As acoustic signals play a central role in cetacean ecology, passive acoustic monitoring (PAM) enables the investigation of acoustic presence and related behaviour of cetaceans based on known species-specific signals. Analysis of acoustic signals, whether clicks, pulses, tonal signals or song can yield insights into foraging behaviour, mating, group composition and coordination, as well as population and individual identity. Despite its potential, the effective application of PAM in ecological research depends critically on reliable signal identification, which becomes particularly challenging when signal characteristics overlap substantially. In the Southern Ocean, where multiple baleen whale species co-occur and produce acoustically similar signals, these challenges remain largely unresolved, yet are critical given the limited applicability of alternative monitoring methods. In this PhD thesis, I evaluate the potential and limitations of PAM for studying baleen whales in the Southern Ocean, given persistent challenges in signal interpretation. This includes assessing the reliability of species-level classification of non-song calls, with a focus on southern right (Eubalaena australis) and humpback whale (Megaptera novaeangliae) upcalls and frequency-modulated calls of blue (Balaenoptera musculus) and fin whales (B. physalus), as well as examining Southern Hemisphere fin whale song characteristics, particularly high-frequency (HF) components, and their suitability for population identification. In Chapter 1, vocalisations comparable to southern right whale upcalls detected off Elephant Island were structurally analysed to confirm species identity. Call characteristics were compared with confirmed southern right and humpback whale vocalisations. The detected upcalls were attributed to southern right whales, with time- frequency slope and bandwidth of the vocalisation identified as the key parameters distinguishing the two species. Chapter 2 combines long-term PAM datasets with animal-borne tag recordings to evaluate whether blue and fin whale frequency-modulated calls can be reliably distinguished at the species level. While the call parameter Duration 90% provides partial separation, a more robust approach was achieved by combining deep-learning-based feature extraction with non-linear dimensionality reduction, allowing calls from long-term datasets to be projected into an embedding space derived from the tag data. In Chapter 3, region-specific HF components of Southern Hemisphere fin whale song are evaluated as acoustic cues to distinguish and monitor fin whale acoustic populations in the Atlantic Sector of the Southern Ocean (ASSO). Passive acoustic data from ten recording positions are used to assess the spatio-temporal distribution of the 86- and 99-Hz HF components present in this region. Results show that while the 99-Hz component was detected at seven recording positions throughout the ASSO, the 86-Hz component was restricted to the western part of the study area, centred around the Western Antarctic Peninsula. Chapter 4 further investigates the use of HF components of Southern Hemisphere fin whale song as reliable acoustic markers for monitoring fin whale acoustic populations, given the indication of frequency declines. Analysis of multi-year passive acoustic datasets from two recording sites in the ASSO, Elephant Island and the Greenwich Meridian, shows that despite gradual interannual and intra-annual variability, these song features remain distinct and recognizable across regions and years, providing a robust cue to identify acoustic populations. Overall, the findings and frameworks I developed in this thesis support more reliable passive acoustic monitoring by improving interspecific call discrimination and enabling robust identification of fin whale acoustic populations, thereby strengthening confidence in acoustic assessments of species presence and population structure in the Southern Ocean. At the same time, methodological limitations in regard to call classification are critically evaluated, including constraints arising from transmission loss effects and deep-learning-based feature extraction, along with potential approaches to overcome these limitations. The thesis further considers how additional baleen whale FM calls could be incorporated into the developed frameworks, as well as future challenges in call classification and possible drivers of similarity among non-song calls across baleen whale species. In regard to song characteristics for population identification in Southern Hemisphere fin whales, two additional features, inter-note intervals and song variants, were examined. Both correspond to the results from the HF components, and all three characteristics were evaluated in terms of their applicability, reliability, and potential for monitoring, including how they could contribute to a circumpolar perspective on fin whale populations. Finally, this thesis emphasises that while standardised analytical frameworks are essential for reliable and comparable PAM, human expertise remains indispensable for detecting changes in signal structure, validating results, and ensuring biological relevance.