中文摘要：

利用WRF-STILT模型模拟玉米种植区生长季（6-9月）小时CO2浓度,并基于美国最大农业种植区‘玉米带’100m高塔CO2浓度观测数据,对WRF-STILT模型的模拟能力及CO2通量的不确定性对模拟结果的影响进行分析。结果表明：（1）WRF-STILT能够模拟高塔观测的CO2浓度日变化特征,模拟值与观测值的均方根误差为13.70mmol×mol^-1,模拟结果偏高7.26mmol×mol^-1。（2）EDGAR和Carbon Tracker两种典型化石燃料的CO2通量,其区域平均值相差〈6%,但两者对CO2浓度增加值的模拟结果相差约10%;（3）CO2通量空间分辨率的差异会导致模拟结果产生偏差,使用区域边长为1o的EDGAR化石燃料CO2通量模拟的浓度贡献值仅为0.1o的0.4倍,且空间分辨率越低,模拟误差越大;（4）白天和夜晚Carbon Tracker模拟的植被生态系统净交换数据是高塔涡度相关方法观测结果的2.26和1.56倍,下垫面分类的误差以及相应的通量模拟误差使模拟的CO2浓度贡献出现12mmol×mol^-1的差异,这是模拟结果偏高7.26mmol×mol^-1的潜在误差来源。研究认为,WRF-STILT模型和高空间及时间分辨率的CO2通量能够较好模拟出农业区生长季的CO2强日变化特征,CO2通量的误差是模拟结果误差的主要来源,研究结果表明该方法具有评估和优化通量的巨大潜力。

英文摘要：

Based on the CO2 concentration observations in U.S. corn belt, which was measured at 100 m height of a tall tower, hourly CO2 concentration was simulated for the growing season（June–September, 2008） with the WRF-STILT model. And the effect of flux uncertainty on modeled CO2 concentration was also analyzed. The results showed as：（1） WRF-STILT model can simulate the observed strong diurnal variation in growing season, with RMSE be 13.70mmol×mol^-1, and it was overestimated by 7.26mmol×mol^-1, the shape and area of intense footprint zonesare different for different months（September August June July）.（2） The difference of regional average anthropogenic CO2 flux for EDGAR and Carbon Tracker was within 6%, when both of them were at the same spatial resolution, the simulated CO2 enhancement difference was close to 10%.（3） Spatial resolution can lead to large bias in the modeled CO2 enhancement, when using 1o emissions, the simulated CO2 enhancement was only 0.4 times of the results using 0.1o emissions, and with the decreases of spatial resolution, the modeled bias increases.（4） Daytime and nighttime NEE of Carbon Tracker is 2.26 and 1.56 times that of tall tower NEE observations, and the misrepresentatives of underlying land use categories can lead to about 12 mmol×mol^-1 bias in the modeled results, which may be the potential reason of bias high for 7.26 mmol×mol^-1. Our study concludes that when combing WRF-STILT model with high quality CO2 flux, the strong diurnal variation of CO2 concentration can be well simulated, and the uncertainty of CO2 flux is the main reason for modeled CO2 concentration bias, it also indicates the potential of evaluating and retrieving prior CO2 flux.