Testing the RothC and DNDC models against long-term dynamics of soil organic carbon stock observed at cropping field soils in North China

•Accurate C input was crucial for modeling turnover of soil organic carbon (SOC).•RothC simulated similar to DNDC for SOC when they used DNDC modeled crop biomass.•RothC simulated well SOC when measured crop residue was as carbon input.

RothC and DNDC, two process-based agroecosystem models widely utilized for quantifying soil carbon (C) sequestration, were tested against datasets of soil organic carbon (SOC) measured in two long-term experiments with winter wheat-summer maize rotation at field soils in North China. In the study, the two models were utilized in conjunction in three modes, i.e., using DNDC alone, using RothC driven by soil C inputs of measured crop residue, and using RothC driven by soil C inputs derived from DNDC simulations. In general, results from the three ways of simulations were in agreement with the measured SOC dynamics for the two experimental sites with varied treatments. However, discrepancies in the results revealed the differences between the two models regarding their functionality. The simulations with RothC driven by the measured crop residue as C input produced the best results on SOC dynamics in comparison with observations. Results from the simulations with DNDC alone were less correlated to observations mainly due to the approximation of the modeled amount of crop residue. Due to the same reason, the simulations with RothC but driven by DNDC-modeled crop residue poorly fit to observations. The results conclude that given the similarity between RothC and DNDC regarding their routines modeling SOC transformations it is crucial for improving the models’ performance to provide accurate C input data (mainly crop residue) to the SOC turnover modules in the models.