Abstract:Gully erosion leads to massive soil loss, which aggravates the migration, loss and morphological changes of organic carbon, and thus induces CO2 emission effects. Most of the previous studies on CO2 emission effects of eroded soil were carried out at watershed and regional scales by model estimation, but not enough attentionwas paid to the CO2 emission effects induced by gully erosion. According to the carbon pool redistribution model, organic carbon and Cesium-137 content of soil profiles from three gullies were used to estimate the CO2 flux induced by gully erosion on the Loess Hilly Region, and the model prediction efficiency and its influencing factors were tested. The results showed that: (1) After long-term erosion, severe erosion occurred in both erosion area and sedimentary area of gullies in the three study areas. The erosion rate in erosion area was between 30.99 and 46.44 mm/a, and that in sedimentary area was between 34.20 and 37.88 mm/a, and the soil loss rate in eroded areas were slightly higher than that in deposition areas. (2) The stronger carbon source existe in both erosion and sedimentary areas. The modeling fluxes in erosion and sedimentary area of gullies ranged from 18.41 to 28.44 g/(m2·a), and from 22.19 to 29.25 g/(m2·a), respectively. (3) Erosion site, soil bulk density, soil organic carbon content, soil erosion amount, average slope of gully, aboveground biomass and underground biomass of vegetation jointly explained the variability of prediction efficiency of the carbon pool redistribution model (R2=0.68), among which erosion site, erosion amount, organic carbon content, soil bulk density, and below-ground vegetation had strong driving effects on prediction efficiency. (4) The neglected parameter of new input organic carbon pool induced by vegetation should be introduced to correct the carbon pool redistribution model and improve the model prediction efficiency. It was clear that the carbon pool redistribution model had a higher prediction efficiency of CO2 flux in gully erosion area compared with deposition area. In order to further improve the prediction accuracy of the model, the new organic carbon pool induced by vegetation should be considered as a calibration parameter incorporated in to the carbon pool redistribution model.