Characterizing Humic Substances from Native Halophyte Soils by Fluorescence Spectroscopy Combined with Parallel Factor Analysis and Canonical Correlation Analysis

Soil is one of the principal substrates of human life and can serve as a reservoir of water and nutrients. Humic substances, indicators of soil fertility, are dominant in soil organic matter. However, soil degradation has been occurring all over the world, usually by soil salinization. Sustainable soil productivity has become an urgent problem to be solved. In this study, fluorescence excitation-emission matrices integrated with parallel factor analysis (PARAFAC) and canonical correlation analysis (CCA) were applied to characterize the components of fulvic acid (FA) and humic acid (HA) substances extracted from soils from the Liaohe River Delta, China. Along the saline gradient, soil samples with four disparate depths were gathered from four aboriginal halophyte communities, i.e., the Suaeda salsa Community (SSC), Chenopodium album Community (CAC), Phragmites australis Community (PAC), and Artemisia selengensis Community (ASC). Six components (C1 to C6) were identified in the FA and HA substances. The FA dominant fractions accounted for an average of 45.81% of the samples, whereas the HA dominant fractions accounted for an average of 42.72%. Mature levels of the HA fractions were higher than those of the FA fractions, so was the condensation degree, microbial activity, and humification degree of the FA fractions. C1 was associated with the ultraviolet FA, C2 was referred to as visible FA, C3 and C4 were relative to ultraviolet HA, C5 represented microbial humic-like substances (MH), and C6 referred to visible HA. C1, C2, C5 and C6 were latent factors of the FA fractions, determined using the CCA method and could possibly be used to differentiate among the SSC, CAC, PAC and ASC samples. C3, C4, C6 and C5 were latent factors of the HA fractions, which might be able to distinguish the ASC samples from the SSC, CAC and PAC samples. Fluorescence spectroscopy combined with the PARAFAC and CCA is a practical technique that is applied to assess the humic substance content of salinized soils.


Introduction
Humic substances (HS), as typical natural supramolecular components, are predominantly stabilized by faint dispersion and originate from biodegradation [1][2][3]. In mineral soils, the HS can constitute 70-80% of organic carbon [4][5][6]. They have a relatively high amount of lignin, originating from plant debris and organic materials, which is prone to be accompanied by fairly high humification The soil samples were hermetically poured into packages, stored in an incubator, and transported to the laboratory as soon as possible.

Physico-Chemical Analyses
A portion of each soil sample was air-dried, crushed and sieved (Φ = 100), while the remaining moist soil samples were analyzed directly. The moist soil samples were analyzed for electric conductivity (EC) and pH, while the air-dried samples were analyzed for the total organic carbon (TOC). The EC was surveyed in the slop (soil/water = 1:1) utilizing an electronic conductivity meter (FE30, Mettler Toledo, Switzerland), and the pH was assayed in the solution (water/soil = 2.5:1) utilizing a pH meter (Sartorius, Germany). A Shimadzu V-CPH analyzer (Japan) was used to measure the TOC [25,26].

Isolation of FA and HA Fractions
The isolation of the HS from the soils was performed according to the methods approved by the International Humic Substances Society [31][32][33]. In detail, a 20 g air-dried, crushed and sieved sample was added to 100 mL of aqueous solution containing 0.1 M Na 2 P 2 O 7 and 0.1 M NaOH (pH = 13). The suspension was sealed and stirred continuously at 150 rpm for 24 h protected by purging with nitrogen. Then the suspension was centrifuged at 3000 rpm for 15 min. The supernatant was further filtered through the 0.7 µm glass fiber membrane to obtain the FA and HA fractions.
The FA and HA fractions were separated with 0.5 M H 2 SO 4 aqueous solution. FA and HA appeared in the supernatant and deposition, respectively. The supernatant was centrifuged at 3000 rpm for 15 min to obtain the FA. 100 mL of the 0.1 M NaOH was added in the deposition to obtain the HA. The FA and HA fractions were further filtrated with activated carbon, then the pH of the solutions was adjusted to 8.0 (±0.05) to avoid the precipitation of the HA fraction which might disturb the further analysis [34,35].

EEM Spectroscopy Detection and PARAFAC
A Fluorescence Spectrophotometer (F-7000, Hitachi, Japan) was used to obtain the EEM spectrum. PMT voltage was set at 700 V. Emission (em) wavelengths were conducted from 260 to 550 nm at 5 nm intervals, while excitation (ex) wavelengths were conducted from 240 to 450 nm. The scanning speed was set at 2400 nm·min −1 and the response time was 0.5 s. The Rayleigh scattering and the upper spectral data were defined as 0, and the water Raman scattering was restrained by using Quinine-sulfate [33,34].
The sum of the 64 EEM spectra (the FA = 32 and the HA = 32) was modeled with the PARAFAC by the DOMFluor toolbox 1.7 in MATLAB platform [22]. The number of fluorescence components was determined using split-half and residual analysis. Six principal components (C1 to C6) were extracted from the EEM spectrum, whose maximum fluorescence intensities (F max ) are representative of the proportional abundances, respectively [22]. The F max of each component was measured as a percentage of the total F max for the six components (%C1 to C6%), which was concerned with the relative abundance.

Multivariable Analyses
The variations of the FA and HA fractions in the different soil profiles were investigated using the matrices of the fluorescent components. The fluorescence indices (Fis)-inversed from the EEM spectra and fluorescent components-were built using the column and Y error plot, and ternary plot. The similarities/dissimilarities of the sampling sites were detected using non-metric multidimensional scaling (nMDS) analysis. The correlations between the fluorescent components and Fis were revealed using bivariate correlation analysis, and the potential factors were sought using canonical correlation analysis (CCA). Statistical analyses were conducted on SPSS 22.0 (SPSS Inc., version 22.0, USA), Origin 8.0 (Microcal, USA) and Canoco 4.5 (Microcomputer Power, USA).

Physico-Chemical Characteristics
The soil EC, pH and TOC were measured in our previous studies on the same samples [26]. A noticeable difference in the TOC can be observed in the different aboriginal halophyte soils along with each soil profile. The average TOC value in the ASC soil profile was 11.36 ± 1.14 mg·kg −1 , approximately 2.69 times greater than that in the SSC soil profile, 2.04 times greater than that in the CAC soil profile, and 1.49 times greater than that in the PAC soil profile. An evident distinction can be found in the TOC in each soil profile, but there is no similar trend found according to the depth ( Figure 2).

EEM Spectroscopy Characteristics
All the samples of the FA fraction show the similar EEMs-two distinct peaks and one weak shoulder (Figure 3a). Peak A situates within the λ Ex from 240 to 280 nm and the λ Em from 420 to 480 nm, could be attributed to the ultraviolet FA substances [35]. Peak C situates at the λ Ex of 320-360 nm and the λ Em of 400-460 nm, could be ascribed to the visible FA substances [36]. Peak M is situated between the peaks A and C, could be viewed as a microbial humic-like substance (MH), which indicates microbial activity [37]. All of the HA fraction samples show the similar EEM spectra too, in which two strong peaks and one weak shoulder coexist (Figure 3b). Peak D (λ Ex/Em = 270-310/450-510 nm) can be ascribed to the ultraviolet HA, and peak H (λ Ex/Em = 360-400/450-510 nm) implies the visible HA exists [36,38]. There was a weak shoulder, known as the peak M, which is believed to be MH.

Assignments and Variations of Fluorescent Components
The PARAFAC was applied to analyze the model EEM spectra matrices of the FA and HA fractions and produce six different spectral components ( Figure 4). Component one (C1) with the wide and long wavelength peak, resembles the identified peak C, which is associated with the visible FA. Component two (C2) with the wide and short wavelength peak, resembles the identified peak A, which is associated with the ultraviolet FA. The locations of component three (C3) and component four (C4) are similar to the identified peak D, which corresponds to the ultraviolet HA [39]. The location of component six (C6) is similar to the identified peak H, which is relative to the visible HA. Component five (C5) is a ubiquitous fluorescent signal known as the identified M (Figure 4e), which is related to the MH [18].
Concerning the four soil profiles, the average F max of each fluorescent component obtained from the FA fraction is larger than that of the corresponding component of the HA fraction (Figure 5a), especially for C1, whose average F max of the FA fraction (588.56 ± 64.65) is 2.74 times greater than that of the HA fraction. This indicates that the fluorescent components in the FA fraction were much more concentrated than those in the HA fraction, as regards the whole soil profiles. The descending sequence of the mean F max of the components of the FA fraction is C2 > C3 > C1 > C4 > C5 > C6, while the descending sequence of the mean F max of the components of the HA fraction is C2 > C3 > C4 > C1 > C6 > C5. The average %C1 of the FA fraction is higher than that of the HA fraction, as well as the %C2, %C3 and %C5 (Figure 5b). The average %C4 of the FA fraction is lower than that of the HA fraction, similar to the %C6. These indicate that the content of the FA substances in the FA fraction is higher than that in the HF fraction, while the content of the HA substances within the former is lower than that in the latter.  For the FA fraction, the mean F max C1 of the ASC soil profile (869.61 ± 163.12) is the highest, followed by the PAC (683.25 ± 163.60), CAC (425.59 ± 130.66) and SSC (375.81 ± 148.44) (Figure 6a). The trends of the average F max C2 and C5 are the same as the average F max C1. The average F max C3 of the ASC soil profile (1017.60 ± 132.57) is the highest, followed by the PAC (786.20 ± 112.01), SSC (641.04 ± 143.09) and CAC (615. 42 ± 192.43). The trends of the average F max C4 and C6 are similar to the average F max C3. The variation of each component is distinct in soil profiles, but there is no trend found in the vertical section. Noticeably the F max of each component in the 0-20 cm soil layer is much higher than those in the other soil layers of the SSC soil profile, while the F max of each component in the 40-60 cm soil layer is higher than those in the other soil layers of the PCA soil profile. Further, the F max of a given component in the 60-80 cm soil layer in the CAC soil profile is higher than those in the other soil layers, this was also the case for the ASC soil profile.
With the HA fraction, the descending sequence of the mean F max C1 in the soil profiles is PAC (311.66 ± 261. 33  ). Evidently, the %C2 is higher than those of the other components, suggesting that C2 is the representative component of the FA fraction. As regards to the HA fraction, the variation range (0.53-38.25%) for the %C5 is the largest, followed by the %C1 (8.77-32.01%), %C6 (3.13-23.77%), %C4 (9.07-25.20%), %C3 (11.91-25.05%) and %C2 (14.16-28.85%). Markedly, the variation ranges for the components in the HA fraction are much larger than those in the FA fraction.
The %FA (C1 + C2) in the FA fraction is approximately 45.81%, about 1.12 times greater than the %HA (C3 + C4 + C6), while the %FA in the HA fraction is 39.38%, 1.28 times less than the %HA. The FA is the main component in the FA fraction, and the HA is the major component in the HA fraction. This is consistent with a redshift in the fluorescent maxima of the HA fraction contrast with the FA fraction ( Figure 3). The redshift in the fluorescent maxima could be attributed to the existence of high molecular weight fractions, electron-withdrawing substitutes, and a higher conjugation level [1,40]. The %C5 (%MH) is 12.88% in the FA fraction, about 1.24 times higher than that in the HA fraction, indicating that the former had more active microbial sources than the latter.
The variations of the %MH, %FA and %HA in the HA fraction are consistently larger than those for the FA fractions (Figure 7). For the FA fraction, the %FA stays relatively stable (41.17-48.16%) within the whole soil profiles (Figure 7a

Deduction of Fluorescence Indices
It has been reported that the quotient of the fluorescence intensities at the Ex/Em 370/450 nm and Ex/Em 370/500 nm (f 450/500) is fluorescent index (FI) which indicates the organic matter sources [41,42]. The f 450/500 value less than 1.40 indicates the main contribution from allochthonous sources (terrestrial-derived organic matter), larger than 1.90 indicates the main contribution of autochthonous sources (microbial-derived organic matter), and between 1.40 and 1.90 indicates the main contribution of both allochthonous and autochthonous sources, i.e., mixture sources [43,44]. The average f 450/500 values of the HA fraction are 1.31 ± 0.12 proving that it is mainly derived from the allochthonous sources, while the mean of the FA fraction is 1.49 ± 0.20 proving that it mainly derived from the mixture sources. For the FA fraction, the descending sequence of the mean f 450/500 values is CAC > SSC > ASC > PAC, and its values stay relatively consistent within each soil profile except for the PAC soil profile (Figure 8a). For the HA fraction, the descending sequence of the mean f 450/500 is ASC > PAC > CAC > SSC, and its values maintain relatively stable too (Figure 8b). A ratio between the C2 and C1 (peak C: peak A, C/A), is well known as a practical FI to indicate the mature level of the organic matter which is reinforced after a rise in the C/A [1]. The average C/A value of the FA fraction (0.57 ± 0.13) is lower than that of the HA fraction (0.69 ± 0.45), indicating that the mature level of the HA fraction is higher than that of the FA fraction. The decreasing order of the average C/A values in the FA fraction is ASC > PAC > CAC > SSC (Figure 8c), while the decreasing order in the HA fraction is PAC > CAC > SSC > ASC (Figure 8d). A ratio between the C4 and the C3 (C4/C3) is defined as an FI to indicate the condensation degree of organic matter, which increases with the increasing of the C4/C3 [33]. The C4/C3 mean of the FA fraction is 0.58 ± 0.09, approximately 1.44 times less than that of the HA fractions. This suggested that the condensation degree of the FA fraction is lower than that of the HA fraction. The increasing order of the C4/C3 means in the FA fraction is PAC < CAC < SSC <ASC (Figure 8e), while the increasing order in the HA fraction is CAC < PAC < ASC < SSC (Figure 8f). A ratio between the C5 and the C1 + C2 (microbial humic-like: fulvic acid, M/F) is well known as an FI to indicate microbial activity, and it increases with a rise in M/F [45]. The M/F mean of the FA fraction is 0.28 ± 0.07, approximately 1.59 times less than that of the HA fraction. This shows that the microbial activity in the FA fraction is lower than that in the HA fraction. For the FA fraction, the decreasing order of the M/F means is CAC > ASC > SSC > PAC, and the M/F variations in the CAC and PAC soil profiles are larger than those within the SSC and ASC soil profiles (Figure 8g). For the HA fraction, the decreasing order of the M/F means is ASC > CAC > SSC > PAC, and there is an obvious variation of the M/F in each soil profile, but no trend can be found in the vertical direction (Figure 8h). A ratio among the C5 with the C3 + C4 + C6 (microbial humic-like: humic acid, M/H), as the M/H, is calculated as an FI to indicate the microbial activity, and it increases with a rise in M/H too [45]. The M/H mean of the FA fraction is 0.32 ± 0.13, approximately 1.68 times less than that of the HA fraction. This identified that the microbial activity in the FA fraction is lower than that in the HA fraction. For the FA and HA fractions, the M/H trends are similar to the M/F (Figure 8i,j). Moreover, a ratio between the C3 + C4 + C6 and C1 + C2 (humic acid: fulvic acid, H/F) has been widely utilized to assess the humification degree of organic matter, and it increases according to an ascent in H/F [35]. The H/F means of the HA fraction (1.10 ± 0.44) is higher than that of the FA fraction (0.90 ± 0.13), indicating that the humification degree of the HA fraction is high compared with that of the FA fraction. For the FA fraction, the H/F means within the SSC and ASC soil profiles are higher than those in the CAC and PAC soil profiles (Figure 8k). For the HA fraction, the H/F mean in the SSC soil profile is the highest, followed by the CAC, PAC and ASC soil profiles (Figure 8l).

nMDS Analyses
The nMDS was used to analyze the FIs and the fluorescent components of the FA and HA fractions to reveal similarities/dissimilarities between 32 soil samples. The nMDS produces a two-dimensional matrix with stress = 0.13 and R 2 = 0.94, suggesting the reliability of the two-dimensional representation [24]. The nMDS map exhibited that the soil samples in the delta are clustered into two groups with the linear equation y = −x, i.e., the FA fraction and HA fraction (Figure 9). This results indirectly show that the FA fraction is significantly different from the HA fraction. The sample desperation of the FA fraction is lower than that of the HA fraction. Therefore, the FA and HA fractions should be analyzed and discussed in a standalone manner.

CCA Analyses
The CCA provides an observable and graphic comprehension of the relationships among the FIs, fluorescent components, and samples. Regarding the FA fractions, the f 450/500, M/F and M/H moved toward the positive direction of the AX1 in the CCA ordination biplot (Figure 11a), while the C/A, C4/C3, H/F and C1 to C6 moved towards the negative direction of the AX1. This indirectly proves that the former showed negative correlations with the latter. The CAC samples remain in the first quadrant, the PAC remains in the second quadrant, the ASC remains in the third quadrant, and the SSC remains in the fourth quadrant. The above results demonstrate that differences in the characteristics of the FA fractions occur among the SSC, CAC, PAC and ASC soil profiles. Based on the law of species and specimens [46], C1, C2, C5 and C6 are the latent factors of the FA fractions, proving that FA is the major component in the FA fractions. The latent factors could differentiate among the SSC, CAC, PAC, ASC samples.  (Figure 11b), while the arrows of the C/A, C4/C3, H/F and C1 to C6 point to the negative direction of the AX1. This suggests that there should be negative correlations between the former and the latter. The samples in the SSC, CAC and PAC soil profiles (except for the SSC4, CAC2 and PAC2) remain in the second and third quadrants, while the samples within the ASC soil profiles remain in the first and fourth quadrants. This results prove that the characteristics of the HA fractions in the SSC soil profiles are different from the other three soil profiles. Based on the law of species and specimens [46], the C3, C4, C6 and C5 are the latent factors of the HA fractions, proving that HA is the representative component in the HA fraction. The latent factors distinguish the ASC samples from the SSC, CAC and PAC samples.

Conclusions
The FA and HA fractions extracted for the Liaohe river delta soils are identified as six fluorescent components: C1 is defined as the ultraviolet FA, C2 is the visible FA, C3 and C4 are the ultraviolet HA, C5 is the MH, and C6 is the visible HA. The obvious variations of fluorescent components for the FA and HA fractions are represented in each soil profile, but there is no trend found in the vertical direction. The variations of C1 + C2 (FA), C3 + C4 + C6 (HA), and C5 in the HA fractions are much larger than those in the FA fractions. FA is the dominant component in the FA fractions, whereas HA is the predominant component in the HA fractions. The HA fractions are mainly derived from the allochthonous sources, while the FA fractions are mainly derived from the mixture sources. The mature level of the HA fractions is higher than that of the FA fractions, this was also the case for the condensation degree, microbial activity and humification degree. C1, C2, C5 and C6 are the latent factors of the FA fractions, which are distinct from the SSC, CAC, PAC, ASC samples. The C3, C4, C6 and C5 are the latent factors of the HA fractions, which are different to the ASC samples from the SSC, CAC and PAC samples. The humic substances from soils with different salinized-vegetation improve the buffering potential of the soil and prevent soil from salinization, which might be conducive to the sustainable development of agriculture.