2.4. Statistical analysis
Classification and regression tree (CART) analysis is a machine learning algorithm, which is able to identify relatively important variables. CART is a non-parametric model based on graphics, which is easier to understand than traditional statistical techniques. Use CART to judge what the environment factors (air temperature (Ta), soil temperature (Ts), aggregated growing season degree day (GDD), photosynthetic photon flux density (PPFD), precipitation (PPT), air relative humidity (RH) and vapor pressure deficit (VPD)) played a major control role on variations of CO2 fluxes. The maximum number of split was 5, and the minimum Proportional Reduction in Error (PRE) was 0.01. The growing season is from May to October, and the other months belong to the non-growing season (Zhang et al. 2008). Linear regression and correlation analysis were used to explore the inter-annual variation characteristics of CO2 fluxes and its response mechanism to environmental factors. We used SYSTAT 13.0 (Systatsoftware Inc, USA) for the CART and linear regression analyses.