Data analysis
All measured data were subjected to the Shapiro-Wilk test for normality before statistical analysis. Logarithmic transformation was carried out when necessary. One- or two-way analysis of variation was conducted to determine the effects of litter and clay mineral types. The means of three replicates were compared using the least significant difference test at P < 0.05. All statistical analyses were conducted using SPSS 21.0 software.
We developed a new mineral-regulating decomposition model (equation (1)) that included a novel parameter (δ ), describing the feedback effect of mineral-protection on cumulative soil respiration and the relative contribution of decomposition of free litter pool and mineral-protected litter pool. We proposed to use δ to quantify the mineral-protection strength of soil or mineral type by fitting the dynamics of cumulative soil respiration into the model using the least square optimization method with SPSS 21 for Windows.
Linear regression was performed to predict the measured formation efficiency of mineral-associated SOM from the measured mineral-protection strength for all the soils mixed with either litter type. Linear multivariable regression was performed to predict the mineral-protection strength of the natural soil material from those of its compositional all pure clay minerals and their relative abundances in natural soil material for each litter type. As the natural soil material contained a mixed layer mineral of vermiculite and illite, theδ value of this clay mineral type was set as the average of those of vermiculite and illite. A simulated error item was added into the multivariable regression model, and the root mean square error (RMSE) between the measured and estimated δ values of the natural soil material was calculated to estimate the quality of the regression (equation (3)):