Most ultrasonic transmitters (pingers) used for behavioral surveys of aquatic animals send a variety of information by changing the interval of the pulses. The reflection of a pulse generally occurs at the sea surface and/or the bottom. Reflection is a critical problem in considering the measurement error generated by the arrival-time delay of the reflected pulse. We conducted a field experiment to evaluate the effect of reflection, and discussed ways to prevent the effect. In this paper, we used a tracking-type ultrasonic telemetry system. The pinger of the system measures its depth by changing the interval of two pulses. In order to evaluate the paths of the reflected pulses, we measured the difference in arrival time between a direct pulse that propagated the direct path to the hydrophone and reflected pulses. We observed the variation in arrival-time difference in field experiments. Results showed that reflected surface pulses easily affected the system. Improvements to the ultrasonic telemetry system helped to prevent the effect of reflection, and we demonstrated the effectiveness of the improved system in an additional field experiment. The improved system could measure the depth data without the effects of reflection. We also discuss a more optimal configuration for preventing the measurement error from the additional experimental data.

Automatic acoustic detection is a useful method for monitoring phonating marine animals. Fin whale calls consist of frequency down-sweeps at around 20 Hz with a duration of approximately 1 s. These low frequency acoustic data were recorded with four hydrophones located on the ocean floor off Kushiro and Tokachi, Hokkaido, Japan. The fin whale calls were automatically detected by calculating the ratio of signal energy to background noise and the duration of the calls. With respect to manually detected calls, correct detection and false alarms using the proposed method were 69.5% and 4.4%, respectively. A total of 2,865,403 calls were detected from data collected during 2,981 days from January 2007 to February 2015. The number of detected calls showed clear seasonal differences, high in October to February period.