Satellite imagery utilization for water surface monitoring faces challenges in the presence of clouds that interfere with the segmentation process. This study aims to overcome the problem by implementing and developing the Multi-temporal Cloud Removal MCR method at the pre-processing stage. This research uses (MCR) approach to produce cloud-free satellite images. The difference from the current MCR method comes from using a red band in the cloud detection step, adding a dilation process on the cloud masking result, and the pixel mapping process. The best MCR results are obtained from cloud cover percentages below 20% with an NRMSE value of 0.156 and a PSNR of 29.815. In addition, this research compares the accuracy of the U-net water segmentation model between the original satellite image and the resulting MCR data. The model evaluation process uses binary cross-entropy loss, dice coefficient, and intersection over union (IoU) to check the model's performance. The model achieved a Dice coefficient of 0,842, IoU of 0.913 and a binary cross-entropy loss of 0,137 for MCR images. For images with cloud cover below 20%, the Dice coefficient and IoU slightly decreased to 0.819 and 0.892 (loss: 0.1563). As cloud cover increased to 20-40%, these values dropped to 0.687 and 0.812 (loss: 0.25) and further decreased to 0.546 and 0.706 (loss: 0.2656) for 40-60% cloud cover. The results demonstrate that the percentage of cloud cover in satellite images significantly influences the accuracy of surface water segmentation. © 2024 IEEE.