The communication of African and Asian elephants based on seismic and acoustic waves has been studied for decades. However, research within anthropogenic zoo environments, particularly with respect to seismic signals, remains limited compared to studies in natural habitats. This study analyses low-frequency elephant rumbles recorded at the Opel-Zoo near Frankfurt am Main, Germany, by comparing characteristics from datasets obtained using non-invasive, co-located seismic and infrasound sensors. Analysis of recordings from August 2024 revealed over 1350 rumbles, indicating significant temporal variability. These rumbles are characterized by signal durations of 1–8 seconds and fundamental frequencies between 10 and 25 Hz, with harmonics above. Due to high seismic background noise during zoo opening hours, infrasound detections are more abundant during the day, while seismic and infrasound detection rates are comparable at night. Our results show that the number of rumbles significantly increases approximately every second night, in agreement with the elephants changing nocturnal housing schedule, with one pair of elephants being much more communicative than the other pair. We find indication for a diurnal correlation of rumble activity with visitor numbers. Many rumbles occur in rapid sequences within minutes, suggesting elephant interaction or external triggers. Most rumbles are accompanied by motion-induced signals associated with locomotion or trampling, phenomena not detectable with infrasound sensors measuring acoustic waves only. This highlights the value of combined seismic and infrasound data. To enable a robust automated classification of rumbles and noise for continuous monitoring, we train and test two CNNs using spectrogram images of the hand-picked seismic and infrasound rumbles as inputs. The model achieves up to 98% classification accuracy, while cross-domain applications demonstrate better generalization and robustness of the CNN trained with seismic data. The seismo-acoustic monitoring approach and resulting findings have the potential to enhance our understanding of zoo elephant behavior, social interactions, and welfare.

On 23 November 2025, the Hayli Gubbi volcano in the Danakil Depression (Afar, northern Ethiopia) erupted in a massive explosion. We present the first seismic characterization of this event using data from ten regional seismological stations. Seismic records reveal a complex eruption sequence consisting of a minute-long precursor, a main explosive phase, and secondary events hours later. Two distinct surface-wave trains indicate that the main phase comprised two major explosions separated by minutes rather than a single event. The explosion sources were localized by jointly inverting epicentral coordinates and wave velocity from surface-wave arrivals. Waveform modeling supports an explosive sequence, as observed signals are reproduced only when two sequential sources are assumed. Pulse-like precursor signals minutes before the main explosions indicate pressurization, conduit processes, or magmatic fracturing prior to failure. This exceptional event provides new insights into the eruption and highlights the need for improved local monitoring in the region.

Abolfazl Komeazi*, Goethe University Frankfurt, Geosciences, Geophysics, Altenhöferallee 1, Frankfurt, Germany, E-mail: komeazi@geophysik.uni-frankfurt.de; ORCID: 0000-0003-2411-9770Fabian Limberger, Goethe University Frankfurt, Geosciences, Geophysics, Altenhöferallee 1, Frankfurt, Germany, E-mail: f.limberger@geophysik.uni-frankfurt.de; ORCID: 0000-0002-8948-6005Georg Rümpker*, Goethe University Frankfurt, Geosciences, Geophysics, Altenhöferallee 1, Frankfurt, Germany, E-mail: rumpker@geophysik.uni-frankfurt.de; ORCID: 0000-0002-5348-9888* Corresponding author: Abolfazl Komeazi, komeazi@geophysik.uni-frankfurt.de* Corresponding author: Georg Rümpker, rumpker@geophysik.uni-frankfurt.deABSTRACTEarthquake localization using Distributed Acoustic Sensing (DAS) is challenging due to the single-component directional sensitivity of DAS systems. We propose a novel approach for localization that is based on dense DAS recordings and constraints from known structural heterogeneity of the subsurface. Our method employs full-waveform simulations to generate synthetic DAS wavefield images for a range of potential earthquake source locations. A deep convolutional neural network (CNN), based on a U-Net architecture, is trained on these images to map DAS-recorded wavefield patterns to earthquake source coordinates, without the need for identification and picking of P- and S-wave arrivals. We evaluate this wavefield-based localization technique based on a challenging synthetic case study involving DAS recordings in a single vertical borehole only. We consider different velocity models of varying geological complexity. The results show that the CNN effectively learns location-specific wavefield signatures influenced by subsurface heterogeneity. Uncertainties can be reduced significantly by adding recordings from a second borehole. While the results are based on idealized 2D synthetic modeling, the method offers a promising approach for improving microseismic monitoring when detailed information on the heterogeneous velocity structure is available (such as that derived from seismic surveys).PLAIN LANGUAGE SUMMARYLocating earthquakes using Distributed Acoustic Sensing (DAS) fiber optic cables in a single borehole is difficult because DAS primarily measures motion along one direction. Our study explores a new method that utilizes known variations in underground geology. These structures alter how seismic waves travel, creating unique patterns, or wavefields, in the DAS recordings depending on the origin of the earthquake. We use computer simulations based on geological models to generate a catalog of these wavefield patterns from many potential earthquake locations. A machine learning algorithm is then trained to recognize these patterns and map them directly to the source location. We tested this approach using simplified 2-dimensional models and found it can improve earthquake localization accuracy compared to standard techniques. While this approach is still based on simulated data, it shows strong potential for improving earthquake monitoring using a single DAS cable, especially in areas where deploying many sensors is difficult or expensive.