評估改良式統計降尺度模式應用於日雨量模擬之研究

Assessment of an improved statistical downscaling model for daily precipitation simulation

廖信華(Hsin-Hua Liao)

指導教授 ： 林國峰

國立臺灣大學/工學院/土木工程學研究所/碩士(2013年)

https://doi.org/10.6342/NTU.2013.03020

摘要

氣候變遷研究中常使用大氣環流模式(General Circulation Models, GCMs)輸出值配合預設情境推估未來之水文變化。由於大氣環流模式屬於全球尺度，無法直接反映局部區域的氣候特性，因此必須透過降尺度方法將大氣環流模式輸出值轉換到區域尺度。降尺度方法主要有動力降尺度及統計降尺度兩種，其中統計降尺度較為簡單且成本較低，因此被廣泛地使用。但過去研究中指出統計降尺度方法容易有低估的情形，極端值低估狀況更明顯，因此本研究為了提升統計降尺度在極端值的模擬表現，使用分類的概念提出一個改良式統計降尺度模式。
本研究提出的分類降尺度模式，其主要架構包含兩個部分:颱風警報判斷模式與雨量降尺度模式。首先颱風警報判斷模式是以支援向量機(Support Vector Machine, SVM)建立NCEP/NCAR大氣變數與海上颱風警報之間的非線性關係，判斷該日是否發布海上颱風警報。其次，雨量降尺度模式將各測站的觀測雨量分類為颱風雨量與非颱風雨量，以改良式自組織映射線性輸出模式(Improved Self-organizing Linear Output Map, ISOLO)分別建立大氣變數與颱風雨量及大氣變數與非颱風雨量之間的關係，再將兩者模擬雨量合併成日雨量時間序列。最後與各測站未分類降尺度模式之結果比較。
本研究提出一個提升統計降尺度在極端值模擬的分類降尺度模式，並分別以海上颱風警報歷史資料與各測站觀測雨量來驗證提出之模式。結果顯示颱風警報判斷模式具有良好的判斷能力且表現穩定。分類降尺度所模擬之雨量在整體準確度有所提高之外，極端值的模擬也有明顯改善，顯示本研究提出之分類降尺度模式具有良好的模擬能力。總結而言，本研究提出的分類降尺度模式能有效運用在降尺度上，有助於後續氣候變遷相關研究。

關鍵字

氣候變遷 ； 統計降尺度 ； 支援向量機 ； 改良式自組織映射線性輸出模式 ； 極端值

並列摘要

Today, General circulation models (GCMs) are the main tools used to assess the impact of specific climate change scenarios on the future climate. They provide simulation of weather variable at coarse scale which cannot reflect characteristic of regional climate. To fill the gap between GCM runs and regional climate, downscaling methods have been developed to downscale GCM runs to finer scale (the dynamical downscaling methods and statistical downscaling methods). Statistical downscaling methods are so easy and inexpensive that they have been widely used. However some researches reveal that statistical downscaling methods tend to underestimate variables and poorly in extreme-value.
In this study we present, develop and validate an improved statistical downscaling model named classified downscaling model to solve this problem. The proposed model is consists of two parts: typhoon-warning judgment model and downscaling model. Firstly, typhoon warning judgment model is built by support vector machine (SVM) to establish the relationship between NCEP/NCAR weather variables and Typhoon warning. Typhoon-warning judgment model is to judge the day of typhoon warning. Secondly, the observed precipitation of each station is classified to typhoon precipitation and non-typhoon precipitation. Then the improved self-organizing linear output model (ISOLO) is used as downscaling model to establish typhoon precipitation model and non-typhoon precipitation model separately. The two simulated results are combined to daily precipitation time series. Finally, the classified downscaling model is compared with the unclassified downscaling model in each station.
On the other hand, typhoon-warning records and observed precipitation in each station are used to validate the proposed model separately. The results show that typhoon-warning judgment model displays well and stable. The performance of the classified downscaling model promotes not only the whole accuracy but also the simulation of extreme-value. In conclusion, the purposed model is useful for downscaling and expected to be helpful to support climate change researches.

並列關鍵字

climate change ； downscaling model ； support vector machine ； improved self-organizing linear output model ； extreme-value

參考文獻

2.Chang, F.J., Chang, L.C. and Wang, Y.S. (2007). “Enforced self-organizing map neural networks for river flood forecasting.” Hydrological Processes 21(6): 741-749.

3.Charles, S.P., Bates, B.C., Smith, I.N. and Hughes, J.P. (2004). “Statistical downscaling of daily precipitation from observed and modelled atmospheric fields.” Hydrological Processes 18(8): 1373-1394.

4.Chen, S.T., Yu, P.S. and Tang, Y.H. (2010). “Statistical downscaling of daily precipitation using support vector machines and multivariate analysis.” Journal of Hydrology 385: 13-22.

5.Chen, H., Xiang, T.T., Zhou, X. and Xu, C.Y. (2011). “Impacts of climate change on the Qingjiang Watershed's runoff change trend in China.” Stochastic Environmental Research and Risk Assessment 26(6): 847-858.

6.Crane, R.G. and Hewitson, B.C. (1998). “Doubled CO2 precipitation changes for the susquehanna basin: Down-scaling from the genesis general circulation model.” International Journal of Climatology 18(1): 65-76.