Abstract: Identifying factors influencing survivorship is key to understanding population persistence. Although satellite telemetry is a powerful tool for studying remote animal ecology and behaviour it is rarely used for demographic studies because distinguishing the death of the animal (individual mortality) from failure of the tag (mechanical tag failure) has proven difficult. Southern elephant seals present an opportunity to separate tag failure from animal mortality thanks to the availability of large tracking datasets, broad knowledge of demographic rates, and because for these large animals, satellite tags are known not to influence mortality rates. A key rationale for investigating satellite telemetry to estimate mortality as compared to using traditional Capture-Mark-Recapture methods is the potential for obtaining spatially and temporally specific information, particularly while the animals are at sea and largely unobservable. We used satellite tag data from 182 seals from Isles Kerguelen, deployed between 2004 and 2018. Of these, 76 (42%) tags transmitted for the full post-moult foraging trip (max. 265 days for females and max. 305 days for sub-adult males) with the remaining 107 tags (58%) ceasing transmission at sea. We found that contrary to expectations, behavioural choices seem not to influence tag failure rates by mechanical means, rather the signals we detected seemed to align with previously described variation in mortality between groups. There was evidence, albeit limited, for an increase in tag failure for adult females in years with negative Southern Annular Mode (lower Southern Ocean productivity). We speculate that this increase in failure may suggest higher mortality in these years. Also, males using the Kerguelen Plateau had higher tag failure rates than those in the sea-ice zone, perhaps indicative of higher mortality. We suspect that these differences in tag failure rates between groups reflect variation in predator exposure and foraging success. This suggests satellite telemetry could be used to infer mortality events for southern elephant seals while they are at sea.

Abstract: The diving behavior of southern elephant seals, Mirounga leonina, is investigated through the analysis of time-depth dive profiles. The originality of this work is to consider dive profiles as continuous curves. For this purpose, a Functional Data Analysis (FDA) approach is proposed for the shape analysis of a collection of dive profiles. Complexity of dive shapes is characterized by a mixture of three main shape variations accounting for about 80% of the entire variability: U or V shape, vertical depth variability during the bottom time, and skewed left or right. Model-based clustering allows the identification of eight dive shape clusters in a quick and automated way. Connection between shape patterns and classical descriptors, as well as the number of prey capture events, is achieved, showing that the clusters are coherent to specific foraging behaviors previously identified in the literature labeled as drift, exploratory and active dives. Finally, FDA is compared to classical methods relying on the computation of discrete dive descriptors. Results show that taking the shape of the dive as a whole is more resilient to corrupted or incomplete sampled data. FDA is, therefore, an efficient tool adapted for processing and comparing dive data with different sampling frequencies and for improving the quality and the accuracy of transmitted data.

Abstract: Machine learning is becoming increasingly important in scientific and technological progress, due to its ability to create models that describe complex data and generalize well. The wealth of publicly-available seismic data nowadays requires automated, fast, and reliable tools to carry out a multitude of tasks, such as the detection of small, local earthquakes in areas characterized by sparsity of receivers. A similar application of machine learning, however, should be built on a large amount of labeled seismograms, which is neither immediate to obtain nor to compile. In this study we present a large dataset of seismograms recorded along the vertical, north, and east components of 1487 broad-band or very broad-band receivers distributed worldwide; this includes 629,095 3-component seismograms generated by 304,878 local earthquakes and labeled as EQ, and 615,847 ones labeled as noise (AN). Application of machine learning to this dataset shows that a simple Convolutional Neural Network of 67,939 parameters allows discriminating between earthquakes and noise single-station recordings, even if applied in regions not represented in the training set. Achieving an accuracy of 96.7, 95.3, and 93.2% on training, validation, and test set, respectively, we prove that the large variety of geological and tectonic settings covered by our data supports the generalization capabilities of the algorithm, and makes it applicable to real-time detection of local events. We make the database publicly available, intending to provide the seismological and broader scientific community with a benchmark for time-series to be used as a testing ground in signal processing.