3.2. CO2 flux
There was different tendency of CO2 flux rate at four slope positions between the initial and last part of the 47 day measurement period (Fig. 2a). We used piecewise linear regression of the CO2 flux and soil temperature to parse the data into two measurement subperiods. The 28th day was determined to be the turning point. For the initial 27 days, position, time, and interaction of position and time significantly affected CO2 flux(p < 0.001). From summit to toe-slope, CO2 flux declined with altitude. The flux showed an interaction between slope position and time: CO2 flux at summit decreased over time, CO2 flux at toe-slope increased, and CO2 flux at shoulder-slope and back-slope kept steady. There was no obvious difference in fluxes among the four positions at the 27th day (Fig. 2a). For the last 20 days, position significantly affected CO2flux (p < 0.05). There was a similar trend of CO2 flux reducing gradually with time at all four positions. Post hoc analysis of LSD means showed that there were significant differences in CO2 flux between all four positions in the initial 27 days; however, in the last 20 days, significant differences only existed between summit and the other three positions (Table 2).