Analysis of the response of grassland biomass and soil organic carbon to precipitation changes based on the DNDC model in the loess hilly region

Grassland ecosystems represent an important terrestrial ecosystem in the semi-arid region of the Loess Plateau, and changes in precipitation patterns have an impact on the carbon cycle process in this region. In situ precipitation gradient experiments were carried out in naturally restored grassland (NR) and Bothriochloa ischaemum artificially restored grassland (BI). Natural precipitation was used as a control, and nine precipitation gradients (P – 80, P – 60, P – 40, P – 20, CK, P + 20, P + 40, P + 60, P + 80) were established by decreasing the natural value by 80%, 60%, 40% and 20% and increasing this value by 20%, 40%, 60% and 80%. In addition, a denitrification-decomposition (DNDC) model was used to simulate and analyze the effects of precipitation change on soil organic carbon (SOC) and aboveground and underground biomass. The results showed that 1) compared with natural precipitation, the SOC content of NR presented the greatest increase (21.4% and 21.5%) when the precipitation increased or decreased by 80%, respectively, although significant differences were not observed between these two treatments (P > 0.05). The SOC content of BI reached the peak value (13.41 g·kg⁻¹) when the rainfall decreased by 20%. Overall, the SOC content of BI was higher than that of NR. 2) The biomass of the two types of grassland restoration fluctuated with rainfall changes. When the rainfall decreased by 20% and increased by 40%, the underground and aboveground biomass of NR increased by 53.6% and 52.4%, respectively, whereas when the rainfall increased by 20% and decreased by 40%, these parameters achieved the maximum values in BI. Moreover, the aboveground biomass of NR was obviously higher than that of BI, while the underground biomass presented the opposite results. 3) The DNDC model results showed that the change trends of SOC content and simulated aboveground and underground biomass values of NR and BI were basically consistent with the measured results, with RMSE% values of 10.0, 7.5, and 6.7 for NR and 11.7, 1.3, and 1.0 for BI, respectively; and R² values of 0.578, 0.989, and 0.989 for NR and 0.776, 0.998, and 0.999 for BI, respectively. However, the model was more consistent with the simulation of biomass, and the simulation of BI was more accurate. Therefore, the DNDC model is appropriate for simulating the SOC and biomass of grassland ecosystems in the loess hilly region.