We detected 12 landquakes in Taiwan with collapse areas between 0.27 and 2.48 km² associated with rock collapse, rockslide and debris and/or rock avalanche during the passage of Typhoon Morakot in 2009. These events were recorded by seismic stations of the Broadband Array in Taiwan for Seismology. Their locations were determined by a cross-correlation technique that maximizes the coherency of horizontal envelope function among seismic stations with a mean location error of 1.92 km. We applied time-frequency analysis to estimate the bandwidth of seismic energy generated by the landquakes. The predominant frequency ranges from 0.5 to 5.0 Hz, with higher-frequency signals likely caused by block impact. We extracted signal duration (S_D), peak ground velocity, rise time (T_R), area of velocity envelope function (A_E) from the closest station and estimated the collapse area (A_C) and run-out distance (D_R) by mapping satellite images. Based on aforementioned seismic and geometrical parameters, we defined the potential of initial impact (P_I), the frequency of rock impact signal (f_I) and the mean quasi-front velocity (V_f) as indicators of landquake types, especially for events with dam formation (dam-formation-type events). We also derived an empirical linear relationship between the envelope area (A_E) and collapse area (A_C) with a high correlation coefficient of 0.83. Our automatic approach is very effective for rapid determination of landquake centroid location and collapse area, and for identifying dam-formation event using records from existing real-time broad-band seismic networks, thus providing an important alternative for landquake hazard mitigation.

Hazards from gravity-driven instabilities on hillslope (termed ‘landquake’ in this study) are an important problem facing us today. Rapid detection of landquake events is crucial for hazard mitigation and emergency response. Based on the real-time broadband data in Taiwan, we have developed a near real-time landquake monitoring system, which is a fully automatic process based on waveform inversion that yields source information (e.g., location and mechanism) and identifies the landquake source by examining waveform fitness for different types of source mechanisms. This system has been successfully tested offline using seismic records during the passage of the 2009 Typhoon Morakot in Taiwan and has been in online operation during the typhoon season in 2015. In practice, certain levels of station coverage (station gap < 180°), signal-to-noise ratio (SNR ≥ 5.0), and a threshold of event size (volume >10⁶ m³ and area > 0.20 km²) are required to ensure good performance (fitness > 0.6 for successful source identification) of the system, which can be readily implemented in other places in the world with real-time seismic networks and high landquake activities.

Landslide hazard motivates the need for a deeper understanding of the events that occur before, during, and after catastrophic slope failures. Due to the destructive nature of such events, in situ observation is often difficult or impossible. Here, we use data from a network of 58 seismic stations to characterise a large landslide at the Askja caldera, Iceland, on 21 July 2014. High data quality and extensive network coverage allow us to analyse both long- and short-period signals associated with the landslide, and thereby obtain information about its triggering, initiation, timing, and propagation. At long periods, a landslide force history inversion shows that the Askja landslide was a single, large event starting at the SE corner of the caldera lake at 23:24:05 UTC and propagating to the NW in the following 2 min. The bulk sliding mass was 7–16 × 10¹⁰ kg, equivalent to a collapsed volume of 35–80 × 10⁶ m³. The sliding mass was displaced downslope by 1260 ± 250 m. At short periods, a seismic tremor was observed for 30 min before the landslide. The tremor is approximately harmonic with a fundamental frequency of 2.3 Hz and shows time-dependent changes of its frequency content. We attribute the seismic tremor to stick-slip motion along the landslide failure plane. Accelerating motion leading up to the catastrophic slope failure culminated in an aseismic quiescent period for 2 min before the landslide. We propose that precursory seismic signals may be useful in landslide early-warning systems. The 8 h after the main landslide failure are characterised by smaller slope failures originating from the destabilised caldera wall decaying in frequency and magnitude. We introduce the term afterslides for this subsequent, declining slope activity after a large landslide.

Flooding resulting from the bursting of dams formed by landquake events such as rock avalanches, landslides and debris flows can lead to serious bank erosion and inundation of populated areas near rivers. Seismic waves can be generated by landquake events which can be described as time-dependent forces (unloading/reloading cycles) acting on the Earth. In this study, we conduct inversions of long-period (LP, period ≥20 s) waveforms for the landquake force histories (LFHs) of ten events, which provide quantitative characterization of the initiation, propagation and termination stages of the slope failures. When the results obtained from LP waveforms are analyzed together with high-frequency (HF, 1–3 Hz) seismic signals, we find a relatively strong late-arriving seismic phase (dubbed Dam-forming phase or D-phase) recorded clearly in the HF waveforms at the closest stations, which potentially marks the time when the collapsed masses sliding into river and perhaps even impacting the topographic barrier on the opposite bank. Consequently, our approach to analyzing the LP and HF waveforms developed in this study has a high potential for identifying five dam-forming landquake events (DFLEs) in near real-time using broadband seismic records, which can provide timely warnings of the impending floods to downstream residents.

Continuous seismic records near river channels can be used to quantify the energy induced by river sediment transport. During the 2011 typhoon season, we deployed a seismic array along the Chishan River in the mountain area of southern Taiwan, where there is strong variability in water discharge and high sedimentation rates. We observe hysteresis in the high-frequency (5–15 Hz) seismic noise level relative to the associated hydrological parameters. In addition, our seismic noise analysis reveals an asymmetry and a high coherence in noise cross-correlation functions for several station pairs during the typhoon passage, which corresponds to sediment particles and turbulent flows impacting along the riverbed where the river bends sharply. Based on spectral characteristics of the seismic records, we also detected 20 landslide/debris flow events, which we use to estimate the sediment supply. Comparison of sediment flux between seismologically determined bedload and derived suspended load indicates temporal changes in the sediment flux ratio, which imply a complex transition process from the bedload regime to the suspension regime between typhoon passage and off-typhoon periods. Our study demonstrates the possibility of seismologically monitoring river bedload transport, thus providing valuable additional information for studying fluvial bedrock erosion and mountain landscape evolution.

One purpose of landslide research is to establish early warning thresholds for rainfall-induced landslides. Insufficient observations of past events have inhibited the analysis of critical rainfall conditions triggering landslides. This difficulty may be resolved by extracting the timing of landslide occurrences through analysis of seismic signals. In this study, seismic records of the Broadband Array in Taiwan for Seismology were examined to identify ground motion triggered by large landslides that occurred in the years 2005 to 2014. A total of 62 landslide-induced seismic signals were identified. The seismic signals were analyzed to determine the timing of landslide occurrences, and the rainfall conditions at those times – including rainfall intensity (I), duration (D), and effective rainfall (Rt) – were assessed. Three common rainfall threshold models (I–D, I–Rt, and Rt–D) were compared, and the crucial factors of a forecast warning model were found to be duration and effective rainfall. In addition, rainfall information related to the 62 landslides was analyzed to establish a critical height of water model, (I − 1.5) ⋅ D = 430.2. The critical height of water model was applied to data from Typhoon Soudelor of 2015, and the model issued a large landslide warning for southern Taiwan.