**3. Assessment of compost maturity by spectroscopy techniques**

Various parameters are commonly used to evaluate compost quality. [9] In general, these parameters include germination index (GI), water-soluble organic carbon (WSOC), watersoluble organic nitrogen (WSON), pH, electrical conductivity (EC), moisture, and total organic matter (TOM) content. It is accepted that any sole parameter cannot determine compost maturity, which must be assessed by a combination of different physical, chemical, and biological properties (**Table 1**). However, all these approaches are expensive or time-consum‐ ing when a large number of samples are involved. [1,10] It is reported that spectroscopy techniques have many advantages over traditional chemical analyses, such as its ease of sample preparation, rapid spectrum acquisition, nondestructive analysis, and portability. [11]


**Figure 3.** Trough composting system (a) and windrow composting system (b) in China.

4 Organic Fertilizers - From Basic Concepts to Applied Outcomes

Therefore, the size of a windrow is determined by its porosity.

**3. Assessment of compost maturity by spectroscopy techniques**

In China, trough composting system and windrow composting system are the main compost‐ ing processes, with windrow composting system being more popular in Jiangsu Province (**Figure 3**). Windrow composting consists of placing the mixture of raw materials in long narrow piles that are agitated or turned on a regular basis. The turning mixes the materials during composting and enhances passive aeration. Generally, the heights of windrows are in a range of 90 to 180 cm. Correspondingly, the width of them varies in a range of 100 to 300 cm. In general, the size, shape, and spacing of the windrows are determined by the turning equipment. During aeration, the rate of air exchange depends on the porosity of the windrow.

Various parameters are commonly used to evaluate compost quality. [9] In general, these parameters include germination index (GI), water-soluble organic carbon (WSOC), water**Table 1.** Current criteria evaluated in the literature to characterize compost quality [9]

Near-infrared reflectance spectroscopy (NIRS) has been shown to rapidly (within 1 min) assess compost quality. [12–14] However, it is unclear whether NIRS can also be applied to rapidly determine the quality of commercial organic fertilizers, owing to the complexity and unpre‐ dictability of the raw materials of the commercial products. **Figure 4** shows a distinct appear‐ ance of commercial organic fertilizers with composting samples. A total of 104 commercial organic fertilizers were collected from full-scale compost factories in Jiangsu Province, China. These factories treat organic matter from animal manure and other agricultural organic residues. These factories produce approximately 5000–150,000 tons of commercial organic fertilizers per year.

**Figure 4.** Typical commercial organic fertilizers, including powered (A, C) and granular (B, D) fertilizers. These photos suggest that the commercial organic fertilizers are more even than samples from the composting process.

We can see that all the NIR spectra collected from commercial organic fertilizers in Jiangsu Province, China (**Figure 5**) were divided into two groups of signals with different slopes under 1400 nm: one group has an increased curvature with a significant absorbance peak at a wavelength of approximately 1420 nm, while another is more flat and has only a small absorption at this position. This is because the second significant spectral peak is mainly at approximately 1950 nm (**Figure 5**). The band at 1420 nm is associated with the O–H and aliphatic C–H, while that at 1950 nm is assigned to the amide N–H and O–H. Because the NIR spectrum contains all strength information of the chemical bond, chemical composition, electronegativity, etc, the absorption peaks are heavily overlapped. In addition, other inter‐ ference information, such as scattering, diffusion, special reflection, refractive index, and reflected light polarization, also has an important influence on the NIR spectrum. Thus, the quantitative predictions are difficult directly through NIR spectra alone.

**Figure 5.** Spectra of NIR of a total of 104 commercial organic fertilizers. [10]

organic fertilizers were collected from full-scale compost factories in Jiangsu Province, China. These factories treat organic matter from animal manure and other agricultural organic residues. These factories produce approximately 5000–150,000 tons of commercial organic

**Figure 4.** Typical commercial organic fertilizers, including powered (A, C) and granular (B, D) fertilizers. These photos

We can see that all the NIR spectra collected from commercial organic fertilizers in Jiangsu Province, China (**Figure 5**) were divided into two groups of signals with different slopes under 1400 nm: one group has an increased curvature with a significant absorbance peak at a wavelength of approximately 1420 nm, while another is more flat and has only a small absorption at this position. This is because the second significant spectral peak is mainly at approximately 1950 nm (**Figure 5**). The band at 1420 nm is associated with the O–H and aliphatic C–H, while that at 1950 nm is assigned to the amide N–H and O–H. Because the NIR spectrum contains all strength information of the chemical bond, chemical composition, electronegativity, etc, the absorption peaks are heavily overlapped. In addition, other inter‐ ference information, such as scattering, diffusion, special reflection, refractive index, and

suggest that the commercial organic fertilizers are more even than samples from the composting process.

fertilizers per year.

6 Organic Fertilizers - From Basic Concepts to Applied Outcomes

Multivariate analyses are required to discern the spectral characteristics of commercial organic fertilizers with the support of chemometric methods, for example, partial least squares (PLS) analysis in this study. The results of the NIRS calibration and validation for the quality indices of commercial organic fertilizers are listed in **Table 2** and Figures 6 and 7. The NIR calibrations allowed accurate predictions of WSON, TOM, pH, and GI (*R*<sup>2</sup> = 0.73−0.93 and RPD = 1.47−2.96). However, the results were less accurate for moisture (*R*<sup>2</sup> = 0.91, *r*<sup>2</sup> = 0.79, RPD = 2.22), TN (*R*<sup>2</sup> = 0.98, *r*<sup>2</sup> = 0.80, RPD = 2.25), and EC (*R*<sup>2</sup> = 0.99, *r*<sup>2</sup> = 0.74, RPD = 2.27). In addition, the WSOC had the worst prediction (*R*<sup>2</sup> = 0.88, *r*<sup>2</sup> = 0.76, RPD = 2.10). Therefore, predictions were moderately successful for TOM, TN, WSON, pH, EC, GI, and moisture, but failed for WSOC.

**Figure 6.** The measured and predicted values of the quality indices of commercial organic fertilizers in the calibration data set. [10] The best fit is shown by red line.

**Figure 7.** The measured and predicted values of the quality indices of commercial organic fertilizers in the prediction data set. [10] The best fit is shown by red line.


TOM, total organic matter; TN, total nitrogen; WSOC, water-soluble organic carbon; WSON, water-soluble organic nitrogen; EC, electrical conductivity; GI, germination index; PC, number of principal components; *R*<sup>2</sup> , the coefficient of determination for the calibration set; RMSECV, the root mean squared error in cross-validation; *r*<sup>2</sup> , the coefficient of determination for the validation set; RMSEP, room mean squared error of prediction; RPD, the ratio of the standard deviation in the validation set over the room mean squared error of prediction.

**Table 2.** NIRS calibration and validation results for quality indices of commercial organic fertilizers [10]

Similar to NIRS, fluorescence excitation–emission matrix (EEM) spectroscopy is extensively utilized to detect protein-like, fulvic-acid-like, and humic-acid-like substances. These materials are directly proportional to fluorescence intensity at low concentrations and thus are applied to assess compost maturity [15]. Fluorescence spectroscopy has been widely used as a tool to assess compost maturity, owing to high instrumental sensitivity. [16] However, analysis of fluorescence EEM has generally been limited to visual identification of peaks or development of ratios of fluorescence intensities in different regions of the spectrum. These techniques lack the ability to capture the heterogeneity of samples. It has been reported that as opposed to individual main peak positions analysis, analyzing the full fluorescence EEMs can provide much information. [17] Additionally, the composition complexity of WEOM in compost samples often results in the overlapped fluorophores in the EEM spectra. As a result, the EEM spectra are difficult to interpret.

Recent work has demonstrated that parallel factor (PARAFAC) analysis can be used to decompose full fluorescence EEMs into different independent groups of fluorescent compo‐ nents. [18] Therefore, EEM-PARAFAC analysis is able to assess compost maturity and also to be a potential monitoring tool for rapidly characterizing compost maturity. For this purpose, 62 full-scale compost facilities in nine Provinces of China, yielding compost from animal manures and other industrial organic residues of different maturities, were selected. Then, these compost samples were used to extract water-soluble OM (WEOM) and characterized by fluorescence EEM spectra. EEM-PARAFAC analysis was then conducted for assessment of compost maturity.

**Figure 8** shows the typical fluorescence EEM contours of WEOM. The EEM contours of WEOM between mature and immature composts were obviously distinct. Specifically, the former had only two peaks (A and B), while the latter exhibited four peaks (A, B, C, and D) (**Figure 8**).

**Figure 7.** The measured and predicted values of the quality indices of commercial organic fertilizers in the prediction

Moisture content (%) 19 0.94 0.23 0.67 0.36 2.27 −0.03 0.76 TOM (g/kg) 12 0.85 17.69 0.69 24.76 1.71 −9.03 0.72 TN (g/kg) 17 0.98 3.08 0.80 3.63 2.25 −0.29 0.77 WSOC (g/kg) 7 0.85 3.82 0.55 3.21 1.43 0.06 0.69 WSON (g/kg) 8 0.86 1.81 0.77 1.60 1.47 0.05 0.74 pH 9 0.86 0.36 0.75 0.41 1.96 −0.10 0.73 EC (mS/cm) 17 0.99 0.46 0.74 0.54 2.27 −0.04 0.78 GI (%) 18 0.84 4.81 0.68 9.52 1.73 −2.68 0.63 TOM, total organic matter; TN, total nitrogen; WSOC, water-soluble organic carbon; WSON, water-soluble organic

**PC** *R***<sup>2</sup> RMSECV** *r***<sup>2</sup> RMSEP RPD Bias Slope**

, the coefficient of

, the coefficient of

**Parameters Calibration set Validation set**

nitrogen; EC, electrical conductivity; GI, germination index; PC, number of principal components; *R*<sup>2</sup>

**Table 2.** NIRS calibration and validation results for quality indices of commercial organic fertilizers [10]

determination for the validation set; RMSEP, room mean squared error of prediction; RPD, the ratio of the standard

Similar to NIRS, fluorescence excitation–emission matrix (EEM) spectroscopy is extensively utilized to detect protein-like, fulvic-acid-like, and humic-acid-like substances. These materials

determination for the calibration set; RMSECV, the root mean squared error in cross-validation; *r*<sup>2</sup>

deviation in the validation set over the room mean squared error of prediction.

data set. [10] The best fit is shown by red line.

8 Organic Fertilizers - From Basic Concepts to Applied Outcomes

**Figure 8.** Typical EEM contours of mature (a) and immature (b) composts, [1] normalized to 1 mg DOC/L. Note: The arrows refer to the location of peaks. Specifically, two peaks at Ex/Em of 230/420 (B) and 330/420 (A) were presented in mature composts, whereas four peaks at Ex/Em of 220/340 (D), 280/340 (C), 220/410 (B), and 330/410 (A) were detected in the immature composts.

The fluorescence EEM contours were interpreted using the protocol of the previous literature. [17] The typical mature compost contained humic-acid-like substances (i.e., Peak A: Ex/Em = 330/420) and fulvic-acid-like substances (i.e., Peak B: Ex/Em = 230/420) in the WEOM of compost. The typical immature WEOM of compost, however, in addition to containing humic acid–like substances (Peak A) and fulvic acid–like substances (Peak B), also contained trypto‐ phan-like substances (i.e., Peak D: Ex/Em = 220/340) and a few soluble microbial by-product– like (SMP) substances (i.e., Peak C: Ex/Em = 280/340). The above results were consistent with those of Marhuenda-Egea et al. in that the composting process was a degradation of the original tyrosine-like and tryptophan-like materials and an increase in the humic-like and fulvic-like materials. [19] As composting time increased, molecular heterogeneity decreased, while aromatic polycondensation, level of conjugated chromophores, and humification degree increased, as shown by fluorescence EEM spectra. [20] Therefore, fluorescence EEMs of the WEOM from composts had the potential to be used as a monitoring tool for rapidly assessing the maturity of compost.

Preprocessing of the fluorescence EEMs is essential to acquire a correct component number before PARAFAC analysis. After removing the Rayleigh and Raman scatters, the sum of squared residuals in the excitation (Ex) and emission (Em) directions for three different models are plotted in **Figure 9a**. By comparing the two-, three-, and four-component models, it is clear that the step from the two- to three-component model showed great improvement of fit, whereas the step from the three- to four-component model offered only some enhancement of fit. This result suggests that three components are adequate for this data. Split-half analysis further validated the close-to-perfect correspondence between the Ex and Em loadings for the three components, using the four independent random halves (**Figure 9b**). Thus, three components were suitable for all of the examined samples. The EEM contours are shown in

**Figure 9.** PARAFAC analysis of 60 compost samples. (a) Sum of squared error for determining the component num‐ bers. (b) Split-half analysis for model validation. (c) EEM contours of identified components. [1]

**Figure 9c**. All the EEMs in this study could be decomposed into a three-component model by PARAFAC analysis, with component 1 [Ex/Em = (230, 330)/410], component 2 [Ex/Em = (250, 350)/450], and component 3 [Ex/Em = (220, 280)/340]. The three components belonged to humiclike, fulvic-like, and protein-like substances, respectively.

Clearly, component 1 was humic fluorophores derived from both terrestrial-like and marinelike, and component 2 was terrestrial-like humic fluorophores. In contrast, component 3 was a tryptophan-like substance. In addition, all the components displayed the same Em wave‐ length at different Ex wavelengths, possibly attributable to the rapid internal conversion of excited electrons to the lowest vibration level of the first excited state. **Table 3** shows that component 1 was strongly correlated with components 2 and 3, whereas components 2 and 3 had no significant correlation, indicating that component 1 has a common source with components 2 and 3, but components 2 and 3 do not have the same source.


a Correlation is significant at the 0.05 level (two-tailed).

The fluorescence EEM contours were interpreted using the protocol of the previous literature. [17] The typical mature compost contained humic-acid-like substances (i.e., Peak A: Ex/Em = 330/420) and fulvic-acid-like substances (i.e., Peak B: Ex/Em = 230/420) in the WEOM of compost. The typical immature WEOM of compost, however, in addition to containing humic acid–like substances (Peak A) and fulvic acid–like substances (Peak B), also contained trypto‐ phan-like substances (i.e., Peak D: Ex/Em = 220/340) and a few soluble microbial by-product– like (SMP) substances (i.e., Peak C: Ex/Em = 280/340). The above results were consistent with those of Marhuenda-Egea et al. in that the composting process was a degradation of the original tyrosine-like and tryptophan-like materials and an increase in the humic-like and fulvic-like materials. [19] As composting time increased, molecular heterogeneity decreased, while aromatic polycondensation, level of conjugated chromophores, and humification degree increased, as shown by fluorescence EEM spectra. [20] Therefore, fluorescence EEMs of the WEOM from composts had the potential to be used as a monitoring tool for rapidly assessing

Preprocessing of the fluorescence EEMs is essential to acquire a correct component number before PARAFAC analysis. After removing the Rayleigh and Raman scatters, the sum of squared residuals in the excitation (Ex) and emission (Em) directions for three different models are plotted in **Figure 9a**. By comparing the two-, three-, and four-component models, it is clear that the step from the two- to three-component model showed great improvement of fit, whereas the step from the three- to four-component model offered only some enhancement of fit. This result suggests that three components are adequate for this data. Split-half analysis further validated the close-to-perfect correspondence between the Ex and Em loadings for the three components, using the four independent random halves (**Figure 9b**). Thus, three components were suitable for all of the examined samples. The EEM contours are shown in

**Figure 9.** PARAFAC analysis of 60 compost samples. (a) Sum of squared error for determining the component num‐

bers. (b) Split-half analysis for model validation. (c) EEM contours of identified components. [1]

the maturity of compost.

10 Organic Fertilizers - From Basic Concepts to Applied Outcomes

b Correlation is significant at the 0.01 level (two-tailed).

c Correlation is significant at the 0.001 level (two-tailed). OM, organic matter; MI, mineralization index; GI, germination index; OUR, oxygen uptake rate; CER, CO2 evolution rate.

**Table 3.** Pearson correlation between log (scores) of PARAFAC components and selected stability indices (*n* = 60) [1]

In addition, components 1 and 3 had a strong correlation (*R* > 0.35, *p* < 0.01) with TOC, OM, CER, GI, and OUR. But component 2 was only significantly (*R* > 0.37, *p* < 0.01) correlated with TOC and CER, and weakly (*R* = 0.30, *p* < 0.05) correlated with OUR (**Table 4**). Therefore, components 1 and 3 are more suitable to assess compost maturity than component 2. Further‐ more, the regression equations in **Table 4** indicated that the composts could be identified as mature when the log scores of components 1 and 3 were higher than 3.69 ± 0.06 and 3.49 ± 0.09, respectively.


TOC, total organic matter; OM, organic matter; GI, germination index; OUR, oxygen uptake rate; CER, CO2 evolution rate.

**Table 4.** Calculated log (scores) of components from regression equations (*n* = 60) [1]

Some investigators have shown great interest in the assessment of compost maturity using fluorescence intensity. Our results demonstrate that log scores of components 1 and 3, identified by the DOMFluor–PARAFAC approach, can be applied to assess the maturity of composts which cover a large range of waste sources. This is attributable to the fact that the full fluorescence spectroscopy analysis provides a basis for capturing subtle changes in the fluorescence spectra. In summary, the assessment of compost maturity by the DOMFluor– PARAFAC approach is not wastesource–specific.
