Accurate assessment of underwater acoustic propagation relies heavily on understanding variability in the waveguide. Typically, prevailing methods in the field use environmental data exclusively to evaluate water column fluctuations. This study breaks new ground by integrating a graph-based approach, merging insight from acoustics and environmental data to comprehensively analyze ocean fluctuations. The proposed framework leverages data obtained during the Shallow Water ‘06 Experiment (SW06) from a densely measured region in the experimental area. The data used in this work include acoustic broadband signals below 1 kHz from a 48-element L- shape array and environmental measurements from 59 thermistors across 16 mooring arrays, affected by the passing of internal waves. The spatial relationship in the data, facilitated by sensor locations, is harnessed through a graph built upon underlying features from the datasets, providing a nuanced understanding of environmental variability in the water column. The proposed framework is not confined to theoretical constructs; it is substantiated through rigorous model evaluation on the measured data. This validation process robustly demonstrates the generalization power of the proposed architecture, affirming its effectiveness in a highly fluctuating water column and showcasing its potential for advancing underwater acoustic research. [Work supported by ONR Ocean Acoustics program.]