3.7.Redundancy discriminatory analysis
The CANOCO software was used to analyze the data in relation to carbon sources and environmental factors. The maximum gradient length in the four axes was 0.464, with a value below 3. Therefore, linear model analysis (RDA) was selected. The eigenvalues of the first two ordination axes (carbon source diversity an environmental factors) were 0.573 and 0.213 (Table 4), while the correlation coefficients of the two ordination axes were 1. The first two ordination eigenvalues accounted for 99.8% of the total eigenvalues. The correlation coefficients between the first two ranking axes and environmental factors were extremely high, accounting for 99.9% of the total variance. The axes RDA1 and RDA2 explained 46.4 and 34.9% of the variation among intercropping system communities, respectively (Table 5).
The greatest differences between treatment communities were observed when comparing mulberry and alfalfa and related to differences in pH, OM, SWC and AN, according to axis length and angle. The second greatest differences were observed between ANE and AN0 and were due to the differences in the activities of PPO, CAT, and POD (Fig. 6). Moreover, POD, CAT, and OPP were positively related to A0 and AN0, while pH, AN, and SWC were negatively related to AE and ANE. The third greatest differences were observed between ME and MN0, and between MNE and M0, which were related to differences in SWC, AN, SUR, POD, pH, and OM, respectively. The parameters SUR, AN, and SWC were positively related with M0, while pH, OM, and SWC were positively related to M0 and MNE.