台灣三維風場改善—應用NCEP再分析資料與內插方法之改良

Abstract

本研究以過往本研究室發展的全台灣地區三維風場為架構，針對產生第一步風場的客觀分析中使用的Barnes權重函數作改良，使經變分分析後格點的模擬值更符合監測站的觀察值。因為客觀分析產生的初始風場嚴重影響後續風場處理的品質，而過往客觀分析使用的權重函數並沒有考慮測站的密度，因此在本研究中嘗試使用可變掃描半徑，測站密度高的地區使用小的掃描半徑，測站密度低的地區使用大的掃描半徑，此外每個格點只使用與它距離最近的數個測站作內插，避免距離過遠的測站也納入格點的計算。結果顯示，使用可變半徑及使用前五個測站作內插會得到更好的風場，其風速絕對誤差值可達0.5m/s或以下，最低誤差值達到0.16m/s，相關性均在0.9以上。 另外本研究加入美國海洋及大氣管理中心的提供的NCEP(Nation Center Environment of Prediction)再分析資料彌補海面測站的不足，取代過往增加海面虛擬測站的方法。以線性內插把NCEP再分析資料內插到設置的海面虛擬測站上，進行客觀分析。結果顯示，以NCEP再分析資料作為模擬範圍的邊界值，能得到合理的風場。

In this work, objective analysis is the first step to Establish 3-D wind field in Taiwan. Initial wind field, which is generated by objective analysis, affect the result of final wind field seriously. To reduce the error, we improved the weighting function of objective analysis used in the past and cause the grid point’s simulation value is more in line with the observation value. The weighting function used in the past does not take into account the density of stations in an aera. A variable radius of influence is used in this stuy.We used a small radius of influence in a high density of stations aera and a large radius of influence in a small density of stations aera.In addition, each grid point is interpolated by the nearest stations. The result of using variable radius of influence obtained a better wind field, the average wind speed error between simulation and observation could be less than 0.5m/s and the minium wind speed error could be 0.16m/s. The wind speed correlation is above 0.9. Moreover, NCEP reanalysis data, which is provided by U.S. Oceanic and Atmospheric Administration Center, was added in our simulation to replace the extrapolate method used in past. Each virtual station’s data is interpolated by NCEP reanalysis data. The results show that by using NCEP reanalysis data as the boundary of domain obtain more reasonable wind field.