Tobias Schöffl

Debris flows are frequent natural hazards whose destructiveness is controlled by the dynamics of their flow fronts, surges behind the front and large boulders. Understanding the mechanisms underlying the spatiotemporal variations in flow depth and velocity is limited by a lack of catchment‐scale measurements. In this study, we present and analyze flow‐depth and velocity measurements from a new monitoring setup which consists of high‐frequency 3D LiDAR scanners installed at three different locations along the active debris‐flow fan of the Illgraben. For the event analyzed herein, we observe that (a) the LiDAR‐based velocities are in excellent agreement with measurements from a Pulse‐Doppler (PD) radar and with manually tracked feature velocities; (b) the flow front decelerates as it travels along the fan and a watery pre‐surge develops, likely due to a combination of segregation and vertical shear, which transport woody debris and small boulders to the front, as well as a horizontal velocity profile, required for transportation of large boulders through a mechanism we term “centerline advection”; (c) roll waves begin to develop on the lower part of the fan by coalescence of free surface instabilities and they exceed the front velocity by up to 2 to 3; (d) surges later in the event do not show an accumulation of boulders at the crest, but can accelerate individual boulders to velocities up to 2 the front. This study demonstrates that radar and LiDAR‐based measurements improve our understanding of debris‐flow processes, including front propagation and development of waves.

Debris flows are fast-moving, destructive mass transport processes that frequently occur in mountainous regions, posing severe threats to infrastructure and communities. Despite extensive research on debris flows, high-resolution velocity and flow depth data from full-scale natural events to test and parameterize empirical equations remain scarce. This study utilizes pulse-Doppler (PD) radar to continuously track debris-flow velocities at Illgraben, Switzerland, during the 2022 season. We analysed three debris flows and one debris flood, initially assessing four empirical mean velocity estimation equations: Newtonian Laminar Flow, Dilatant Grain Shearing, Manning-Strickler, and Chézy.Flow resistance coefficients were back-calculated for each equation to evaluate their applicability and define their plausible value ranges. Based on these findings, we identified optimal Manning-Strickler (n = 0.16; n = 0.09) and Dilatant Grain Shearing (ξ = 25.5; ξ = 51.2) coefficients for the three debris flows and the debris flood, respectively, to estimate discharge and volume, highlighting both the strengths and limitations of these approaches. However, substantial variability in M-S and DGS coefficients—both within individual events and across different flows—challenges the conventional assumption of constant friction coefficients in debris-flow modeling.Additionally, analysis of the relationship between flow height and velocity revealed a progressive decrease in yield stress across successive surges in two events, indicating a transition toward more fluidized flow behaviour. These findings contribute critical data for refining debris-flow models and improving predictive capabilities.

The Grafenhöfe site in Innervillgraten, Austria, has been affected by ongoing rockfall activity following the destabilization of a steep rock slope due to storm-induced deforestation (storm Vaia in 2018) and subsequent mass movements. The slope, composed primarily of marble and mica schist, experienced initial failures in September 2023, leading to repeated rockfall events and the partial infill of the Grafenbach torrent. In January 2024, further destabilization resulted in large-scale rock detachments, prompting immediate safety measures, including the evacuation of a farmstead.To enhance safety during the construction of protective dams, the Austrian Torrent and Avalanche Control (WLV) implemented a pulse-Doppler Radar system (IBTP Koschuch) for real-time rockfall detection. The system provided continuous high-resolution monitoring, triggering alarms within two seconds upon detecting rockfall movement. This allowed for rapid response and significantly reduced exposure of construction workers to hazards on-site. For monitoring the deformation of the detached rock mass, an automatic continuous terrestrial survey was installed on the opposite slope. This provided insight into the development of the slope's deformation, allowing for the avoidance of construction during a potential large-scale failure of the rock mass.The radar system, which was specifically developed for alpine settings, detects Doppler spectra, ensuring reliable detection independent of weather and light conditions. Integrated with an automated alerting network, it facilitated direct communication with construction teams and authorities, enabling proactive safety management. Beyond immediate hazard mitigation, the radar data provided a valuable basis for refining WLV's safety strategy, as evidenced by the correlation of rockfall detections with rainfall data, revealing a strong dependency: all detected rockfalls coincided with precipitation events, while rainfall did not always led to rockfall. This enabled an optimised risk management approach, where construction activities were suspended during rainfall, effectively minimising exposure to potential rockfall events.This study underscores the effectiveness of real-time monitoring for adaptive hazard mitigation during high-risk construction projects. Future work will focus on refining detection algorithms and integrating AI-based predictive models to enhance early warning capabilities for rockfall hazards.

Intelligent identification of debris flow using Doppler-based radar remains unexplored. In this study, we have implemented 12 deep learning models and a voting strategy to develop classification models for identifying the debris flow, using 24,000 samples across eight categories of targets obtained from field experiments. Every deep learning model showed significant proficiency in classification, achieving the highest accuracy of 95.46% for the multi-object classification. Among the individual models, the vgg16 model excelled in debris flow identification, exhibiting a high precision and a low false alarm rate. Furthermore, the majority-based voting strategy could improve the reliability of individual models for the classification task.

Debris flows represent a severe hazard in Alpine regions. The initiation of debris flows is connected to several hydro-meteorological and geomorphological factors. For operational early warning and assessment of climate change impact, knowledge about critical rainfall conditions is needed. For several regions of the European Alps intensity-duration (I-D) thresholds for debris flows have been derived in recent years. In this study we provide triggering rainfall conditions of high temporal and spatial resolution for > 3700 documented torrent processes including debris flows that occurred in Austria between 2003 and 2022. Additionally, we estimate the change in their probability of occurrence in a future climate, based on an ensemble of bias corrected and localized EURO-CORDEX simulations. We find slightly steeper I-D curves for debris flows than for torrent floods and no clear trend indicating substantial influence of antecedent rainfall on the triggering rainfall. For all process types, it is shown that both the probability of occurrence and the areas affected by triggering precipitation events increase substantially in the future, with clear dependences on the emission scenarios (RCPs). The results of this study provide a basis for improved event forecasting in a changing climate.

Microwave radar, utilizing the differences in Doppler frequencies from moving target echoes, offers remote sensing capabilities and continuous all‐weather monitoring for geological disasters. However, intelligent identification of debris flow signals using such radar remains unexplored. Therefore, we implemented 12 deep learning models coupled with a voting strategy to develop classification models for identifying the debris flow, using 24,000 samples across eight categories of targets obtained from field experiments. Each model demonstrated significant proficiency in classification, achieving a remarkable highest accuracy of 95.46% for the multi‐object classification. Among the individual models, the vgg16 model with a simple and deep architecture excelled in debris flow identification, exhibiting a high precision and a low false alarm rate. The voting strategy further improved the reliability of individual deep learning model. We propose that employing radar‐based deep learning techniques combined with extensive field data represents a crucial advancement in the monitoring and early warning of debris flow.

As a consequence of their natural occurrence and the frequent formation of multiple surges with high sediment loads, debris flows are considered one of the most hazardous gravity‐driven mass movements in mountain regions. Field measurements of surge dynamics are an essential link in the chain of understanding fundamental process dynamics and engineering protection against debris flows. However, continuous information on the velocities of multiple consecutive surges within a single debris‐flow event with high temporal resolution is rare. In this study, we present a new pulse‐Doppler radar (PD radar) for high‐resolution real‐time debris‐flow monitoring. We analyze PD radar data sets over a torrent length of 250 m for two debris flows that occurred at the Gadria creek (IT), on 26 July 2019 and 10 August 2020. The radar data were validated with independently derived data from particle image velocimetry and manually tracked velocities. We observe that between surges, the flow frequently comes to a complete halt and is re‐mobilized by subsequent surges, resembling erosion‐deposition waves in granular flows. In addition, our data confirm that surges can superimpose and merge. We anticipate that the outcomes of this work serve as a blueprint for future high‐resolution observations of debris‐flow surge dynamics with PD radar and that our findings provide new insights into the physical principles of natural debris flows.

As a consequence of their spontaneous occurrence, and frequent formation of multiple surges with high sediment loads, debris flows are considered one of the most hazardous gravity-driven mass movements in montane regions. Field measurements of surface velocities are an essential link in the chain of understanding fundamental process dynamics and applied protection against debris flows. In order to measure the velocities of multiple consecutive surges within a single debris-flow event, a PD radar (pulse-Doppler high-frequency radar) sensor for high-resolution real-time debris-flow monitoring has been developed. In this contribution we present PD radar measurements of two debris flows, that occurred at the Gadria creek in Italy on July 26, 2019, and August 10, 2020, over a torrent length of 250 meters. We record over 55 high-amplitude surges that overlap and superimpose at the front of the debris flow, but also subsequently throughout the debris-flow body. Our results demonstrate applicability of a PD radar for debris-flow monitoring and serve as a data source for modelling surge dynamics in debris flows.