Using remote sensing image for the aeolian morphology and high resolution ancient monsoon indicators, the authors restore how the wind blows in Qaidam Basin since late Holocene. The results show that the prevailing winds of the basin were northwest winds, at the same time there were a small amount of westerly and north wind, with the summer monsoon and westerly circulation having little effect on the wind models. The eastern dune and Yardang landforms of the basin recorded the westerly winds of the last glacial, but the landforms in the midwest indicated that the Asian winter monsoon developed from the northwest during late Holocene. The changes in the atmospheric circulation model took place around 4000 aBP years ago, which was an important node, basin climate shifting from warm and humid to cold and dry.

Response surface methodology (RSM) is utilized to analyze and optimize the experimental condition of V(Ⅴ) removal in single chamber microbial fuel cell (MFC). The influences of initial vanadium concentration, COD and electrolyte conductivity on V(Ⅴ) removal efficiencies are investigated when sodium acetate is adopted as carbon source in single chamber MFC. The results of RSM with three factors at three levels reveal that the initial vanadium concentration has the most significant effect on V(Ⅴ) removal efficiencies, followed by the COD, whereas the influence of the electrolyte conductivity is minimal. The optimal experimental condition obtained by RSM is initial V(Ⅴ) concentration of 75.44 mg/L, COD of 1007.48 mg/L and electrolyte conductivity of 11.98 mS/cm, in which condition the theoretical maximum V(Ⅴ) removal efficiency of 80.31% can be achieved. Verification test is also conducted and the results further confirm the reliability of the optimization. The results of the present study will promote the application of MFC in the remediation of vanadium contaminated groundwater.

Based on the shallow tree and dependency tree, the authors introduce the structural representations, NPST (new phrase-based shallow tree) and NPDT (new phrase-based dependency tree) to Tree-LSTM to compute sentence similarity. Experimental results manifest that the proposed approach achieves a higher performance than the baseline.