基于伴随方法的大气污染溯源

On adjoint method based atmospheric emission source tracing

大气污染防治的核心是找准污染源头,厘清污染成因,实现靶向治理,提高控制效率.本文建立了全国空气质量高分辨率预报与污染控制决策支持系统(NARS,呐思系统),实现了气象与大气化学的监测、同化、预报、溯源、排放源反演和动态优化控制等大气污染闭环防控,可为大气污染防控提供一整套解决方案,其中所建立的CAMx伴随溯源模式,实现了排放源动态反演和网格化定量溯源,可快速定量追溯导致目标区域未来7天大气污染的排放源及其贡献率时空分布.针对2016年9月~2017年3月北京主城区PM_(2.5)集中污染时间段进行了排放源反演、气象场预报、空气质量预报和网格化溯源.与京津冀地区国控点监测结果进行了对比分析,结果表明污染过程、污染等级和污染物浓度预报的准确率分别为100%、88.8%和84.7%,预报值和监测值之间的相关系数为0.81.网格化溯源结果表明导致北京主城区PM_(2.5)污染的排放源基本来自于北京西南方向这一条大气污染物传输通道,北京本地、河北、天津及周边地区排放源对北京主城区PM_(2.5)浓度分别贡献了66%、29%、5%.就重污染过程而言,京津冀排放总量的19%导致了北京主城区80%的PM_(2.5)重度及以上污染,其中北京本地占京津冀排放总量的9%贡献63%、河北占京津冀排放总量的10%贡献17%.就整体污染天气而言,京津冀排放的26%导致了北京主城区80%的PM_(2.5)轻度以上污染,其中北京本地占京津冀排放总量的9%贡献61%、河北占排放总量的15%贡献18%,天津占排放总量的2%贡献1%.导致北京主城区PM_(2.5)污染的排放源主要分布在北京城区和南部区域、保定和石家庄所辖的部分区县,贡献排名前6位的区县均为北京辖区,贡献率合计为48%,前20个区县的总体贡献为73%.将动态反演排放源方法与调查排放清单相结合,应用伴随溯源模式对预报结果进行同步大气污染溯源,可为大气重污染应急控制找准控制对象,并进行损益评估,运用自然控制论,实现大气污染应急优化控制.

The utmost important task for air pollution prevention and control is to identify the source of pollution such that targeted management and governance can be greatly enhanced. In this paper, we have develoepd an air quality forecasting system called ＂the national air quality high-resolution forecast and pollution control decision support system（NARS）＂, which integrates atmospheric environment monitoring and analysis, meteorological numerical forecasting, emission source inversion, air quality forecasting, meteorological and atmospheric chemical data assimilation, emission sources tracing,profit and loss assessment together with dynamic optimal control. The system could be applied to perform the closed-loop prevention and control for atmospheric pollution by monitoring, forecasting, assimilation, traceability, emission source inversion and optimal control, and the system stands a great promise of providing a solution for current air pollution prevention and control strategy. In the NARS, a CAMx based adjoint model was developed, and was further demonstrated to perform dynamic inversion and grid-based quantitative emission source tracing. Our results show that the system can trace quickly and quantitatively both spatial and temporal distributions of emission sources and their contribution rates resulting from severe air pollution for certain target area for the coming 7 days. As an application of the NARS, the emission inversion, meteorological field simulation, air quality forecasting and grid traceability were studied for the time period from September 2016 to March 2017 in Beijing. Compared with the observation of national control station in Beijing-Tianjin-Hebei region（BTH）, the results showed that the forecast accuracies of the pollution processes, the pollution level and the pollutant concentration are close to 100%, 88.8% and 84.7%, respectively, and the correlation coefficient between the modeling and observation is 0.81. The results of grid traceability showed that the emissions of PM（2.5） pollution in the main urban area of Beijing are mainly from the air pollutant transmission channel in the southwest of.The emission contributions of BTH plus the surrounding areas are respectively 66%, 29% and 5%. The 19% PM（2.5） emissions of BTH have resulted in 80% Beijing PM（2.5） concentration in heavily polluted and severely polluted days. Among them, 9% Beijing local emissions account for 63% PM（2.5）, and 10% Hebei emissions accounted for 17%. The 26% BTH emissions have resulted in 80% PM（2.5） in Beijing＇s main city for slightly or heavily polluted days. Furthermore, Beijing emissions accounting for 9% of BTH emissions have contributed 61%; meanwhile, Hebei emissions accounting for 15% of BTH emissions have contributed the 18% of the pollution; finally, Tianjin 2% accouts for the leaving 1%. The emissions leading to the PM（2.5） pollution in Beijing main city mainly distribute in the city zone and the sourthern of Beijing, some districts or counties of Baoding and Shijiazhuang. The top six districts or counties accounting for 48% pollutants are located in Beijing, while the top 20 districts or counties contribute 73%. Here, using the the adjoint model developed, we are able to analyze synchronously the emisison apportionment of the modeling results through combining the dynamic inversion and the monitoring emission inventories. Using both profit-loss assessment and natural cybernetics, the NARS is able to locate the emisison sources for optimal emisison control during a heavy air pollution episode.