4.4. synergistic effect and prediction models
Mande believed that the soil temperature and moisture explained the spatial and temporal variation in soil CO2 flux. Likewise, the changes in the soil properties and forest carbon significantly increased the soil CO2 efflux indicating a strong positive correlation(Mande et al., 2015). In our study, multivariate correlation analysis was carried out on soil temperature, moisture, concentration of organic carbon and soil CO2flux at different layers (5-20cm, 80-200cm). The results showed that there was a significant correlation (P<0.01)whether at the deep layer (80-200 cm) or shallow layers. All of above suggested that soil temperature, moisture, and soil organic carbon substrates all exhibited synergistic responses to soil CO2 flux whether at deep layers or shallow layers.
Chen et al. developed a T&P&C-model including SOC as an additional predictor of annual R-s. This extended but still simple model performed better than the T&P-model and explained 69%, 89%, and 47% of the interannual and intersite variability of R-s for croplands, grasslands and forests, respectively. And the modeling efficiency of the T&P&C-model was nearly 60% across ecosystems. (Chen et al., 2010). In this study, based on the traditional hydrothermal two-factor models (T&M models), combined with the environment and substrate characteristics of artificial Robinia pseudoacacia , a three-factor model of temperature(T), moisture(M) and organic carbon substrate (C) was established at the deep and shallow layer, respectively(Table 4). Better estimates of CO2 flux at deep layers than shallow layers would be obtained with the new model driven by temperature(T), moisture(M) and organic carbon substrate together. In order to further verify the superiority of the T&M&C model, we selected vertical CO2 flux data of different ecosystems in different research areas through literature, and compared the accuracy of the T&M&C model with the T&M model. The prediction accuracy (RMSE/MAE) was increased by average of 20~26%, and T&M&C model was of better results in prediction effect. Therefore, the new multi-factor model driven by climate and soil properties can better estimate CO2 flux at different layers and should be widely used to predict global carbon emissions.