Diffusive flux of CH4 and N2O from agricultural river networks: Regression tree and importance analysis

Abstract

River networks in subtropical agricultural hilly region become an inconvenient greenhouse gas (GHG, methane and nitrous oxide) source because of the influence of human activities, which has caused large uncertainties for refinement of national GHG inventories and their global budget. Based on field monitoring experiments at high temporal resolution, we employed regression tree and importance analysis to identify quantitatively factors that influence the diffusive flux of GHGs to provide a scientific basis for reducing GHG emissions and controlling regional carbon and nitrogen losses. The results indicate that significant spatiotemporal variation of methane (CH4) nitrous oxide (N2O) diffusion occurs in all the four reaches (W1, W2, W3 and W4) of Tuojia river networks. Among them, W1 contributed lowest CH4 (22.55 μg C m−2 h−1) and N2O (5.00 μg N m−2 h−1) diffusive flux than the other three (P < 0.05), while W4 offered highest CH4 (166.15 μg C m−2 h−1) and N2O (30.47 μg N m−2 h−1) diffusive flux but with no statistically significant difference between W2 and W3 due to homogeneous extraneous nutrition loading into the two reaches. W4 also contributed largest cumulative flux of CH4 (14.55 kg C ha−1 yr−1) and N2O (2.69 kg N ha−1 yr−1) in Tuojia River networks (P < 0.05). Furthermore, the regression tree and importance analysis indicate that, in the anaerobic environment, dissolved oxygen saturation controlled the production and diffusion for both CH4 and N2O. The findings of this investigation highlighted that decision support tools provide an effective pathway to enhance the GHG mitigation technology research in agroecosystems and simultaneously shed light on the global campaign on refinement of national GHG inventories as well as regional nutrient management.