Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/14188
標題: 地表及地表下降雨-逕流模式之研究
A Study on Surface and Subsurface Rainfall-Runoff Model
作者: 畢嵐杰
Pi, Lan-Chieh
關鍵字: rainfall-runoff model;降雨-逕流模式;optimum theory;Geographic Information System;linear reservoir;recession curve theory;kinematic wave model;tank model;最佳化理論;地理資訊系統;線性水庫;退水曲線理論;運動波模式;水筒模式
出版社: 土木工程學系
摘要: 
本論文旨在建構一套地表及地表下降雨-逕流模式,藉以闡明集水區複雜之水文歷程。論文內容主要分為三個部分進行研究,首先,利用菲利普入滲模式將降雨量區分為有效雨量及入滲量,分別為地表逕流模式(ST Model)及地表下逕流模式(TSTs Model)之輸入量,可避免一般水文模式在進行基流分離所造成之不確定性影響。
第一部份即建構地表逕流模式,採用線性水庫之概念為架構,以運動波模式為基礎配合退水曲線理論推導出地表逕流孔口參數理論式,使得參數結合集水區地文條件,可確切反映地表狀況,突破傳統水筒模式參數無法描述地表情形的缺點。此模式所需之地文參數係運用地理資訊系統及數值高程模型進行計量地形分析而得;另外對於糙度係數則應用遙感探測分析技術及衛星影像資料,以監督式倒傳遞類神經網路分類法判釋集水區之地表覆蓋情形而獲取,可有效掌握集水區內地文之時空變異特性,充分表現實際地面情況。
第二部分為建構地表下逕流模式,係以二個直列式水筒模式為架構,分別模擬中間流及地下水流,此兩部分水流之總和即為地表下逕流量。各筒間皆以線性水庫及水文連續方程式控制流入與流出量之機制。參數係由全域優選方法率定而得,因此本文即針對Multistart Powell法及SCE法兩種全域參數最佳化率定法進行評析與比較,發現在擴大參數搜尋範圍的情形下,無論數值試驗及實際觀測資料皆顯示Multistart Powell法在參數收斂的一致性上明顯優於SCE法,因此地表下逕流模式之參數採用Multistart Powell 法,並配合本文所研議之自動化檢定策略,給定適當的限制條件、懲罰函數機制以及退水區間長度的資料個數至少需95的退水流量資訊,可獲致率定參數收斂一致性的結果。
第三部分則結合前述二部分之逕流模式構築完整的地表及地表下降雨逕流模式(2S-R Model),由於地表逕流機制考慮糙度、坡度等集水區地文條件以及入滲容量等因素,因此可將逕流機制合理地劃分為地表、中間流及地下水逕流3個次逕流分量,且由於考慮地表下逕流機制,使得對於集水區逕流歷線之退水段模擬更為精確,改善過去水文模式高估初始歷線與低估最終歷線之現象,經以合坑溪集水區為模擬研究試區,證實其具有良好之精確度,可更真實地呈現逕流歷線之原貌。

The main purpose of this dissertation is to develop a surface and subsurface rainfall-runoff model that illustrates the complexity of the watershed hydrological process. This research contains three steps. First, using Philip's infiltration model, the rainfall is divided into effective rainfall and infiltration. These two portions are the input for the surface (ST Model) and subsurface runoff models (TSTs Model), respectively. This arrangement avoids the uncertainty caused by the base flow separation procedure in conventional models.
The first step is to establish a surface runoff model. The linear reservoir concept is used as the framework for this model. Based on the kinematic wave model and recession curve theory, a theoretical orifice coefficient formula for surface runoff model is derived. Incorporated with the geomorphic conditions, these parameters better represent the surface conditions. This overcomes the shortcoming in the conventional Tank Model in lacking the ability to describe the surface conditions. The geomorphic parameters of the ST model were analyzed using the Geographic Information System and Digital Elevation Model. Remote Sensing analysis technology and satellite image information was used to produce the coefficients of roughness using a supervised backward propagation neural network model to interpret the surface coverage conditions. The proposed model can effectively handle the temporal and special watershed variation characteristics.
The second step is to establish a subsurface runoff model constructed using Two Serial Tank models simulating the interflow and subsurface flows, respectively. A summation of the quantity of these two flows is the subsurface runoff value. The linear reservoir and continuity equations govern the inflow and outflow control mechanism between these two tanks. The model parameters were obtained using the global optimization method. This research compares the Multistart Powell and SCE methods. The comparison results show that the Multistart Powell method is superior to the SCE method in parameter convergence consistency in numerical test and observation data. The subsurface runoff model parameters are identified using the Multistart method combined with the automatic validation strategies proposed by this research. Given appropriate constraints, the penalty function and no less than 95 flow data are in the recession period. Consistent model convergence parameter results can be achieved.
The third step in this research is to combine the aforementioned runoff models into a comprehensive Surface and Subsurface Rainfall-Runoff Model (2S-R Model). Geomorphic conditions including roughness, slope, and infiltration capacity are considered in the surface runoff mechanism. The runoff mechanism is divided into three runoff components; surface flow, interflow and subsurface flow. Considering the subsurface runoff mechanism, more accurate recession limb in the hydrograph simulations can be carried out to improve the over estimate in the initial stage and under estimate in the recession period over that obtained with conventional hydrological models. Using the observation data from the Her-Kan Creek watershed, the proposed 2S-R model was proven to have better accuracy in rainfall runoff process modeling.
URI: http://hdl.handle.net/11455/14188
Appears in Collections:土木工程學系所

Show full item record
 

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.