2.5. Statistical analyses
All soil spectra were smoothed and normalized before spectral analysis
in MATLAB R2020b (The Math Works, Natick, USA). A one-way analysis of
variance (ANOVA) was used to analyze the effects of various treatments
on grain yield, NAE, changes in soil properties, soil
NH4+-N
and NO3−-N contents, and plant N
content. Differences between treatments were determined by comparing
their means using the least significant difference (LSD) at the 0.05
probability level. A two-way ANOVA was applied to evaluate the main and
interactive effects of crop rotation and fertilization on grain yield,
NAE, changes in soil properties, soil
NH4+-N and
NO3−-N contents, and plant N content.
A pairwise samples test was conducted to compare the significant changes
in soil properties and characteristic spectral bands after various
treatments. Principal component analysis (PCA) was performed to
illustrate the internal structure of spectra. A two-way permutational
multivariate analysis of variance (PERMANOVA) was applied to evaluate
the effect of crop rotation, fertilization, and their interaction on the
changes in spectral structures. Regressions between spectral data and
SOC, TN, and POXC were built by the partial least squares regression
(PLSR) model. The five-fold cross-validation was performed to obtain the
optimal number of latent variables in the PLSR model. The variable
importance in projections of spectral bands in the PLSR model was used
for identifying the great changes in molecular structure for soil
organic matter and TN. Structural equation modeling (SEM) was applied to
determine the direct and indirect effects of selected variables on grain
yield and NAE based on known correlations. The chi-square,P -value, root-mean-square error of approximation (RMSEA), and
good fit index (GFI) were used to evaluate the model fitness. All data
processing, statistical analysis, and visualization were implemented in
R 4.1.0 software (R Development Core Team, 2018).