不同流向演算法與集流閥值對數值地形模型河川網路萃取之影響

Abstract

一般傳統水系萃取透過人工測繪具寬度之河川網路，雖具一定準確性，但作業緩慢耗時。利 用網格式數值地形模型萃取集水區河川網路，較人工測繪快速有效率，但其正確性因流向演算法 與集流閥值設定而異。 本研究以台北市雙溪流域為試區，使用5 m 網格解析度數值地形模型，配合6 種流向演算法 與12 等級集流閥值，進行河川網路自動化萃取，並對照1/5,000 地形圖之河流線，輔以人工判釋， 檢核其網格空間對位誤差。經整合演算結果發現，多流向演算法與無限流向演算法在1 級河與2 級河部位造成誤授比率明顯高於其他演算法，並有斷流現象，但卻能有效凸顯下游外雙溪部份之 河寬表現；集流閥值設定主要影響河川長度與支流數，對河寬影響不大，但其值增加則能有效降 低多流向演算法之誤授率。整體而言，不同流向演算法主要差異在於河寬推估，河川中心線彼此 間差異極小，但與人工測繪基準圖層仍普遍存在1 至8 個網格之河道對位誤差。綜合評估結果， 特定的流向演算法因其在河川網路上下游表現之優劣不同，後續研究建議依坡度與級序分段採用 不同之流向演算法整合運算，將有效提升水系萃取準確率。

Traditionally, stream networks were delineated manually in hydrologic analysis. Though certain degree of accuracy can be achieved, the process requires large amount of time. In comparison, deriving stream networks from grid based digital terrain model (DTM) is more efficient than the manual process. However, the accuracy of the results varies with different flow direction algorithms and threshold values. This study selected the Shuangshi watershed as the study site in Taipei, and utilized digital terrain models with grid size 5m to derive stream networks automatically. Six different flow direction algorithms and twelve threshold values were applied in this study, and the resultant stream networks were manually compared to the stream networks found on 1/5,000 topographic map. The results indicate that: (1) The commission error is much higher in 1st order stream and 2nd order stream, and more effective on stream width when Multiple-flow direction algorithm and D-infinity flow direction algorithm were applied; (2) Different threshold values have more effective on the stream length, and the commission error decreases as the threshold value increases, where the effects on stream width are not prominent; (3) Overall, though the difference between flow direction algorithms is mainly in deriving the width of streams, and the extracted stream centerlines are similar, 1~8 grid-length bias remain existed when comparing to the stream networks shown on the topographic maps. The results obtained from the Shuangshi watershed show that the performance of flow direction algorithms is different when stream order changes. For accurate estimation of stream width, further investigations are needed.