北京空气质量多模式集成预报系统的建立及初步应用

Ensemble air quality multi-model forecast system for Beijing(EMS-Beijing):Model description and preliminary application

介绍了北京空气质量多模式集成预报系统（EMS-Beijing）．系统中区域空气质量模式包括中国科学院大气物理所自主开发的嵌套网格空气质量模式系统（NAQPMS）模式、美国环保署（EPA）开发的Models-3／CMAQ模式及美国Environ公司开发的CAMx模式等，均使用SMOKE排放模型统一处理大气污染排放清单．此系统各模式采用统一的模拟区域和网格分辨率，采用中尺度气象模式MM5提供统一的气象场，并采用算术平均、权重集成等方法集成各空气质量模式结果，并投入北京空气质量业务预报，有效支持了北京奥运会空气质量数值预报．业务预报结果表明：1）采用SMOKE处理的奥运排放清单较接近实际，2008年8月各空气质量模式可吸入颗粒物（PM10）日均值预报结果相对偏差为-3％～17％，与观测数据相关系数在0．7以上；2）在排放清单较接近实际的情况下，PM10日均值多模式算术平均优于单个空气质量模式；3）分析2008年4—11月业务预报表明，采用权重集成方法预报的PM10日均值优于算术平均方法，达61％．总体上，权重集成方法优于算术平均结果．

The ensemble air quality muhi-model forecast system for Beijing （EMS-Beijing） is introduced in this paper. It includes the IAP/CAS nested air quality prediction modeling system （NAQPMS） , the US/EPA community multiscale air quality （CMAQ） modeling system and the US Environ company＇s three dimensional comprehensive air quality model with extensions （CAMx）. The system uses the unified meteorological field and emissions inventory provided respectively by the fifth-generation NCAR/Penn state meso-scale model （MM5） and the sparse matrix operator kernel emissions （SMOKE）. All the models adopt the same nested domains,with the same grid size and resolution. The EMS-Beijing has been used for Beijing daily air quality routine real-time forecast since March 2008, especially,successfully supporting the air quality forecast for the 2008 Beijing Olympic Games. Various ensemble methods such as arithmetic mean and weight integration methods are compared and the results indicate that：1 ） the emission inventory in August 2008 （Olympic Games） processed by SMOKE is close to the actual,with the model bias （MB） of each air quality model being - 3% - 17% for August 2008 ; 2） the arithmetic mean ensemble method has better performance than any other single model in forecasting daily PM10 concentration; 3 ） according to the daily PM10 concentration results during April-November 2008, the weight mean integration method is better than the arithmetic mean method ,with performance increase of 61%.