acid detergent analysis was used to determine ADF, ADL, ADSOM and ADSN (CitationOyanagi et al. 2007). Compost (1 g) was mixed with 100 mL of acid detergent solution (20 g cetyltrimethylammonium bromide dissolved in 1 L 0.5 mol L−1 sulfuric acid). The mixture was boiled for 1 h, filtrated using a glass crucible, washed with hot water under reduced pressure, dried and weighed (a). The dried residue was mixed thoroughly with 15 mL 12 mol L−1 sulfuric acid and kept under room temperature for 4 h and stirred every 30 min. Water was added to make up a total volume of 300 mL and the mixture was boiled for 10 min, filtered, washed with water, dried and weighed (b). The dried residue was incinerated and weighed again (c). ADF and ADL were calculated as (a – c) and (b – c), respectively. The ADSOM was obtained by subtracting crude ash and ADF from the dry weight of the sample. The ADSN was obtained by subtracting nitrogen in the dried sample in (a) from the TN content of the sample. Nitrogen was analyzed using a C/N analyzer (NC-90A; Sumica Corporation, Tokyo, Japan).
Table 1 Composting methods and contents of the livestock manure compost
Near infrared spectroscopy
Reflectance spectra were measured using an NIR spectrophotometer (Model 6500; Foss-NIRSystems, Laurel, MD, USA) with a standard sample cup. Samples were scanned 32 times from 400 to 2500 nm, and data were collected every 2 nm. Spectra were obtained as log(1/R) with a ceramic tile used as a reflectance standard. To verify the absorption bands of the wavelengths for the calibration equations, reflectance spectra of cellulose, lignin and protein (albumin) were determined in the same manner.
Regression analysis
Data were divided into a calibration set (60 samples) and a prediction set (40 samples). These sample sets were selected so that the distributions of each constituent were similarly spread, and care was taken to ensure that the calibration set did not concentrate around any point in the range. For resolution enhancement and baseline correction (CitationOzaki et al. 2007), second derivative spectra of each sample were used for calibration and prediction. When producing the derivatives, segment size and gap size (CitationKawano et al. 1992) were set to 10 and 0, respectively.
The calibration equations were developed by multiple linear regression using the second-derivative values of absorbance (d2log[1/R]) and the constituents of the calibration set. The d2log(1/R) data from 800 nm to 2500 nm, which corresponds to the near infrared region, were used for calibration. The accuracy of the calibration equations was determined by evaluating the differences between the values predicted by the equations and the actual results of the analyses of the prediction set.
All analytic procedures were conducted using the Vision software package ver.3.2 (Foss-NIRSystems Inc.).
Table 2 Correlation coefficients between the constituents of cattle and swine compost
RESULTS
Detergent analysis and nitrogen contents of the samples
Results of the detergent analysis as well as the TN and AVN of the compost samples were widely spread for each type of livestock and composting method (). Swine manure composts showed higher average ADSOM, ADSN, TN and AVN than cattle manure composts. CitationHarada et al. (1993) reported similar results for TN. The enclosed-type composter (fermentation tank of the vertical kiln type) produced swine manure compost with higher ADSOM, ADSN, TN and AVN than the other composting methods.
Relationship between the constituents of the samples
A significant positive correlation was found between ADF and ADL, whereas strong negative correlations were detected between ADF or ADL on the one hand and ADSOM, ADSN, TN or AVN on the other hand ). Significant positive correlations were also observed between ADSOM, ADSN, TN and AVN.
The correlation coefficient for ADSN and AVN (r = 0.941) was higher than that for TN and AVN (r = 0.892). The correlation coefficient for AVN and TN was lower than that reported by CitationFujiwara and Murakami (2007) for poultry manure compost (r = 0.948). The regression equations of estimating AVN using ADSN or TN are given in .
Key wavelengths for calibration
As the second derivative transforms maxima into minima, absorption peaks are turned into troughs. As a result, the correlation coefficients between d2log(1/R) and the contents of the corresponding constituents should be negative (CitationKawano et al. 1992). Therefore, we focused on the wavelengths that showed a high negative correlation coefficient between d2log(1/R) and the constituents, as shown in . A high negative correlation was observed at similar wavelengths for ADF and ADL, and at other wavelengths for ADSOM, ADSN, TN and AVN. also lists the absorption peaks of the second-derivative spectra of cellulose, lignin and protein. Cellulose and/or lignin showed absorption peaks at approximately the wavelengths with high correlations between ADF or ADL and d2log(1/R). Similarly, protein had absorption peaks close to those of ADSOM, ADSN, TN and AVN.
Figure 1 Relationship between (a) acid-detergent-soluble nitrogen and (b) total nitrogen and available nitrogen in cattle (•) and swine (○) manure compost. Se, standard error of the regression estimates.
Table 3 Wavelengths that showed high negative correlations (r > 0.7) between the second derivatives of absorbance and the constituent contents in compost samples of the calibration set
Table 4 Multiple regression and deviations between the predicted and actual constituent contents in cattle and swine compost as determined by near infrared spectroscopy using second derivative spectra
Regression analysis
Results of the multiple regression analyses and predictions derived from the calibration equations are summarized in and . These equations were selected considering a low standard error of prediction (SEP) and the maximum wavelength number without statistical overfitting (CitationMartens and Naes 1987). No significant differences were observed between the results of the analyses and the predicted values (paired t-test, P < 0.05).
To standardize the SEP, the ratio of the standard deviation of the prediction set to SEP (RPD) and the ratio of the prediction set range to SEP (RER) can be used; high values of RPD and RER indicate reliable results (CitationWilliams 1987). CitationMalley et al. (2002) proposed evaluation criteria for calibration equations to be used for soil, animal manure and compost samples. According to these criteria, our calibration equations for ADF and ADSOM were “successful” or “moderately successful”, those for ADL, ADSN and TN were “moderately successful”, and the equation for AVN was “moderately useful”.
Estimating available nitrogen using NIRS
shows comparisons of actual and predicted values of AVN, which were estimated by the regression equations in (ADSN) and (TN) using ADSN and TN values predicted by NIRS with the calibration equations in . The estimation of AVN using ADSN values established by NIRS () appeared more accurate than that using TN (), and was also more accurate than the direct prediction of AVN by NIRS ().
DISCUSSION
Detergent analysis and nitrogen contents of the samples
The results of our analysis showed considerable variability (), probably because of differences in the nature and amounts of the additives. Differences in the maturity of the compost might also have played a role because the results of the detergent analysis change during the composting process of cattle (CitationTakahashi 2004) and swine feces (CitationTakahashi 2006).
Swine manure compost showed higher average TN than cattle manure compost, corroborating the report by CitationHarada et al. (1993). A fermentation tank of the vertical kiln type produced swine manure compost with particularly high TN. One reason for this might be that 80% of the compost produced in this tank did not contain any additives, whereas almost all (94%) of the other swine manure compost did. Another reason could be that the organic nitrogen compounds had not decomposed as completely in the tank as in the compost produced by other methods; processing periods in tanks generally are shorter than the processing periods of the alternative methods.
Key wavelengths for calibration
As shown in , the wavelengths with high negative correlation coefficients between ADF and d2log(1/R) were similar to those with high coefficients between ADL and d2log(1/R). This result appeared in line with the strong correlation (r = 0.897) between ADF and ADL (). The wavelengths mentioned corresponded approximately to the absorption peaks of cellulose and/or lignin (), which supports the notion that ADF represents the sum of cellulose and lignin (CitationOsborne and Fearn 1986; CitationOyanagi et al. 2007).
Figure 2 Actual constituent contents versus values predicted by near infrared spectroscopy in cattle (•) and swine (○) manure compost in the prediction sample set. ADF, acid detergent fiber; ADL, acid detergent lignin; ADSOM, acid-detergent-soluble organic matter; ADSN, acid-detergent-soluble nitrogen; TN, total nitrogen; AVN, available nitrogen; SEP, standard error of the prediction without bias correction. Solid line shows y = x. Calibration equations: ADF (mg g−1) = 295.1 – 18376A2336 + 25270A2296 + 97556A1496 + 44971A1608; ADL (mg g−1) = 185.1 – 43326A2382 + 190074A1268 + 29940A2208 + 97011A1220; ADSOM (mg g−1) = 254.2 – 101816A2180 – 103573A1452 – 26868A1400 – 40041A1702 – 59147A1196; ADSN (mg g−1) = 14.86 – 1739A2300 – 3453A1422 + 982.8A1904 – 4438A1740 + 11172A976; TN (mg g−1) = 23.46 – 3964A2176 + 44346A1266 + 1715A2318 – 4779A2452 + 521.3A1898; AVN (mg g−1) = –2.30 – 636.8A2300 – 2417A2162 – 10271A1174 – 3850A2148, where Aλ is d2log(1/R) at λ (nm).
Figure 3 Actual available nitrogen (AVN) versus predicted AVN based on (a) acid-detergent-soluble nitrogen and (b) total nitrogen in cattle (•) and swine (○) manure compost in the prediction sample set. The predicted AVN values were calculated using the regression equation in Fig. 1a (acid-detergent-soluble nitrogen) and Fig. 1b (total nitrogen) using constituents determined by near infrared spectroscopy with the calibration equation in Fig. 2. SEP, standard error of prediction without bias correction. Solid line shows y = x.
The ADSOM corresponds to the total organic matter minus ADF and includes hemicellulose (CitationOyanagi et al. 2007). In addition, ADSOM appears to include considerable amounts of nitrogen compounds, according to the contents of ADSN (). Moreover, ADSOM showed significant positive correlations with ADSN, TN and AVN (). These facts probably explain why ADSOM, ADSN, TN and AVN had significant correlations with d2log(1/R) at similar wavelengths (). Second-derivative spectra of protein showed absorption peaks at these wavelengths (), and most of these could be attributed to the absorption bands of functional groups containing N–H (CitationWorkman and Weyer 2007).
The wavelengths identified in were considered diagnostic for calibrating each constituent because they showed strong negative correlations with d2log(1/R), and the absorption peaks of substances related to the constituents in question exist close to these wavelengths. Therefore, we carried out multiple regression analyses using these wavelengths.
Evaluation of compost decomposition by NIRS
Acid detergent fiber, ADL and ADSOM have been suggested to be valid indicators of the decomposition of livestock manure compost after application to fields (CitationOyanagi et al. 2007). In the present study, the calibration equations for ADF and ADSOM () were “successful” or “moderately successful”, and the equation for ADL was “moderately successful”, according to the guidelines proposed by CitationMalley et al. (2002). Although the level of accuracy of the ADF, ADL and ADSOM estimates required for practical applications remains to be established, NIRS appears to be a useful alternative to acid detergent analysis for the evaluation of the decomposition of cattle and swine manure compost.
Estimating available nitrogen by NIRS
Based on the results shown in , ADN is more suitable than TN for estimating AVN in cattle and swine manure compost. In terms of RPD and RER, the accuracy of ADSN estimated by NIRS was comparable to that of TN (). These results explained why the estimation of AVN using ADSN values established by NIRS () was more accurate than that using TN (). The direct estimation of AVN by NIRS () was not as accurate as that using the ADSN values established by NIRS. In conclusion, indirect estimation using ADSN as determined by NIRS appears to be a valid method for evaluating the AVN of cattle and swine manure compost.
ACKNOWLEDGMENTS
This work was funded by a grant for a “Research project for utilizing advanced technologies in agriculture, forestry and fisheries” from the Ministry of Agriculture, Forestry and Fisheries of Japan. The authors thank Dr Hitoshi Obata who reviewed a draft of this article and provided helpful suggestions.
Related Research Data
REFERENCES
- Fujiwara , T . 2006 . [Current problems and future prospects of research on the quality evaluation of organic vegetables.] . Hort. Res. (Japan) , 5 : 1 – 5 . (in Japanese)
- Fujiwara , T and Murakami , K . 2007 . Application of near infrared spectroscopy for estimating available nitrogen in poultry manure compost . Soil Sci. Plant Nutr , 53 : 102 – 107 .
- Harada , Y , Haga , K , Osada , T and Koshino , M . 1993 . Quality of compost produced from animal wastes . JARQ , 26 : 238 – 246 .
- Ikumo , H . 2005 . Estimation of potential supply of livestock waste compost to replace chemical fertilizer use in Japan based on 2000 census of agriculture . JARQ , 39 : 83 – 89 .
- Kawano , S , Watanabe , H and Iwamoto , M . 1992 . Determination of sugar content in intact peaches by near infrared spectroscopy with fiber optics in interactance mode . J. Japan. Soc. Hort. Sci , 61 : 445 – 451 .
- Kays , SE . 2004 . “ Analysis of cereal food products ” . In Near-infrared Spectroscopy in Agriculture , Edited by: Roberts , CA , Workman , J Jr and Reeves , JB III . 411 – 438 . Madison : American Society of Agronomy Crop Science Society of America, Soil Science Society of America .
- Malley , DF , Yesmin , L and Eilers , RG . 2002 . Rapid analysis of hog manure and manure-amended soils using near-infrared spectroscopy . Soil Sci. Soc. Am. J , 66 : 1677 – 1686 .
- Martens , H and Naes , T . 1987 . “ Linear prediction and ways to find the calibration coefficients. Multivariate calibration by data compression ” . In Near-Infrared Technology in the Agricultural and Food Industries , Edited by: Williams , P and Norris , K . 59 – 60 . St Paul : American Association of Cereal Chemists .
- Matsumura , S and Sato , T . 2005 . [Usefulness of the acid detergent fiber as an item for evaluating the mineralization characteristics of nitrogen in organic materials.] . Jpn. J. Soil Sci. Plant Nutr , 76 : 645 – 648 . (in Japanese)
- Murakami , K , Kosaka , S and Hara , M . 2007 . [The rapid estimation of available nitrogen in chicken manure compost by determination of uric acid.] . Jpn. J. Soil Sci. Plant Nutr , 78 : 85 – 87 . (in Japanese)
- Osborne , BG and Fearn , T . 1986 . “ Carbohydrates. Applications of near infrared spectroscopy in food analysis ” . In Near Infrared Spectroscopy in Food Analysis , 131 – 136 . Harlow : Longman Scientific & Technical .
- Oyanagi , W and Ando , Y . 2006 . [Evaluation of decomposition property of animal waste compost by detergent analysis.]Abstracts of the Annual Meeting . Jpn. Soc. Soil Sci. Plant Nutr , 52 : 159 (in Japanese)
- Oyanagi , W , Ando , Y and Tanahashi , T . 2007 . [Decomposition property of organic matter in soil and its indicators.] . Jpn. J. Soil Sci. Plant Nutr , 78 : 407 – 410 . (in Japanese)
- Ozaki , Y , Morita , S and Du , Y . 2007 . “ Baseline correction methods. Spectral analysis ” . In Near-Infrared Spectroscopy in Food Science and Technology , Edited by: Ozaki , Y , McClure , WF and Christy , AA . 54 – 56 . Hoboken : John Wiley & Sons .
- Roberts , CA , Stuth , J and Flinn , P . 2004 . “ Analysis of forages and feedstuffs ” . In Near-infrared Spectroscopy in Agriculture , Edited by: Roberts , CA , Workman , J Jr and Reeves , JB III . 231 – 267 . Madison : American Society of Agronomy, Crop Science Society of America, Soil Science Society of America .
- Takahashi , M . 2004 . [Accelerated decomposition of organic matter in the compost of swine feces and rice hulls by addition of slaked lime: evaluation by detergent analysis.] . Nihon Chikusan Gakkaiho , 75 : 587 – 598 . (in Japanese with English summary)
- Takahashi , M . 2006 . [Evaluation of organic matter decomposition during the maturing process in cattle feces compost by detergent analysis.] . Nihon Chikusan Gakkaiho , 77 : 259 – 267 . (in Japanese with English summary)
- Van Kessel , JS and Reeves , JB III . 2002 . Nitrogen mineralization potential of dairy manures and its relationship to composition . Biol. Fertil. Soils , 36 : 118 – 123 .
- Williams , PC . 1987 . “ Interpretation of statistical evaluation of NIR analysis. Variables affecting near-infrared reflectance spectroscopic analysis ” . In Near-infrared Technology in the Agricultural and Food Industries , Edited by: Williams , P and Norris , K . 146 – 148 . St Paul : American Association of Cereal Chemists .
- Workman , J Jr and Weyer , L . 2007 . “ Spectra-structure correlations for near infrared ” . In Practical Guide to Interpretive Near-Infrared Spectroscopy , 239 – 263 . Boca Raton : CRC Press .