Investigating and predicting landslides using a rainfall-runoff model in Southern Norway

Landslides are amongst the most destructive natural hazards, causing damage to infrastructures, such as roads, railways and houses, and can, in a worst-case scenario, take lives. By studying the effect and response of rainfall using the temporal and spatial distribution of the storage and discharge, a better understanding of landslide processes and a more detailed prediction can be possible. This study employs a parameter-parsimonious rainfall-runoff model, the Distance Distribution model (DDD), to simulate hydrological conditions for rainfall induced landslide events. The DDD model represents the subsurface in 2D in that it calculates the storage along a hillslope representing the entire catchment in question. Model simulations for 76 debris avalanches and debris flows in Southern Norway have been investigated at catchment scale and at three points along the hillslope. The main objectives were to determine if the model has any capacity to predict hydrological conditions triggering landslides and to investigate how storage-discharge hysteresis is represented in the model and how it can relate to landslide occurrences. Evaluated for the entire catchment, 70 % of the landslide events occurred during completely saturated conditions and more than 90 % of the events are characterized by sharp gradients and/or a prolonged high saturation in the temporal dynamics of saturation. This results suggests that the DDD model has capacity to predict hydrological conditions triggering landslides. The simulation of overland flow proved to be relevant for landslide occurrence found for 87 % of the events. The results for lower, middle and upper point of hillslope show that the storage has a distribution that varies along the hillslope and with time. The 2D representation of the hillslope has the potential to be used in landslide investigation, however, only if the registration of landslide events improves, starting with landslide initiation points. Simulations employing hysteretic curves indicate that the structure of the DDD model allows addressing the non-linear, hysteretic relationship between storage and discharge. Hysteresis are complex processes, however, and there are still many aspects which are not known, suggesting that further exploration of the changes in storage and discharge, dS/dt and dQ/dt, would be useful. In terms of relating the landslide occurrence with hysteresis, no connections were found. A reduced uncertainty related to the timing of the landslide events and the use of input data of hourly resolution may allow for a better correlation between landslides and hysteresis