Neuro-fuzzy classification of submarine lava flow morphology on the Galapagos Spreading Center, 92 degrees W

Variations in lava flow morphology along the axis of the Galapagos Spreading Center can indicate differences in eruption and emplacement dynamics due to enhanced magma supply from the adjacent Galapagos hotspot. Unfortunately, the ability to discriminate fine-scale lava morphology is usually limited to the small coverage areas of towed camera surveys and submersible operations. This research presents a neuro-fuzzy approach to automated seafloor classification using high-resolution multibeam bathymetry and side-scan backscatter data. The classification method integrates a fuzzy inference system in an adaptive neural network and is capable of rapidly classifying seafloor morphology based on attributes of surface geometry and texture. The system has been applied to the 92°W segment of the western GSC in order to quantify the abundances and distributions of lava flow morphology types. The classified maps reveal the study area terrain consists of approximately 47% pillow lava flows, 31% lobate lava flows, 12% sheet lava flows, and 10% faulted/fissured areas. An accuracy assessment shows the classification has an overall accuracy of 88.4% with a kappa coefficient of 0.84. Analysis of the distribution and abundance of lava flow types shows that low effusion rate pillow lava eruptions are most abundant at the end of the segment, while high effusion rate sheet lava flows are most abundant near the middle of the segment. Changes in lava distribution are directly correlated with the depth to a steady-state axial magma chamber below the ridge.