| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Dep. of Crop Sci., Univ. of Illinois, 1102 S. Goodwin Ave., Urbana, IL 61801
* Corresponding author (gbollero{at}uiuc.edu)
Received for publication May 21, 2002.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Abbreviations: AD, acid detergent • ADF, acid detergent fiber • ADFC, carbon concentration in acid detergent fiber fraction • ADFN, nitrogen concentration in acid detergent fiber fraction • ADSOL, acid detergent soluble fraction • ADSOLC, carbon concentration in acid detergent soluble fraction • ADSOLN, nitrogen concentration in acid detergent soluble fraction • C/N, carbon/nitrogen ratio • CV, coefficient of variation • DCD, decomposition-day(s) • DGD, degree-day(s) • PC, principal component • PCR, principal-component regression • ND, neutral detergent • NDF, neutral detergent fiber • NDFC, carbon concentration in neutral detergent fiber fraction • NDFN, nitrogen concentration in neutral detergent fiber fraction • NDSOL, neutral detergent soluble fraction • NDSOLC, carbon concentration in neutral detergent soluble fraction • NDSOLN, nitrogen concentration in neutral detergent soluble fraction • WCC, winter cover crop
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Residue decomposition is controlled by extrinsic factors (mainly temperature and moisture) and by intrinsic factors of the residues (i.e., biochemical fractions) (Heal et al., 1997). In Ruffo and Bollero (2003), degree-days (DGD) and decomposition-days (DCD) were used as independent variables in a nonlinear approach to estimate WCC residue decomposition and nutrient release rates (k coefficient in the exponential decay function). Degree-days and DCD indexes significantly accounted for the effect of temperature and temperature and moisture. Statistical analysis of the estimated k's showed a lack of significant effect of location and a negligible variance component due to year, suggesting that DGD and DCD successfully accounted for differences in temperature and rainfall between locations and years. Once extrinsic factors are accounted for by DCD or DGD, it could be stated that rates of decomposition and release (k's) are solely a function of the intrinsic factors.
Residue decomposition depends on its biochemical fractions (Heal et al., 1997). The concentrations of nutrients, structural carbohydrates, and other compounds (i.e., lignin and other polyphenols) as well as their ratios have been used as indexes of biochemical quality. More specifically, great efforts have been devoted to develop a residue quality index that best describes C and N residue release rates. For example, in incubation studies, total N concentration (Frankenberger and Abdelmagid, 1985) or its inverse (Quemada and Cabrera, 1995) were reported to be the best indexes for C and N residue release rates of legume and grass residues. Others identified soluble C (Oglesby and Fownes, 1992; Kuo and Sainju, 1998), cellulose (Bending et al., 1998), or lignin (Müller et al., 1988; Giller and Cadisch, 1997) to be most closely related to residue decomposition or C and N mineralization rates. Furthermore, some ratios, such as lignin to N (Vigil and Kissel, 1991) or polyphenol plus lignin to N (Constantinides and Fownes, 1994), have also been used as indexes of residue nutrient release. Mechanistic models such as CENTURY (Parton et al., 1994) use the lignin/N ratio to partition residue biomass into easily decomposable (soluble carbohydrates and proteins) and recalcitrant (fibers and lignin) pools.
The biochemical components controlling residue decomposition change with time. Soluble nutrients are more relevant at earlier decomposition stages and structural carbohydrates or lignin at later stages (Heal et al., 1997). Consequently, the length of the decomposition period being analyzed will determine which fractions have more control or are more relevant in residue decomposition. The C/N ratio (C/N) is the most widely used index of residue quality and predictor of decomposition rate (Heal et al., 1997). However, the use of the initial C/N of the residues does not consider the availability of these nutrients for microbial growth; consequently, it has failed to be a reliable predictor of decomposition or mineralization (Smith et al., 1992; Honeycutt et al., 1993; McKenney et al., 1995). Vigil and Kissel (1995) concluded that N mineralization parameters were estimated poorly by C/N, especially when C/N ranged from 10 to 28. In addition, Gilmour et al. (1998) concluded that decomposition rate variations among years and type of residues were not related to crop species, year, N content, and/or C/N. These authors sustained that the variability in the kinetic parameters needs to be explained. It is therefore accepted that dynamic models that include a more detailed description of decomposition of the various chemical compounds are needed to improve prediction of C and N turnover (Dendooven et al., 1997; Heal et al., 1997). Earlier literature suggests the use of C and N concentration in the residue soluble fractions as a better indicator of residue C and N release processes (Cochran et al., 1980; Reinertsen et al., 1984; Henriksen and Breland, 1999).
The prediction of residue decomposition rates and C and N release rates is a complex problem and can be improved with models that include biochemical fractions and their interactions (Henriksen and Breland, 1999). Gordillo and Cabrera (1997) in an incubation study proposed a two-pool first-order kinetic model to describe N mineralization in broiler litter. Their approach included variables such as total C and N, C/N, uric acid N, water soluble N, and lithium carbonate soluble N as independent variables in multiple regression to select the best predictors of the fast and slow pools. These authors failed to account for multicollinearity among the predictor variables. We hypothesize that accurate prediction of residue decomposition rates and C and N release rates should be based on statistical models that account for a detailed contribution of the biochemical fractions involved and their interactions. The most common approach to develop a predictive model of decomposition or mineralization rates based on residue quality has been to relate decomposition parameters to the different residue biochemical fractions either by multiple regression (Müller et al., 1988; Trinsoutrot et al., 2000) or correlation (Thomas and Asakawa, 1993; Bending et al., 1998). Because the different biochemical fractions are highly correlated (Müller et al., 1988; Kuo and Sainju, 1998), multiple regression by ordinary least squares is not appropriate for the estimation of parameter coefficients due to the presence of multicollinearity among the predictor variables. When multicollinearity exists among the predictor variables, the variances of the parameter estimates are inflated and statistically unstable (Dillon and Goldstein, 1984; Johnson and Wichern, 1998). In addition, the results are difficult to interpret and very sensitive to the inclusion or lack of inclusion of specific variables or to small changes in data points (Dillon and Goldstein, 1984).
Principal-component regression is an effective and commonly used multivariate statistical method to predict the coefficient value of a dependent variable based on several predictor variables with multicollinearity (Jolliffe, 1986; Khattree and Naik, 2000). Concisely, PCR involves the calculation of PCs and then multiple regression using the PCs as predictor variables. Principal components are often effective in summarizing and reducing the dimensionality of a multivariate data set. In addition, the PCs are uncorrelated; thus, the problem of multicollinearity is eliminated. Although PCR has the disadvantage of being a biased estimator of the true parameters, the bias is small compared with the instability of estimates generated when multicollinearity is a serious problem (Dillon and Goldstein, 1984; Jolliffe, 1986).
The sequential biochemical analysis developed by Van Soest et al. (1991) is a simple and widely used technique for feed and forage analyses. This methodology can readily provide the information needed to predict residue decomposition rates and C and N release rates. This methodology has been used in other decomposition and mineralization studies to analyze plant biochemical fractions (Vanlauwe et al., 1994; Quemada and Cabrera, 1995; Henriksen and Breland, 1999).
We hypothesize that biochemical fractions of WCC residues at the time of killing provide sufficient information to predict residue decomposition rates and C and N release rates when used in multivariate models that account for multicollinearity. The objective of this study was to model WCC residue decomposition and C and N release rates based on biochemical fractions of the residue using PCR.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
The experiment was conducted using no-till practices on land that had been in a corn–soybean [Glycine max (L.) Merr.] rotation for at least 5 yr. Winter cover crops were drilled on soybean stubble every year. In the second year, the experiment was conducted in a field adjacent to the field used the previous year. The experimental design was a split-plot arrangement of treatments in a randomized complete block design with four replications. The main-plot treatments were WCC (rye, hairy vetch, and rye–hairy vetch in mixture) and fallow. Main plots were 9 m wide by 20 m long. Split-plot treatments were four levels of N fertilizer: 0, 90, 180, and 270 kg N ha-1. Split plots were 4.5 m wide by 10 m long and accommodated six rows of corn planted at 76 cm.
Winter cover crops were planted in the fall and chemically killed in the following spring before corn planting. Seeding rates were 134 and 34 kg ha-1 for rye and hairy vetch. In the biculture, the seeding rate was 67 kg ha-1 for rye and 23 kg ha-1 for hairy vetch. Hairy vetch was inoculated every year with Rhizobium leguminosarum var. viciae (Urbana Labs, St. Joseph, MO). Winter cover crops were killed with a mixture of 1.1 kg a.i. ha-1 glyphosate [(N-(phosphonomethyl)glycine] and 0.4 kg a.i. ha-1 2,4-D (2,4-dichlorophenoxyacetic acid) on 28 Apr. 1999 and 29 Apr. 2000 at Brownstown and 2 May 1999 and 4 May 2000 at Urbana. Corn was planted approximately 1 wk after WCCs were killed.
Grab samples were collected from the control plots (0 kg N ha-1) at six times during corn growing season to monitor residue decomposition. The first sampling time was done immediately before killing the WCC, and subsequent samples were taken approximately every 3 wk during the growing season, until 1 wk before corn harvest. Grab samples consisted of three subsamples (0.12 m2 each) of WCC residue on the first three sampling times and two on the last three sampling times. The area comprised of the three central corn rows was selected for sampling. Standing residues were cut at ground level with electric shears, and the plant material was gathered by hand. Rye and hairy vetch residue components of the biculture were separated in the laboratory. Residue samples were dried for 3 d at 65°C. After drying, samples were weighed and ground to pass through a 1-mm mesh. Cover crop samples were analyzed for total C and N with an automated Dumas instrument (LECO CHN-2000, LECO Corp., St. Joseph, MI).
Biomass decomposition and C and N residue release were analyzed by fitting a first-order exponential single-pool decay model to the data as described previously (Ruffo and Bollero, 2003). Biomass decomposition rate and C and N release rate were estimated with this model, with time either expressed as DGD with base temperature 0°C (Honeycutt and Potaro, 1990) or DCD (Stroo et al., 1989; Steiner et al., 1999).
Biochemical fractions were separated using the neutral detergent (ND) and acid detergent (AD) sequential procedure developed by Van Soest et al. (1991). Neutral detergent fiber and ADF fractions were determined, and the fractions soluble in ND (NDSOL) and AD (ADSOL) were estimated as the complement of the NDF and ADF fractions. All of these fractions (NDF, ADF, NDSOL, and ADSOL) were corrected for ash content. Ash concentration was determined by dry combustion of a 1.5-g sample at 650°C for 24 h.
In addition, the C and N concentrations were determined in the NDF (NDFC and NDFN) and ADF (ADFC and ADFN) fractions with an automated Dumas instrument (LECO CHN-2000, LECO Corp., St. Joseph, MI). As suggested by Van Soest et al. (1991), the N and C concentrations were analyzed in direct ADF (i.e., without predigesting with ND). The NDSOLC, NDSOLN, ADSOLC, and ADSOLN were calculated as:
 | [1] |
 | [2] |
 | [3] |
 | [4] |
Total C, N, NDF, and ADF are expressed in g kg-1 dry matter, and NDFC, NDFN, ADFC, ADFN, NDSOLC, NDSOLN, ADSOLC, and ADSOLN are expressed as g kg-1 NDFC, NDFN, ADFC, ADFN, NDSOL, or ADSOL. All of these variables were used in multivariate statistical analysis.
All variable means, standard error of the means, coefficient of variations (CVs), skewnesses (measure of the symmetry or asymmetry of the distribution), and kurtosis (measure of the flatness or peakness of the distribution) were calculated with the UNIVARIATE procedure of SAS (SAS Inst., 2000). The correlations among residue fractions were analyzed using the CORR procedure of SAS (SAS Inst., 2000). Principal-component analysis was performed on the correlation matrix of the residue fractions using the PRINCOMP procedure of SAS (SAS Inst., 2000). The correlation matrix was used because variables showed large differences in the ranges of their values as suggested by Khattree and Naik (2000). Each PC is a linear combination of the original standard variables having the eigenvectors (loading or contribution of each original variable to the PC) as coefficients. Principal components with eigenvalues (a measure of the variance of each PC) larger than 0.1 were retained as predictor variables for the regression analysis as suggested by Jolliffe (1986). The REG procedure of SAS (SAS Inst., 2000) with the stepwise option was used to find the best predictor of biomass decomposition rate and C and N release rates, with time expressed as either DGD or DCD, using previously obtained PCs as independent variables. Regression models were evaluated based on significant contributions, R2 values, and Mallows' Cp (Mallows, 1973) to ensure that the stepwise model included the least possible number of PCs.
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
Carbon concentration was stable between ND and AD fractions. It has been widely reported that total C concentration in plant tissue is very stable and close to 400 g kg-1 (Honeycutt et al., 1993; Kuo et al., 1997). Specifically, the constituents of the cell wall (NDF and ADF) are structural polysaccharides, which have a relatively stable C concentration (Chesson, 1997).
The difference in N concentrations between soluble and insoluble fractions along with their similar C concentrations produced an important variation of the C/N within these fractions. For example, in this study, the mean C/N of the WCC residue was 17. However, the C/N was 52 for NDF and only 11 for ND soluble. Henriksen and Breland (1999) reported that the C/N of NDSOLs of 10 different plant residues ranged from 8 [white clover (Trifolium repens L.) foliage] to 21 [white cabbage (Brassica oleracea L. var. capitata L. f. alba) leaves]. They also found that the C/N of the NDF was always larger (12- to 323-fold) than the C/N of the NDSOL. Furthermore, Kuo and Sainju (1998) reported an overall C/N of 10 and 23 for hairy vetch and rye residues. In the same study, the C/N of the NDSOL for the combined WCC residues was 11.5, which is in agreement with our study.
Correlations among biochemical fractions are presented in Table 2. Nitrogen and ADSOLN were the fractions with the largest number of highly significant (p < 0.0001) correlations. In contrast, NDFC, C, and ADSOLC showed the least number of highly significant correlations. Neutral detergent fiber and ADF were negatively correlated with all of the N fractions, and this is in agreement with Müller et al. (1988). The negative correlations between cell wall components and N concentration also agree with Chesson (1997), who suggested that there is an inverse relationship between N concentration and cell wall thickness, which is estimated by NDF and ADF fractions. Total N was positively correlated with NDFN and ADFN and with NDSOLN and ADSOLN. The correlations between total N and NDSOLN and ADSOLN were particularly high. Kuo and Sainju (1998) found an equally high correlation (r = 0.96) between total N and water soluble N concentration in hairy vetch and rye or ryegrass (Lolium multiflorum L.) residue mixtures. Because most of the N is present in the cell protoplast, which is soluble in ND and AD, strong correlations between N and these soluble fractions are expected.
|
The use of correlated variables in predictive statistical procedures such as multiple regression is conducive to multicollinearity. Multicollinearity leads to large variances of the regression parameters that yield unstable coefficients, making their interpretation and their use in predictive models difficult (Jolliffe, 1986). As an alternative, PCR is a predictive procedure that overcomes multicollinearity without reducing the number of variables considered. The first step for PCR is to obtain the PCs. Principal-component scores are uncorrelated random variables; thus, multicollinearity is avoided.
The biochemical fractions correlation matrix was used to calculate the PCs among the different fractions (Table 3). Principal-components 1 to 4 explained more than 90% of the total variance (PC1 = 48%, PC2 = 23%, PC3 = 14%, and PC4 = 7%). As suggested by Jolliffe (1986), only PCs with eigenvalues larger than 0.1 (PC1 to PC9) were retained to be used as predictors of biomass decomposition rates and C and N release rates. Principal-components 1, 4, 5, and 7 were selected by both criteria (stepwise and Mallows' Cp) for biomass decomposition rate and C and N release rates for both DGD and DCD. In addition, all the regressions were highly significant (p < 0.001), and the adjusted R2's were larger than 0.70 (Tables 4 and 5).
|
|
|
Principal-component 1 explained the largest proportion of the variability in decomposition and release rates (Tables 4 and 5). The large loadings of NDF and ADF in PC1 were consistent with the fact that NDF and ADF components (i.e., hemicellulose, cellulose, and lignin) had lower degradation rates than did soluble compounds (Paul and Clark, 1996; Kumar and Goh, 2000). Therefore, the larger the NDF and ADF concentration was in the residue, the lower the decomposition rate. Other authors reported that hemicellulose (a component of the NDF) had a negative effect on N release from plant material (Müller et al., 1988) while Janzen and Kucey (1988) found a negative correlation between cellulose and hemicellulose with CO2 evolved. On the other hand, N, NDSOLN, and ADSOLN had the highest loading rates among the positively loaded fractions. Other authors have reported similar effects of total N concentration and soluble N fractions on relative decomposition rate (Müller et al., 1988; Janzen and Kucey, 1988) and N mineralization from plant residues (Iritani and Arnold, 1960; Frankenberger and Abdelmagid, 1985; Kuo and Sainju, 1998; Trinsoutrot et al., 2000). It is generally accepted that residue decomposition is N limited; consequently, N-rich materials decompose faster than materials with low N concentration. However, the large loading of the soluble fractions suggests that the availability of N plays an important role too. Soluble fractions of N (NDSOLN and ADSOLN) are readily available to microorganisms and can stimulate microbial biomass growth and activity, affecting the residue decomposition at later stages (Reinertsen et al., 1984).
Principal-component 4 explained the second largest proportion of the variability in N release rates. In PC4, NDSOLN and ADSOLN contrast NDFN and NDFC. This contrast substantiated that N release rate is mostly controlled by the N availability in the soluble fractions rather than its total concentration in the residue (Kaboneka et al., 1997; Müller et al., 1988). In PC7, NDF and NDFN appeared with significant positive loadings. In the regression model, the negative coefficient associated with PC7 showed that NDF and NDFN fractions reduce biomass decomposition and C release rates. The interpretations of PC7 and PC1 are consistent with each other.
Principal-component regression explained biomass decomposition and C and N release rates, significantly weighing the cell wall (NDF and ADF) and the different N fractions (mainly NDFN and NSOLN). These results suggest that residue N availability is more critical in controlling biomass decomposition and C and N release than total N concentration or C/N. Total N concentration and C/N have failed to predict residue decomposition or N release in several studies when their ranges were small (Dendooven et al., 1997; Heal et al., 1997; Honeycutt et al., 1993; Gilmour et al., 1998).
Regression models of biomass decomposition and C and N release rates had higher adjusted R2 when regressed over DCD compared with DGD. Regression models using either DCD or DGD as independent variables selected the same predictor variables and with equal sign. Thus, the biological interpretation of the predictor variables is the same using DCD or DGD.
The algorithms obtained in this study (Tables 4 and 5) allowed the estimation of WCC biomass decomposition and C and N release, considering the effect of the environment (temperature and moisture) and residue biochemical fractions. With the results of the chemical analysis and calculation of the different C and N fractions, the score for each selected PC is calculated and then used in the multiple-regression model (Tables 4 and 5). Then, the DCD- or DGD-based rates can be estimated and used to calculate the amount of biomass remaining or the amount of N and C released at different times after WCC killing date. Also, the effect of weather on the dynamics of biomass decomposition and C and N release is accounted for with either DCD or DGD.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Principal-component regression allowed the prediction of biomass decomposition and C and N release rates based on residue biochemical fractions. Because these rates are based on time expressed as either DGD or DCD, the effect of temperature (DGD) or temperature and moisture (DCD) are also considered when the decomposition and C and N release rates are included in the decomposition model. The results of the chemical analysis (Van Soest et al., 1991) used as inputs can be obtained rapidly and economically because these methods are routinely performed in feed analysis worldwide. Although these decomposition models need to be validated in a wider range of environments, we believe that they have great potential to provide useful information for WCC management, breeding, and modeling purposes. In addition, this methodology has great potential to explain decomposition rates of other types of agricultural residues.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  M. A. Cavigelli, J. R. Teasdale, and A. E. Conklin Long-Term Agronomic Performance of Organic and Conventional Field Crops in the Mid-Atlantic Region Agron. J., May 7, 2008; 100(3): 785 - 794. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. B. Villamil, G. A. Bollero, R. G. Darmody, F. W. Simmons, and D. G. Bullock No-Till Corn/Soybean Systems Including Winter Cover Crops: Effects on Soil Properties Soil Sci. Soc. Am. J., September 20, 2006; 70(6): 1936 - 1944. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 F. E. Miguez and G. A. Bollero Winter Cover Crops in Illinois: Evaluation of Ecophysiological Characteristics of Corn Crop Sci., May 18, 2006; 46(4): 1536 - 1545. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. C. B. Dubeux Jr., L. E. Sollenberger, S. M. Interrante, J. M. B. Vendramini, and R. L. Stewart Jr. Litter Decomposition and Mineralization in Bahiagrass Pastures Managed at Different Intensities Crop Sci., April 25, 2006; 46(3): 1305 - 1310. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
F. E. Miguez and G. A. Bollero Review of Corn Yield Response under Winter Cover Cropping Systems Using Meta-Analytic Methods Crop Sci., September 23, 2005; 45(6): 2318 - 2329. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Crop Science | Vadose Zone Journal | |||
| Journal of Natural Resources and Life Sciences Education |
Soil Science Society of America Journal | ||||
| Journal of Plant Registrations | Journal of Environmental Quality |
The Plant Genome | |||
