本計畫將結合邊界層密集觀測實驗及高解析度氣象，空品數值模式建立台灣複雜地形下之邊界層資料同化系統。欲同化的觀測資料包含陸地地面觀測，無人機，氣膠光達及探空，海上海研船探空。透過這些觀測資料的高垂直解析度及高觀測頻率特性使模式能掌握台灣複雜地形下邊界層時空變化。此外，透過資料同化參數與策略進一步探討如何讓提高邊界層觀測之同化效益，及其對高影響天氣事件預報之重要性。此外，透過系集模擬的建立將提供建立臺灣近地表預報之敏感區域，以建立常態型觀測降低預報誤差。而利用邊界層資料同化架構所建立之四維分析資料將提供了解台灣複雜地形下大氣邊界層發展，背風面低層氣流變化之動，熱力機制，乃至邊界層內污染物傳輸機制，有助於提供後續精進空品及風能預報之關鍵技術。為解決近地表模式偏差問題，本計畫將採用統計方法建立預報偏差校正，並將此法嫁接於資料同化之分析循環步驟中。所採用的統計方法將包含傳統統計迴歸模式及類神經網路工具。 ;This project aims to establish a high-resolution ensemble data assimilation system for the planetary boundary layer over the Taiwan complex terrain (PBL EDA system) by combining the intense observations in PBL and high-resolution meteorology and air-quality model. We plan to assimilate surface data, UAV, aerosol lidar, and regular and ship-based radiosondes. By taking advantage of the high vertical resolution of the observations, prediction initialized from the PBL EDA analysis can better represent the PBL variations under the complex terrain of Taiwan. Besides, the ensemble simulation helps to establish the sensitivity of near-surface prediction, which allows us to explore the possibility of installing the regular PBL observations.Based on the PBL DA framework, the 4-dimensional analysis product provides the potential to understand the development of PBL under the complex terrain, the variability of leeside flow, and how these may modulate the transportation of the air pollutant in the PBL. The PBL EDA system allows building the key to improve air quality and wind energy prediction. In order to mitigate the near-surface model bias, this project will develop the bias correction methods based on multi-variable regression and neural networks. These methods will be integrated into the PBL data assimilation framework.
