Maritime remote sensing is an interdisciplinary field situated at the intersection of naval hydrodynamics, physical oceanography, and sensor system physics. The expanding information available from sensors fixed on ocean environments brings an ever-increasing demand for new analysis and data fusion methodologies. However, the data sets are often provided without ground truth or complete specification of the measurement scenario. Data quality issues pose challenges when trying to extract whether or not a particular feature is present in the collected data, or whether a particular approach represents an improvement or not. This study provides a foundational methodology for fusing information from multiple data sets. The research implements neural network classifier, trained on simulated synthetic aperture radar (SAR) data sets. The SAR data are generated using existing modeling and simulation capability developed by the authors. We use mutliple SAR modalities represent the concept of various sensors. The classifier is used to identify signatures within an input SAR image. We implement data fusion on the feature level and the decision level. Feature-level fusion, where features are extracted from disparate data sources separately then fused together for analysis, grants no strong performance benefits, but this methods shows promise for more complex implementations. Decision-level fusion, where each data source is analyzed independently and conclusions drawn from each analysis before a final, aggregate classification is made, shows stronger classification performance improvements. Potential improvements for future work are also outlined.