Final Report Summary - ROSNZ (Landslides, floods and erosion: new insights from event-based field measurements in the Southern Alps of New-Zealand and stochastic 2D numerical modelling ) Erosion of mountain belts is fundamentally episodic, driven by floods incising bedrock and flushing landslide-derived debris. Landslides and floods are characterised by a non-negligible probability of very large extremes which is of particular concern for society. This non-negligible probability also raises the issue of the contribution of extreme landsliding and flooding events to river incision at geological timescales; this has fundamental consequences for the way geomorphologists understand landscape development. Landslides and floods interact along the drainage network to generate a complex spatiotemporal response in which punctuated sediment supply by landslides can have dramatic and prolonged impact on upstream and downstream parts of the river. For instance, even moderate landslides might alter the downstream course of a river to require mitigation measures in order to protect roads or infrastructure from locally increased rates of bank erosion. Large landslides can dam rivers, cause upstream aggradation and catastrophic downstream flooding with substantial aggradation if the dam fails. Landslides can induce aggradation further downstream with an unknown delay, increasing the risk of overbank flooding, bank erosion and the cost of mitigating these effects if infrastructures are at risk. A typical issue is to try to predict the catchment response to a large earthquake triggering extensive landsliding (e.g. following a rupture on the Alpine Fault of the Southern Alps of New-Zealand). No quantitative method is currently able to explore this complex two-dimensional (2D) dynamics at secular to geological timescales. To fill this gap, the ROSNZ project aimed at collecting new data on the impact of landslides on river evolution (outgoing phase, Department of Geological Sciences, Univ. Canterbury, N-Z) and to develop a new numerical model of stochastic mountain landscape evolution incorporating a new bedrock landsliding algorithm (return phase, Geosciences Rennes, CNRS, Fr). A third objective was to further improve the fellow's career by widening its scientific expertise on the area of landslide hazards (outgoing host training), taking active leadership in research management (return host training) and develop the scientific links between the outgoing and return host organisations. Outgoing phase: During 13 months, the fellow documented the impact of landslides and floods on channel morphology using repeat terrestrial Light detection and ranging (LIDAR) surveys of rapidly incising rivers in New Zealand. A major achievement was the acquisition and analysis of a dataset of 7 landslide-affected bedrock reaches of up to 1 km long in two rivers (Rangitikei and Otira Gorge). Surveys before and after a 10-year return flood in the Rangitikei river allowed to measure 3D patterns of bank erosion and up to 2 cm of bedrock incision in a few hours. We also managed to document the impact of a single large storm in the Otira gorge that resulted in more than 2 m of bed aggradation downstream of landslides, and movement of meter-size boulders. These data also highlighted the dominant role of fracture density in controlling local rates of incision. We also quantified the impact of large rockfalls deposits in driving opposite riverbank incision. Our preliminary results suggest that in the Rangitikei river, even modest cliff failure can drive significant erosion of the opposite bank owing to the high bedrock erodibility. Large landslides (e.g. reducing the river width by 50 %) could even produce enough channel deviation, and have a residence time long enough, to generate strath terraces downstream, which are ubiquitous in this area. A major contribution of this phase was the development of two open-source software for the advanced processing of LIDAR datasets: 1. a classification software allowing automatic tagging of points depending on their 3D multi-scale geometry. It showed excellent capability in separating vegetation from ground or cliffs, but also fresh rock from gravels or rockfalls (Brodu and Lague, 2012). 2. a 3D point cloud comparison software dedicated to high-accuracy surface change detection with robust uncertainty estimates. Combined with the high-definition topographic data acquired in the field, we managed to detect down to 5 mm of bedrock incision between surveys (Lague et al., submitted). These new techniques and their application in geomorphology have already had a significant impact on the geosciences community as the fellow was invited to present them at the largest international conference on earth sciences (AGU fall meeting, December 2012, San Francisco, US). Return phase: during 12 months, the fellow used the previously acquired expertise on landslide hazards and field measurements of the impact of floods on river dynamics to explore the longer term consequences of stochastic forcings through numerical modelling. Two aspects were studied: 1. Evidences for the dominant role of stochastic floods and thresholds in nature: the fellow synthesised a large number of published data on mountain river incision and geometry to demonstrate that the most widely used model of long-term river incision (the stream power incision model that neglects stochastic flooding) yield strongly biased predictions. He demonstrated how an explicit description of stochastic flooding resolves an important inconsistency of published data and demonstrated the existence of composite transient dynamics emerging from non-linear effects due to the combination of thresholds and stochastic flooding. These results fundamentally change how geomorphologists should model landscape evolution, and challenges many recent studies trying to invert river profiles to infer tectonic deformation patterns. This work is submitted as a solicited State of Science paper in the European journal Earth Surface Processes and Landforms. The fellow further demonstrated the need to account for the stochastic forcing of landslides. 2. Exploring the combined effect of stochastic landsliding and flooding on landscape dynamics: using the training he received during the outgoing phase of the project, the fellow and scientists in charge developed a new efficient 2D bedrock landsliding algorithm. Its implementation in the landscape evolution model Euro developed at the return host institution is fast enough to simulate the rupture and runout of thousands of landslides and explore the consequences of extreme landsliding events on landscape dynamics (sediment fluxes, drainage network evolution, ...). The landslide model is based on a physical description of landslide rupture propagation, and an explicit description of runout and deposition solved with cellular automata. Despite its simplicity, the model reproduces the main scaling characteristics of observed bedrock landslides in New-Zealand (geometry, frequency-magnitude and runout). Combining stochastic flooding and landsliding in the code ROS (which already incorporated fluvial processes), we ran the first ever fully stochastic 2D landscape evolution model. This represents a major achievement of the project. Simulation results show a large instability of the drainage network and alternating phases of sediment storage and release at secular to millennial timescales. These predictions strongly differ from previous continuous models and highlight how large fluctuations in sedimentary fluxes and river incision rates can naturally occur without change in external forcings (i.e. climate, tectonics). Career development: the fellow's career was boosted through the active collaboration that has developed between the outgoing and return host institution and an active involvement in research management: he took up a position of Associate Editor at the Journal of Geophysical Research, became head of the geomorphology team at the return institution and developed European partnership to submit an ERC-synergy proposal in 2014. His past and current research on the dynamics of mountain rivers has been recognised by the Gordon Warwick Medal he received from the British Society of Geomorphology in June 2012. Perspectives: The ROSNZ research and training project met all his objectives. The outcome consists of scientific achievements but also practical tools (in the form of software) that can be used by geoscientists to monitor and assess geohazards using terrestrial LIDAR data and landscape evolution simulations. Data and models developed in this project will be used by the outgoing and return host institutions that have now teamed up to explore the sedimentary hazards related to large earthquake on the Alpine fault of the Southern Alps (N-Z).
