Mineral-protection strength
We assumed that decomposing litter residues within mineral matrices could be partitioned into free (not protected) and mineral- protected litter pools (through mineral-organic association and aggregation). During litter decomposition, the free litter pool would be transformed into the mineral-protected litter pool, which could continue to decompose, but at a slower rate than the free litter pool. Such mineral protection would alter total soil respiration and relative contributions from these pools. The stronger the mineral protection, the lower the total soil respiration and the smaller the contribution from the mineral-protected litter pool. Based on this feedback effect of mineral protection on soil respiration, we developed a novel mineral-mediated decomposition model (Eq. 1) to quantify mineral-protection strength,δ , for a specific mineral/soil:
\(C_{t}=\frac{1}{\delta}(1-e^{-k_{1}t})+(100-\frac{1}{\delta})(1-e^{-k_{2}t})\)(1)
where, Ct is the cumulative soil respiration (% of litter C) at time t ; and k 1 andk 2 are the decomposition rate constants of the mineral-protected and free litter pools, respectively. The two pools sum to 100% of total soil respiration at any time.
Our model well (R2 > 0.96) described the dynamics of cumulative soil respiration for all the soils (Fig. 6a). The simulated mineral-protection strength ranged from 0.18 to 0.49 (Supplementary Table 2) and was affected by litter and clay mineral types and their interaction. We observed that the vermiculitic soils had larger mineral-protection strength than did the illite and kaolinite soils for both litter types, and that the mineral-protection strength was larger for maize litter than for soya litter in all model soils except for the vermiculitic soil. The the vermiculitic soil had lower protection strength for maize litter than for soya litter.
We found that the mineral-protection strength explained 67% to 96% of the variation of the formation efficiency of SOM (Fig. 6b). In addition, the mineral-protection strength of the model soil consisting of multiple natural soil clay minerals (δ soil) was predicted well from the mineral-protection strengths (δclayi ) and the relative abundances (%clayi) of the compositional clay minerals in the soil using a multilinear regression equation (Eq. 2). The absolute root-mean square error (ɛ) of the simulation was < 0.05,
δ soil=∑%clayi *δclayi + ɛ (2)
The relative abundances of the clay minerals in the natural soil material quantified by using the decomposition method of X-ray diffraction spectra31 were 6% for kaolinite, 47% for illite, 21% for vermiculite and 26% for a mixed layer mineral of illite and vermiculite. The average of the mineral-protection strengths of illite and vermiculite was used as the mineral-protection strength for the mixed layer mineral of illite and vermiculite for either litter type. As such, the mineral-protection strength of the mixed layer mineral was the largest source of simulation error for either litter type.