<![CDATA[Flooding is one of the most frequent hydrometeorological disasters in Indonesia, causing significant damage, with its frequency continuing to increase due to climate change and land-use alterations. This condition necessitates a rapid and accurate flood monitoring system. At present, two commonly used methods for flood mapping are thresholding and machine learning. This study aims to compare the performance of flood mapping methods based on SAR imagery using classical thresholding approaches and modern machine learning techniques. The thresholding method used is the Otsu thresholding method, whilst the machine learning approach utilises the XGBoost and LightGBM algorithms. The analysis results show that thresholding methods achieve higher accuracy in identifying flooded areas compared to machine learning methods in this case study. The overall accuracy of the Otsu method is 0.995. Meanwhile, the overall accuracy values of the XGBoost and LightGBM machine learning methods are 0.975 and 0.970, respectively. These findings indicate that thresholding approaches remain relevant and effective for flood mapping in Indonesia. However, the development of machine learning methods can also be a good alternative approach to flood mapping.]]>
Copyrights © 2026
