A methodology for detecting events within time series data based on a model of normal behavior has enabled the detection of storm effects within the usual tidally driven plume bending time series. The event detection method uses a graph-based modeling technique to detect the difference between normal or background behavior and events (Ozer et al., 2013). A state-based model specifies a behavior pattern; then the method matches every time step with a stage in the pattern and reports time sequences where the desired pattern is found. <br /><br />An extended and ongoing sequence of 3D acoustic images has been collected by the Cabled Observatory Vent Imaging System, or COVIS, since its installation in September 29, 2010. The plumes above the clusters of smokers on Grotto mound, Main Endeavour Field, Juan de Fuca Ridge, are identified by increased backscatter intensity; a skeleton-based centerline algorithm estimates the path of the plume cores and yields an overall estimate of orientation changes. While most of the variability in the orientation of the plume centerline likely stems from the tidal currents local to Grotto, there are many instances where a) the amount of change in direction over the 3 hour sampling interval is smaller (<30°) or larger (>120°) than that predicted by tidally-driven models (e.g. ~90° over 3 hours based on the anticipated 180° over 6 hours) and b) little change in direction and/or amount of bending over an extended time interval. The event detection method is applied to identify what other phenomena are occurring. <br /><br />Two basic models have been implemented for the COVIS plume centerline orientation data: 1) a simple sloshing model that looks for failures to follow a regular pattern of change and 2) a storm model that looks for extended periods of consistent orientation of the plume centerline. A variety of models are tested for stability and relevance. The prediction of storm events will be compared with sea surface buoy records and land records of storm events. The predictions and successful models of bending in response to normal tides will be compared with other measures of the Grotto vents to improve understanding of the normal variation in this system. Figure 1 shows a detected activity that may reflect internal waves generated by an atmospheric storm and propagated downward through the deep ocean.
Bemis, K., Ozer, S., Xu, G., Silver, D., & Rona, P. A. (2013). A Method for Detecting the Impact of Atmospheric Storms on Hydrothermal Systems. In American Geophysical Union Fall Meeting 2013. San Francisco: American Geophysical Union.