基于中尺度气象模式（MM5）、排放源模式（SMOKE）和大气化学模式（CMAQ），耦合本地排放源清单，建立并业务化运行了珠三角区域大气灰霾数值模式预报系统。该系统考虑了气象过程、气相与液相化学过程、非均相化学过程、气溶胶过程和干湿沉降过程，能够很好地预报各种气溶胶对能见度的消光贡献以及灰霾现象的发生、持续和消亡过程。从预报的能见度和站点观测资料以及MODIS AOD区域分布对比来看，该套模式系统对典型灰霾过程的能见度时空分布有较好的模拟能力。从全年业务预报运行来看，能见度预报和观测值有较好的一致性。在旱季灰霾多发季节，预报效果较好。全年平均来看，市区TS评分达到0.67，预报效率EH达到0.89；郊区，TS评分达到0.37，预报效率EH达到0.83。研究表明，模式系统对灰霾以及能见度量级有较好预报能力，能达到业务运行要求。

In this study, haze numerical model forecast system was established and run daily by coupling MM5, SMOKE and CMAQ models over Pearl River Delta which used the local sources emission inventories. The meteorological processes, gas-phase chemistry, aqueous-phase chemistry, aerosol processes, heteorogeneous chemistry and deposition processes can be calculated in this modeling system. And the aerosol and extinction contribution, as well as processes of haze initiating, weakening, and intensifying, could be well simulated. By comparison with the forecast and site observations visibility, as well as the regional distribution of MODIS AOD, the spatial and temporal distribution of visibility could be well simulated by this modeling system during a typical haze process. Judging by the running annual business forecast, visibility forecasts were in good agreement with observations. The forecast was better in the dry season (haze season).In urban areas, the annual average TS score was up to 0.67, forecast efficiency EH was up to 0.89; In suburban districts, the TS was 0.37, the EH was up to 0.83. The study showed that, the model system had a good forecasting ability for haze and magnitude of visibility to achieve business operational requirements.