资源环境科技发展态势分析平台

Recent studies have shown significant challenges for atmospheric models to simulate tropospheric ozone (O-3) and its precursors in the Arctic. In this study, ground-based data were combined with a global 3-D chemical transport model (GEOS-Chem) to examine the abundance and seasonal variations of O-3 and its precursors at Summit, Greenland (72.34 degrees N, 38.29 degrees W; 3212m a.s.l.). Model simulations for atmospheric nitrogen oxides (NOx), peroxyacetyl nitrate (PAN), ethane (C2H6), propane (C3H8), carbon monoxide (CO), and O-3 for the period July 2008-June 2010 were compared with observations. The model performed well in simulating certain species (such as CO and C3H8), but some significant discrepancies were identified for other species and further investigated. The model generally underestimated NOx and PAN (by similar to 50 and 30 %, respectively) for March-June. Likely contributing factors to the low bias include missing NOx and PAN emissions from snowpack chemistry in the model. At the same time, the model overestimated NOx mixing ratios by more than a factor of 2 in wintertime, with episodic NOx mixing ratios up to 15 times higher than the typical NOx levels at Summit. Further investigation showed that these simulated episodic NOx spikes were always associated with transport events from Europe, but the exact cause remained unclear. The model systematically overesti-mated C2H6 mixing ratios by approximately 20% relative to observations. This discrepancy can be resolved by decreasing anthropogenic C2H6 emissions over Asia and the US by similar to 20 %, from 5.4 to 4.4 Tg year(-1). GEOS-Chem was able to reproduce the seasonal variability of O-3 and its spring maximum. However, compared with observations, it underestimated surface O-3 by approximately 13% (6.5 ppbv) from April to July. This low bias appeared to be driven by several factors including missing snowpack emissions of NOx and nitrous acid in the model, the weak simulated stratosphere-to-troposphere exchange flux of O-3 over the summit, and the coarse model resolution.