小波分析方法應用於坡地集水區水文模擬之研究

本文係建立一種可以同時分析流量資料噪訊及模擬降雨－逕流相互關係之時間序列小波串聯模式，用以從事本省上游集水區降雨－逕流歷程之模擬。一般水文預報方法須先建立一合適之模式，並有賴足夠且正確之資料以用於模式之建構；然而資料常受諸多因素之干擾，致使降雨－逕流間之因果關係不易確定，是爲造成模式誤差之主因。本文乃應用小波理論研析流量資料之物理特性並建立水文模式，期能提供水文預報之另一嶄新方法。由於小波分析具有強大之資料抽取能力，能將逕流資料中之干擾因素抽離；而類神經網絡模式具有大量參數之模式機制，能模擬並預測非線性之水文現象。鑒於本省山坡地之地形崎嶇、地貌多變，降雨落於集水區上匯流至集水區出口處，其歷程遭受許多氣象、水文、地文及人文等複雜因素之干擾，造成降雨－逕流間之關係往往呈現高度之不確定性。小波分析及類神經網絡模式能提供有效之解決方法，故本文應用小波分析之噪音抽取能力及類神經網絡模式之非線性計算能力，期以建立一兼具資料噪訊分析、現象模擬及結果預測之水文模式。文中並分別套用A、B及C三種不同模式組合以試驗小波分析於各不同分析模式之適用性。為了驗證模式之適用性，本文將所建立之降雨－逕流模式應用於德基水庫上游集水區之颱洪事件做為模擬之依據，並得到令人滿意之驗證結果，且由時間序列小波A、B及C三種組合模式所模擬之結果均具有良好之精確性。因此，本文所研擬之分析方式及模擬結果可提供本省流域水資源規劃及防洪工程設計之參考應用。

關鍵字

小波分析；類神經網路；水文模擬

並列摘要

The main purposes in this paper are to build up a time-series wavelet analysis model which reflects runoff noise and rainfall-runoff relationships, and to apply to the rainfall-runoff simulation of upland watersheds in Taiwan. For the forecasting approaches commonly applied, a suitable model is firstly adopted. A great deal of correct sampling data are then required to build the structure of model. However, the rainfall-runoff relationships are interfered by some uncertain factors, which cause major simulated errors of model. It is of great importance to analyze theoretically the physical characteristics of hydrological models. The method of wavelet analysis can provide the possibility of hydrological simulation for a project watershed. With powerful ability of data classification, wavelet analysis can separate the disturbance from runoff data. On the other hand, with a large number of model parameters, the artificial neural network (ANN) model can be applied to simulate and forecast the nonlinear hydrological phenomena. The runoff process is often influenced by atmospheric conditions, hydrological characteristics, physiographic factors and human activities. Therefore, the rainfall-runoff relationships existed to be highly uncertain. By combination of wavelet analysis and ANN model can provide an effective way to solve the problems. With the separated ability of noise in wavelet analysis and solution of nonlinear equations of ANN model, the hydrological model with noise analysis, field simulation and future forecasting can be constructed. To verify the appropriateness of the approaches adopted, the upland watershed of the Te-Chi Reservoir in middle Taiwan is chosen as a project area. The results of the approaches proposed show a satisfactory simulation. Therefore, it is justifiable to reveal that this study can be further employed to estimate and forecast the peak of flood during the typhoon period in upland watersheds and play an important role on the planning of flood mitigation in Taiwan.

並列關鍵字

Wavelet analysis ； Artificial neural network ； Hydrological simulation

主題瀏覽