Abstract The Data Assimilation Research Testbed (DART) has been extended to be able to assimilate the column-average dry-air mole fraction of CO2 (XCO2) retrievals from the Orbiting Carbon Observatory 2 (OCO-2) satellite. Atmospheric CO2 concentrations over the Midwestern United States were estimated by the Weather Research and Forecasting model coupled with Chemistry (WRF-Chem), with and without assimilating OCO-2 retrievals by the extended-DART. To focus on evaluating the effect of the assimilation, the study period was deliberately set to January 2016, the coldest month in the dormant season, to minimize the influence of biogenic CO2 flux. Independent ground-based and flight observations, as well as the CarbonTracker 2017 products (CT2017), were used to evaluate the results of two distinct approaches. Comparing to the estimated CO2 concentration distribution without assimilating the OCO-2 retrievals, the overall root mean square error (RMSE) and mean bias error (MBE) between the results with the assimilation and the observations were averagely reduced by 20.65% and 78.49%, the overall difference in the RMSE and MBE with respect to CT2017 were averagely reduced by 48.29% and 28.61%, respectively. Experiments showed that the assimilation of OCO-2 retrievals by the extended-DART could make the estimated CO2 concentration distribution significantly more consistent with the observations and the CarbonTracker products.

中文翻译：

---

用 WRF-chem/DART 同化 OCO-2 检索：美国中西部的案例研究

摘要 数据同化研究试验台 (DART) 已扩展为能够同化来自轨道碳观测站 2 (OCO-2) 卫星的 CO2 (XCO2) 反演的柱平均干空气摩尔分数。美国中西部的大气 CO2 浓度是通过天气研究和预测模型结合化学 (WRF-Chem) 估算的，使用和不使用扩展 DART 同化 OCO-2 反演。为重点评价同化效应，特意将研究时间定为2016年1月，即休眠季节最冷的月份，以尽量减少生物源CO2通量的影响。独立的地面和飞行观测以及 CarbonTracker 2017 产品 (CT2017) 用于评估两种不同方法的结果。与未同化 OCO-2 反演的估计 CO2 浓度分布相比，同化结果与观测结果之间的总体均方根误差 (RMSE) 和平均偏差误差 (MBE) 平均降低了 20.65% 和 78.49%，相对于 CT2017，RMSE 和 MBE 的总体差异分别平均减少了 48.29% 和 28.61%。实验表明，扩展 DART 对 OCO-2 反演的同化可以使估计的 CO2 浓度分布与观测结果和 CarbonTracker 产品更加一致。相对于 CT2017，RMSE 和 MBE 的总体差异分别平均减少了 48.29% 和 28.61%。实验表明，扩展 DART 对 OCO-2 反演的同化可以使估计的 CO2 浓度分布与观测结果和 CarbonTracker 产品更加一致。相对于 CT2017，RMSE 和 MBE 的总体差异分别平均减少了 48.29% 和 28.61%。实验表明，扩展 DART 对 OCO-2 反演的同化可以使估计的 CO2 浓度分布与观测结果和 CarbonTracker 产品更加一致。