Coupling landslide susceptibility and connectivity models for cost-effective mitigation of landslide-derived sediment delivered to streams

Abstract

<p>We present a morphometric landslide connectivity model using lasso logistic regression that aims to predict the likelihood of sediment delivery to streams following landslide initiation. The model is used to investigate determinants of sediment connectivity for a landslide-triggering storm event in 1977 in the Wairarapa, New Zealand. A range of connectivity scenarios are explored by defining a set of sinks and simulating varying rates of sediment generation during flood events of increasing magnitude. Sediment delivery ratios ranged between 0.21 and 0.29, depending on the definition of sink. When accounting only for direct deposition into streams, the event sediment yield was estimated at 3548 t/km<sup>2</sup>. However, when assuming continued sediment recruitment at reduced rate by overland flow across saturated soils as well as post-event reworking of deposited material, the maximum sediment yield was estimated at 9033 t/km<sup>2</sup>. This estimate of event sediment yield is approximately 2.5 times greater than estimates only considering landslide deposits connecting to streams directly and challenges previous assumptions related to sediment delivery ratios and the importance of shallow landsliding for sediment budgets.</p><p>Furthermore, the influence of coalescing landslide deposits in determining sediment delivery is assessed. An important outcome of the lasso regression is the potential consequence of over-reliance on individual landslide scar size as a predictor of connectivity. Without considering contributions of landslide-derived material from multiple sources to the transport of delivery downslope, the run-out distance is likely to be underestimated for shallow landslides.</p><p>Finally, sediment mobility from shallow landslides is quantified by coupling statistical landslide susceptibility and sediment connectivity models in a modular form.  We present results from scenario modelling for fifty silvopastoral farms and quantify the cost-effectiveness of targeted versus non-targeted approaches to shallow landslide mitigation at the scale of individual trees. The cost-effectiveness of targeted mitigation of landslide-derived sediment is found to be approximately 10-fold greater than a non-targeted approach. An important finding of the landslide susceptibility and connectivity modelling is that, in total, only 6.5% of farmland is the potential source of approximately two-thirds of landslide-derived sediment across the 50 farms. Due to existing vegetation, this area has already been reduced to 4.7%. Further reductions in future sediment delivery can be achieved by increasing slope stability on these highly susceptible and connected slopes through additional biological mitigation.</p>