Flood mapping is an essential activity for assessing and responding after disasters, particularly in areas impacted by natural calamities. This study introduces a deep learning model for post-disaster flood mapping utilizing Sentinel-2 satellite images, employing UNet Attention for segmenting water bodies and SNUNet-CD for detecting changes. The system effectively identifies and illustrates areas impacted by flooding by examining Sentinel-2 images taken before and after the flood event. Our model demonstrates exceptional performance with an accuracy of 94.22%, an F1-score of 85.03%, and a Dice Similarity Coefficient (DSC) of 76.71%, surpassing current methods like UNet and FCN32s. The suggested model exhibits better performance in segmenting water bodies and detecting flood changes. The framework is scalable, adaptable to various datasets, and capable of integrating multimodal data, making it a valuable tool for disaster response, climate change analysis, and long-term environmental monitoring. The research also highlights future research directions, which include real-time optimization, integration with decision support systems, and advanced post-processing methods, with the goal of making the model more applicable and effective in addressing global challenges.