Résumé
Recently marine organisms have emerged as ocean-sensing platforms, generating data useful to biologists and oceanographers. However, sensor calibrations, performance over time, and effects of behavior and sampling frequency on data quality have not been adequately examined. We performed temperature calibration trials on 36 Wildlife Computers Mk9 time-depth recorders (stated accuracy 0.1 degrees C +/- 0.05 degrees C) versus Rosemount Model 162CE standard platinum resistance thermometer (0.001 degrees C, +/- 0.0001 degrees C). Sixty-four percent of trials were within +/- 0.1 degrees C. Subtracting 0.05 degrees C from all calibrations brought this to 83%. Six instruments calibrated before and after deployment on free-ranging pinnipeds, showed no significant drift in temperature measurements. Mk9 performance was tested against a Seabird (c) CTD (SBE-19). Root mean square (rms) difference between CTD and Mk9 temperature was 0.15 degrees C (max. 0.45 degrees C, min. 0.10 degrees C). Applying a 1-s time-lag improved the data fit. To assess animal effects on thermal data, Mk9s were deployed on juvenile elephant seals released between CTD casts. Overall rms were within +/- 0.1 degrees C. Subsampling of Mk9 data revealed a sampling frequency of 10 s was sufficient to accurately resolve thermal features, e. g., thermocline depth, for these CTD experiments. We present four equations to assess temperature data quality from diving marine animals around the globe carrying different thermistors.