EPA-074-109-007

109

4,160 千元

空氣品質監測

2020/01/21

2020/12/31

馮智勇

監資處

陳亦絢;江家慧

多采科技有限公司

大氣空氣

CMAQ 空氣品質動力模式、客觀預報、貝氏模型平均法

Community Multi-Scale Air Quality (CMAQ) modeling system, Objective forecast, Bayesian model averaging (BMA)

楊智閔

02-29334260#38

zhiminyang@manysplendid.com

近年臺灣與離島地區空氣品質，除了受到季節性境外移入的中 國沙塵暴與霾害影響外，本土排放污染源也是影響空氣品質重要的 因素。為提升空氣品質預報能力，提供客觀空氣品質預報指引，前 期計畫完成 CMAQ-OP1.0 空氣品質數值動力模式與其自動化預報流 程，於中央氣象局超級電腦(CWB-HPC)上的建置，並每日定期提供預 報結果；完成新一代數值動力模式 CMAQ-OP2.0 之評估，並在 CWB-HPC 上建置完成 12UTC 預報之自動化預報流程；利用統計校正方法 DCA 及 DMOS，建置有動力模式的偏差校正預報；引入貝氏模型平均法， 發展客觀整合多預報的綜合預報模式…等工作。本計畫延續過去工 作成果，持續維運現行自動化空氣品質數值模式與偏差校正預報作 業，並進一步拓展自動化預報作業，增建 CMAQ-OP2.0 00UTC 預報， 及評估 1 日 4 報與預報時間延長至 120 小時的可能性。考量網頁安 全性以及平台擴充、維運的便利性，建置一個新的展示平台，並新 增有降雨空間分布，以及預報與觀測在測站位置的時序比對，提供 預報人員參考使用。除此之外，亦分析現行統計校正預報，提出可 能的改善方案，以提升校正預報的成效。

CMAQ 空氣品質動力模式、客觀預報、貝氏模型平均法

期末報告.pdf

24MB

2021/1/1 上午 12:00:00

In addition to seasonal transboundary pollution included sandstorm and haze from China, the local pollution also play an important role in Taiwan. To improve the ability of air quality forecast and provide objective forecast guideline. The following works have been done in previous plans. Numerical dynamic air quality forecast model CMAQ-OP1.0 and its automatic forecasting procedures have been established on high performance computer facility in Central Weather Bureau (CWB-HPC), and providing forecast results every day. The assessment of next version numerical dynamic air quality model CMAQ-OP2.0 have been completed. And the automatic forecasting procedures of 12UTC also have been established on CWB-HPC. The statistical correction method DCA and DMOS were used to build up the correction forecast. To develop the objective composite forecast which integrate multi-forecast, the Bayesian model averaging method is also introduced. Following the results of pervious plan, this work continues maintaining and expanding existing automatic forecasting procedures. Which establishes 00UTC forecast of CMAQ-OP2.0 and assesses the feasibility of 4 forecasts in one day and expanding of forecast time from 84 to 120 hours. Considering the safety and convenience for maintaining of forecast display platform, new platform is built, comparison of precipitation spatial distribution and time series of forecast and observation in site position are also provided for forecaster. Besides, we also analyze statistic correction forecast, and propose the improvement plan to elevate the correction forecast results.

Community Multi-Scale Air Quality (CMAQ) modeling system, Objective forecast, Bayesian model averaging (BMA)