Evaluation of three simulation models used to describe plant residue decomposition in soil

We present a protocol for comparing and evaluating selected components of computer simulation models. The plant residue decomposition algorithms from three published simulation models were evaluated systematically by sequentially `pluggingʹ each into a single base model thus testing only one aspect: the hypothesized mechanisms of residue decomposition. By isolating the plant decomposition algorithms, we confirmed that simulated dynamics of plant-derived N within the microbial and mineral soil fractions were sensitive to structural and kinetic differences among models. Using lack-of-fit analysis, we found that simulated values were significantly different from experimental data indicating that none of the models were flexible enough to function accurately under conditions different from those used to calibrate them. Models were also evaluated relative to one another. Algorithms from the Phoenix model gave the closest fit between simulated and experimental values of 15N recovery in biomass (the most sensitive experimentally verifiable variable). We speculate that plant residue decomposition in the Phoenix model, with its functional approach to plant residue fractionation and explicit inclusion of microbial biomass as a factor in decomposition, has the greatest mechanistic integrity of the three models examined.