Additional file 3: Table S1. of Coupling spectral analysis and hidden Markov models for the segmentation of behavioural patterns

Parameter values conditional to behavioral states for the autoregressive process component of the simulated depth time series. Table S2. Parameter values conditional to behavioral states for the periodic signal component of the simulated depth time series. Table S3. Confusion matrix for cross-validation between the simulated known states and the HMM estimated states. Table S4. Depth ranges for each 5 behavioural states. Table S5. Mean values and standard deviations of NNMF factors and SlpLog-Log variables per HMM states. Figure S1. Spectral signature and activity levels of movements between 6 and 72 h ( orange and blue dotted lines indicate diurnal and tidal periodicities, respectively), averaged over time (A); and an index of movements randomness and activity levels (B). Individual #A11325 tagged at La Turballe. Figure S2. The optimal number of factorization ranks (red dotted line) of the NNMF analysis based on the cophenetic coefficient (A) and the RSS curve (B). Figure S3. NNMF outputs obtained fromperiodograms between 6 and 72 h for all individuals and sites pooled together. Periodograms associated with each factor of the selected 9-dimensional NNMF (#a). Coefficients time series of the NNMF decomposition of the daily periodograms (#b). Figure S4. Mean normalized periodogram (S6-72 h) associated with each behavioural state inferred from HMM ran with three (A) to ten (H) latent states. Figure S5. Model selection criterions: BIC (A), model entropy (B) and ICL (C). The red dotted line indicates the five-states HMM we retained. Figure S6. Known (A, C, E) and estimated (B, D, F) behavioural states for three simulated individual series with different state switching dynamics (Table S1). Figure S7. Spectral signature and activity levels associated to each behavioural states of the fitted three-state HMM for all simulated individuals pooled together. The orange and blue dotted lines indicate diurnal and tidal periodicities, respectively. (DOCX 1349 kb)