GWLF平衡模式因其簡易性與便利性而廣泛被應用在相關領域。然而此類水文模式較少考慮集水區中集流時間對流量所造成的流量延遲效應，以至於部分大型集水區日流量模擬中，部分模擬流量洪峰發生時間先於實際流量洪峰發生時間，造成模擬效率降低。 本研究利用集流時間與日雨型建立一能代表集水區延遲效應之參數，分別使用日雨量(方法一)與時雨量(方法二)兩種修改方法。首先使用日雨量進行修改(方法一)。集水區集流時間藉由Rziha公式決定，一日降雨型態以穩態降雨假設。綜合上述結果可得集水區之延遲係數(Lag coefficient, Lc)，此參數可將每日地表逕流分成該日流出集水區與隔日流出兩部分。 其二，使用時雨量資料進行修改(方法二)。使用集水區平均集流時間(算法同上)結合當日雨量決定當日之延遲比例，進而決定當日地表逕流之延遲量。不同於日雨量修改，時雨量資料有能力決定每日之延遲比例(某日應被延遲之雨量/某日總雨量)，而此延遲比例會隨每日降雨型態而改變。 本研究針對1989-2005年翡翠水庫集水區、石門水庫集水區、曾文水庫集水區、蘭陽溪流域中水文資料進行延遲係數建立與模式檢定。模式檢定結果，原始模式日流量模擬相關係數R2分別為0.72、0.76、0.77、0.51，而方法一修改後R2分別上升至0.80、0.91、0.78、0.62。因時雨量資料有限，故時雨量修改只針對石門水庫集水區與曾文水庫集水區作探討。檢定結果R2分別從0.76與0.77上升至0.94與0.78，且在個別事件上較方法一更能校正模擬與觀測的洪峰時間差異之問題。 GWLF (Generalized Watershed Loading Functions) model is known for its clear concept and simplicity and is widely applied to flow simulation. However, one disadvantage of lump parameter hydrological model is that it does not take into consideration the time of concentration therefore it usually provides low efficiency for each simulation. In this study, a concentration time of 24-hour rainfall pattern was applied to establish a coefficient which assisted in representing the surface lag effect in watersheds. First, daily rainfall data is used for modifying the model (Method 1). The time of concentration is calculated by the Rziha formula. 24-hour rainfall pattern is assumed to be in a state of default. The results that are obtained are combined to produce the lag coefficient. Daily discharge can be separated into two parts daily out flow from the watershed and daily lag in the watershed. Second, hourly rainfall data is then used to modify the model (Method 2). The formula mentioned above was used to calculate the time of concentration. The daily lag coefficient is determined by hourly rainfall data while the coefficient varies on a daily basis. Four watersheds (Feitsui Reservoir, Shimen Reservoir, Zengwen Reservoir, Lanyang River) were selected. 17 years of hydrological data was collected for each watershed between the years 1989 to 2005. The (R-square) results show great improvement with the modification of GWLF model (method 1) from 0.72 (Feitsui Reservoir watershed), 0.76 (Shimen Reservoir watershed), 0.77 (Zengwen Reservoir watershed), 0.51(Lanyang river watershed) to 0.80, 0.91, 0.78, 0.62 respectively. The results obtained from the second method show further improvements. The results improved from 0.91 (Shimen Reservoir watershed) to 0.94 and 0.77 (Zengwen Reservoir watershed) to 0.78 which proved to be much more accurate in verification and single events.
