Identifying subarctic river thermal and mechanical ice break-up using seismic sensing

Sammanfattning: River-ice break-up in high-latitude regions, despite its brevity, is a fundamental process, representing the most dynamic and complex period of fluvial processes. Moreover, ice break-up has significant cascading ecological effects, with a different severity for mechanical vs. thermal break-up, and thus, motivates the importance of monitoring efforts. Classical research methods, such as fieldwork or analysis of photographs and aerial imagery, offer a general perspective on the timing of ice break-up but have safety and logistic issues caused by the dangers of unstable ice cover, the lag times between event occurrence and observation, and the frequent low visibilities. The emerging field of environmental seismology, which studies surface processes through seismic signals, provides an alternative solution to these shortcomings by continuously recording high temporal resolution data. Seismic sensing can potentially record any event within a set distance if the produced signal is powerful enough. Three geophones had monitored the subarctic Sävar River reach for 185 days to test the efficiency of seismic methods to capture ice-cracking events, and based on their characteristics, to identify thermal vs. mechanical ice break-up. With visual and multivariate analysis, seismic methods provided a conservative set of 2 228 events, detected at milliseconds precision, described, and located. Besides, both trigger lag times and principal component analysis depicted correlations between environmental drivers and ice-cracking events. The automatic picker based on duration and trigger thresholds required manual supervision because of the initial numerous false signals that accounted for 96% of total initial events. Ice-cracking signals as short as 0.2s and frequencies of 8-40 Hz with an average power of -117 dB were statistically defined, classified, and described by case events as two types, associated, based on their spectral and temporal patterns, with the two ice break-up modes. With an estimated Rayleigh wave velocity of 680 m/s, all ice-cracking signals' locations were within the instrumented area. Trigger lag times analysis improved detection and showed a strong link between ice-cracking events and drivers of lag times less than three hours, including near-immediate responses (< 2s). With multivariate analysis, the lag times showed a mainly climatic control for thermal melting and a primarily fluvial control in mechanical ice break-up. The combination of statistical and seismic analysis provides, despite the considerable manual screening, a valid and potentially site-transferable method to extract and describe ice-cracking signals and thus identify ice break-up modes in northern rivers.