As water is considered one of the essential assets of nature, the recognition of the availability of water at a specific location can help government bodies to take necessary action toward water conservation. Monitoring water from satellite images is considered one of the most difficult areas of pattern recognition. In this manner, a novel multi-level feature fusion approach is proposed to predict the pattern of water concerning a specific location to analyze the scale and availability. The proposed framework can access the spatial features from sentinel-2 images by utilizing the concept of structural learning. For evaluating the prediction performance, the calculated outcomes are compared with the traditional and modern pattern recognition approaches. It has been observed that the proposed approach is more robust in terms of pattern analysis as compared to the state-of-the-art approaches. Moreover, the performance of the proposed approach is evaluated on different training and testing ratios such as 70:30, 75:25, and 80:20. In this manner, the calculated outcomes define the pattern recognition efficiency of the proposed approach over the state-of-the-art approaches by achieving 94.51% of accuracy.