集水區邊界自動劃定之評估與改善

Abstract

The watershed delineation is the basis of hydrology routing andflood control engineering. In order to reduce the disadvantage of delineating watershed by manual option and save the time of dataprocessing, the DTM of Liyutan watershed and Chungkongshi watershedwere used in this research to investigate the problem that may cause inwatershed automatic delineating with image processing softwareEASI/PACE. The results of watersheds which were delineated by (1)automatic delineation approach via computer, (2) by manual operationwith the topographic map and by field check with GPS were comparedaccordingly to discuss their accuracy. Four space processing methodswere utilized to modify the watershed boundary delineation with computer. In addition, the different resolution of DTM which weremodified with resampling were used to process the watershed automaticdelineation. These were to investigate the effect that caused by the inaccuracy of the automatic delineation. The results was summarized as following:1.Even though there were some restrictions in using DTM to acquire the boundary of automatic delineated watershed,with some help of manual judgment we might increase its accuracy.2.The area error percent of automatic delineated watershed was far less than the perimeter,circularity ratio and compactness error percent.3.The error percent of perimeter,circularity ratio and compactness of automatic delineated watershed increase with respect to the circularity ratio and compactness ratio. The major reason was the difference of data format between raster and vector.4.In order to decrease the error of perimeter,circularity ratio and compactness of delineated watersheds, both commands-CLEAN and GENERALIZE of Arc/Info,attributed the most significant effect in converting raster format to vector format.5.Although it might reduce the area error percent by means of resampling with higher resolution of DTM data and delineating watershed boundary automatically,it might also increase the error percent of perimeter,circularity ratio and compactness simultaneously. So one should weigh its advantage and disadvantage when he wants to delineate the watershed boundary automatically.

集水區之劃定為水文演算、防洪工程等之基礎，為改善人為劃定集水 區之缺失及節省處理時間，本研究以鯉魚潭集水區及中港溪集水區之DTM ，利用加拿大PCI公司出版之影像處理軟體EASI/PACE進行集水區自動劃定 之研究。首先將由電腦自動劃定所得的集水區，與由人為依地形圖判釋所 得之集水區相互比較，並利用GPS進行現場檢測，並以此探討自動劃定過 程可能產生的問題及自動劃定結果的準確性。其次將自動劃定集水區所得 網格式之資料轉換成向量式後，分別以四種方法加以處理，並比較其改善 誤差的程度。最後利用再取樣方式改變DTM的解析度，以不同解析度的DTM 再次進行集水區自動劃定，比較其劃定結果以探討提高或降低DTM解析度 對集水區自動劃定所生誤差率的影響。茲將研究結果摘述如后:1.以DTM為 材料進行自動劃定集水區邊界雖有其限制，然若輔以一 些人為之判釋， 即可提高其效率與準確度。2.以DTM自動劃定所得之集水區面積誤差率， 遠小於周長、圓比值、 密集度等之誤差率。3.集水區自動劃定所得的周 長、圓比值興密集度等之誤差率，由於 網格式與向量式資料間格式的差 異，會隨著集水區圓比值與密集 度的增加而增加。4.對於改善自動劃定 所得集水區周長、圓比值與密集度等之誤差 率，以Arc/Info之指令 Clean 及Generalize的處理最具顯著效 果。5.以再取樣方式提高DTM的 解析度再進行集水區自動劃定，雖可減 少自動劃定集水區的面積誤差率 ，但相對地卻會增加周長、圓比 值與密集度等的誤差率，宜斟酌為之。