| 摘要: | 在眾多自然災害中，洪水因其頻繁發生及其對社會和環境的嚴重影響而尤為突出。本研究旨在通過整合遙感資料、機器學習技術和CMIP6全球氣候模型（GCM）資料，為印尼德馬克地區（Demak District）開發一個洪水潛勢易淹的模型。研究方法包含使用Sentinel-1 SAR資料作為洪水清單，繪製當前洪水易淹區的地圖，並應用MLP-NN、隨機森林、支援向量機（SVM）和XGBoost等機器學習演算法來預測洪水易淹地區。此外，基於CMIP6 GCM的降水資料，針對三種共用社會經濟路徑（SSP）情景（SSP1-2.6、SSP2-4.5和SSP5-8.5）下，預測了2021年至2100年期間未來洪水的易發性。為了提高未來預測的可靠性，研究採用了多模型集成（MME）方法，結合了多個GCM的輸出結果，其中 MME 的 GCM 選擇基於對稱不確定性 (SU) 和評級指標 (RM)，以確保使用最可靠的模型來降低模型不確定性。結果顯示，XGBoost 模型的預測能力最佳，AUC 為 0.9291，其次是隨機森林、SVM 和 MLP-NN。分析的結果易呈現，洪水易發的敏感性對於排放情形有顯著增加，尤其是在較高排放情景（SSP5-8.5）下，極高敏感性區域將從當前時期的 16.67% 增長到 2081-2100 年的 27.43%。洪水敏感性的增加與預測的降水量增加相一致，尤其是在較高排放路徑下，如 MME 所預測之結果。整體的研究結果證實了整合 Sentinel-1 SAR 圖像用於生成洪水清單對於克服資料限制和提高模型準確性至關重要。本研究建構了一個預測洪水風險的預測模式，並適用於不同的排放情境，對於德馬克地區的洪水預防管理的策略具相當的參考價值。另一方面，研究結果亦強調，潛勢易淹的地區需要進行自我調整空間規劃，以應對氣候變化情景下日益增長的洪水風險，尤其是在高排放路徑下。研究結果同時顯示採用對稱不確定性和評級指標的多模型集成方法對於改進洪水敏感性預測和減少洪水風險評估中的不確定性的重要性。;Among various natural hazards, floods stand out due to their frequency and severe impact on society and the environment. This study aimed to develop a flood susceptibility model for Demak District, Indonesia, by integrating remote sensing data, machine learning techniques, and CMIP6 Global Climate Models (GCMs) data. The approach involved mapping current flood susceptibility using Sentinel-1 SAR data as the flood inventory and applying machine learning algorithms such as MLP-NN, Random Forest, Support Vector Machine (SVM), and XGBoost to predict flood-prone areas. Additionally, future flood susceptibility was projected using CMIP6 GCM precipitation data under three Shared Socioeconomic Pathways (SSP) scenarios (SSP1-2.6, SSP2-4.5, and SSP5-8.5), covering the 2021–2100 period. To enhance the reliability of future projections, a multi-model ensemble (MME) approach was employed, combining the outputs of multiple GCMs. The selection of GCMs for the MME was based on Symmetrical Uncertainty (SU) and Rating Metric (RM), ensuring that the most reliable models were used to reduce model uncertainties. The results showed that the XGBoost model demonstrated the best performance with an AUC of 0.9291, followed by Random Forest, SVM, and MLP-NN. The analysis revealed a significant increase in flood susceptibility, especially under higher emission scenarios (SSP5-8.5), with very high susceptibility areas growing from 16.67% in the current period to 27.43% by 2081–2100. This increase in flood susceptibility corresponds with projected increases in precipitation, particularly under higher emission pathways, as seen in the MME projections. The integration of Sentinel-1 SAR imagery for flood inventory generation proved essential in overcoming data limitations and enhancing the accuracy of the model. In conclusion, this study provides a robust framework for predicting flood risks, which is crucial for guiding flood management strategies in Demak District. The results underscore the need for adaptive spatial planning to address growing flood risks under climate change scenarios, especially in high emission pathways. The findings also highlight the importance of multi-model ensemble approaches, with Symmetrical Uncertainty and Rating Metric, to improve flood susceptibility projections and reduce uncertainties in flood risk assessments. |
