使用修改版Budyko框架研究氣候與景觀對水文分配的影響

Abstract

了解水–能量互動和集水區動態對於可持續水資源管理至關重要，尤其是在資料稀缺、氣候和地形複雜的地區。這篇論文探討了在修改後的Budyko框架（Choudhury–Yang方程）下的徑流組成動態，重點研究了埃塞俄比亞的集水區。它涉及兩個相互關聯的研究問題： (1) 蒸散指數短期偏離長期Budyko曲線如何影響徑流生成；以及 (2) 降水量(P)、潛在蒸發蒸散 (PET)和地形參數(n)的個別變化如何決定地表徑流 (SR)和地下徑流 (SSR)的組成彈性。 在第一項研究中，使用 SWAT 模型來模擬衣索比亞流域的水文過程，並將其校正後的輸出與長期改良的 Budyko 框架結合。該模型表現良好（NSE = 0.78），曲線數（curve number）、可用水容量以及地下水延遲被認為是主要參數。景觀參數 (n) 是從 Choudhury-Yang 方程式中最佳化的，其 86% 的變異性可由流域坡度和植被覆蓋比例解釋。年度蒸發指數相對長期 Budyko 曲線的短期垂直偏差，與地表逕流和回流呈顯著負相關，顯示其對惡劣氣候下乾旱和逕流評估具有診斷價值。 在第二項研究中，我們量化了降水 (P)、潛在蒸散 (PET) 和有效降雨日數 (n) 的變化對地表和次表層徑流彈性的貢獻，彈性定義為各驅動因子按比例變化時（子）表層徑流的百分比變化。我們選取了涵蓋完整乾旱度梯度的六個埃塞俄比亞子流域（1964–2023 年）。Mann–Kendall 和 Pettitt 測試將這 60 年的水文氣象記錄劃分為兩個不同時期以進行歸因分析。基於 ERA5-Land 再分析數據，與當地觀測和區域研究的驗證證實了 ERA5-Land 徑流估算的可靠性，皮爾森相關係數 r 值在 0.49 到 0.83 之間。趨勢與變點分析顯示，大多數子流域的 PET 上升，乾旱子流域的降水下降，而四個子流域的有效降雨日數減少，說明土地持續退化，反映了全球變暖和人類干擾的綜合影響。ERA5-Land/Budyko 模型分析顯示，SR 彈性主要受 P 控制（平均貢獻 49.3%），其次是 PET（26.2%）和 n（24.5%），強調了水資源可用性對其的主導影響，尤其是在乾旱子流域。SSR 彈性對 PET（33.6%）和 n（39.7%）更敏感，反映出能量供應和地形特性對慢流過程的更強控制。氣候驅動（P 和 PET 的結合）成為主要控制因素，根據 ERA5-Land/Budyko 歸因分析，平均佔流量彈性 66.7%。 實證證據顯示，景觀參數的下降會加劇氣候驅動的水文反應，而透過復原而增加的景觀參數則會強化景觀特徵對水文過程的影響。研究結果表明，經過改良的Budyko框架，並擴展納入短期時間偏差和組件特定的彈性，提供了一個可轉移的分析工具，用於理解不同氣候區域的水文變化。這些結果提供了新的見解，也證實了有效的水資源管理應該超越單純的總徑流方法，採用針對組件、兼顧流域特性的策略，依據各個流域獨特的氣候和景觀敏感性設計，並考慮全球資料受限地區的氣候變異與景觀動態。

Understanding water–energy interactions and catchment dynamics is crucial for sustainable water resources management, especially in data-scarce regions with complex climate and landscape characteristics. This dissertation explores the dynamics of runoff components within the modified Budyko framework (Choudhury–Yang equation), focusing on Ethiopian catchments. It addresses two interconnected research questions: (1) how short-term deviations of the evaporative index from the long-term Budyko curve affect runoff generation; and (2) how the individual contributions of changes in precipitation (P), potential evapotranspiration (PET), and the landscape parameter (n) govern elasticity of surface (SR) and subsurface runoff (SSR). In objective one, the SWAT model was used to simulate the hydrological processes in the Ethiopian watershed, and its calibrated outputs were integrated with the long-term modified Budyko framework. The model demonstrated satisfactory performance (NSE = 0.78), and the curve number, available water capacity, and groundwater delay were identified as primary parameters. The landscape parameter (n) was optimized from the Choudhury-Yang equation, and 86% of its variability was explained by catchment slope and fractional vegetation cover. Short-term vertical deviations of the annual evaporative index from the long-term Budyko curve were significantly and negatively correlated with surface runoff and return flow, demonstrating their diagnostic value for drought and runoff assessment in harsh climate situations. In objective two, we quantified the contributions of changes in P, PET, and n to SR and SSR elasticity, defined as the percentage change in (sub)surface runoff for a proportional change in each driver, across six Ethiopian sub-basins covering a full aridity gradient (1964–2023). The Mann–Kendall and Pettitt tests divided the 60-year hydrometeorological record into two distinct periods for elasticity analysis. Based on ERA5-Land reanalysis data, validation against local observations and regional studies confirms the reliability of ERA5-Land runoff estimates, with Pearson r values ranging from 0.49 to 0.83. Trend and change-point analyses revealed rising PET in most sub-basins and declining P in arid sub-basins, with decreasing n in four sub-basins indicating progressive land degradation reflecting the combined effects of global warming and human disturbance. ERA5-Land/Budyko model analysis revealed that SR elasticity was predominantly controlled by P (49.3% average contribution), followed by PET (26.2%) and n (24.5%), emphasizing its dominant link to water availability, particularly in arid sub-basins. SSR elasticity was more sensitive to PET (33.6%) and n (39.7%), reflecting the stronger control of energy supply and landscape properties on slow-flow processes. Climate forcing (P and PET combined) emerged as the primary control, accounting for an average of 66.7% of runoff elasticity based on ERA5-Land/Budyko attribution. Empirical evidence suggests that declines in the landscape parameter intensify climate-driven hydrological responses, whereas restoration-induced increases strengthen the influence of landscape characteristics on hydrological processes. The findings demonstrate that the modified Budyko framework, expanded to include both short-term temporal deviations and component-specific elasticities, provides a transferable analytic tool for understanding hydrological change across diverse climate regions. The results provide new insights and confirm that effective water management should move beyond total runoff approaches and adopt component-specific, catchment-aware strategies tailored to the unique climate and landscape sensitivities of individual catchments, accounting for both climatic variability and landscape dynamics across data-constrained regions globally.