利用分區式隨機碎形編碼模擬河川特性並探討土地特徵改變對降雨—逕流歷程之影響

Abstract

本研究目的乃將大自然之河川網路型態、河川長度、河川集水區面積及河川坡度等特性，利用碎形理論中自相似性及尺度不變性之概念加以建構，並將碎形生成以隨機化之機制來進行繁衍。由於隨機性之加入將使擬生成之碎形圖形具變異性，因此研究中以一套嶄新之編碼方式，將圖形中各分段予以數碼化，並藉由碎形之特性於編碼中有效地找出河網型態及河川特性，進一步將求得之參數代入以運動波概念為主軸之地貌型瞬時單位歷線架構中。 研究中藉由不同河網生成之基本型態，以模擬土地特徵及利用之改變，並分析此變化造成降雨—逕流歷程之影響。最後選用基隆河流域五堵集水區作為模式應用區域，探討以隨機碎形河網所得之河網型態與河川特性於水文模擬之適用性。 研究結果顯示，若改變河川基本型態或地貌參數如：主流變短、集水區面積減小、坡度越陡、河網密度增大、曼寧係數縮小等情形，皆會造成瞬時單位歷線之峰值增加及洪峰到達時間前移之效應，其中又以漫地流曼寧係數、河川坡度及河川長度等因素影響較大，集水區面積與河渠流曼寧係數等因素影響較小。而於五堵集水區之套配中可看出，本模式僅需概略假設集水區之河川特性，蒐集少許地文資料，即可有效地模擬該區域之水文狀況。本研究所採用之方法對資料不足之集水區，亦可提供一種有效且可快速地推估集水區河川特性及水文歷程模擬之方法。

The purpose of this study is to utilize fractal theory includes characterized by self affinity and invariance to scale constructing shapes and features of river networks. By using random generation of fractal, the pattern, length, watershed’s area, slope of river network can be created easily. Random generate process will cause the results of fractal pictures be highly varied, for this reason a brand-new coding method is applied. This mode will derive shapes and features of river networks effectively. To analyze the impacts of land use changes on rainfall-runoff process, different fractal basic patterns are applied to simulate omnigenous terrain features or developments. Finally, Wu-tu watershed is selected to verify suitability of this model. The results suggest when change occurs in fractal basic patterns or terrain’s parameters like: shorten main stream length, diminish watershed’s area, increase slope, enlarge probability of fractal generation, and decrease Mannning’s coefficient will all advance the time of concentration and grow in quantity of flood peak. In contrast to the above-mentioned factors, Mannning’s coefficient of overland, slope and main stream length have more effect than watershed’s area and Mannning’s coefficient of channel. Appling the model in Wu-tu watershed reveals this model only need to determine the fractal basic pattern of watershed and collect some informations of terrain features, then it can simulate the rainfall-runoff process effectively and quickly. This model is of great use to a project watershed with insufficient data.