A hydro-climatological approach to predicting regional landslide probability using Landlab

Abstract

We develop a hydroclimatological approach to the modeling of regional shallow landslide initiation that integrates spatial and temporal dimensions of parameter uncertainty to estimate an annual probability of landslide initiation based on Monte Carlo simulations. The physically based model couples the infinite-slope stability model with a steady-state subsurface flow representation and operates in a digital elevation model. Spatially distributed gridded data for soil properties and vegetation classification are used for parameter estimation of probability distributions that characterize model input uncertainty. Hydrologic forcing to the model is through annual maximum daily recharge to subsurface flow obtained from a macroscale hydrologic model. We demonstrate the model in a steep mountainous region in northern Washington, USA, over 2700 km 2 . The influence of soil depth on the probability of landslide initiation is investigated through comparisons among model output produced using three different soil depth scenarios reflecting the uncertainty of soil depth and its potential long-term variability. We found elevation-dependent patterns in probability of landslide initiation that showed the stabilizing effects of forests at low elevations, an increased landslide probability with forest decline at midelevations (1400 to 2400 m), and soil limitation and steep topographic controls at high alpine elevations and in post-glacial landscapes. These dominant controls manifest themselves in a bimodal distribution of spatial annual landslide probability. Model testing with limited observations revealed similarly moderate model confidence for the three hazard maps, suggesting suitable use as relative hazard products. The model is available as a component in Landlab, an open-source, Python-based landscape earth systems modeling environment, and is designed to be easily reproduced utilizing HydroShare cyberinfrastructure. Dunne, 1997a, b) and ecosystem dynamics (Pollock, 1998;May et al., 2009). Landslide hazard maps are a common tool used to characterize the relative potential for landslide occurrence in space, either qualitatively (using susceptibility levels) or quantitatively (using modeled landslide probabilities) (van Westen et al., 2006;Raia et al., 2014).

Cracking “Open” Technology in Ecohydrology

“…Tesfatsion et al (2017) took an open-source modeling tool from the OpenDanubia project and stripped it of a number of elements, including its graphical user interface, in order to expose model functions as an application programming interface (API), and thus make it more generally applicable. Also of interest within this area of research are the effects of hydrologic processes on terrestrial systems, such as shorelines (McCall et al 2014), and their impacts on geomorphological processes, such as landslides (Strauch et al 2017).…”

“…To this end, many researchers have leveraged open datasets, as well as preserved their data for others to use. As an example of the latter point, Strauch et al (2017), in their project on landslide probability, ensured that the data they produced was appropriately documented and shared by first entering it into CUAHSI's HydroShare. They also devoted a portion of the final publication to explaining how and where to find datasets and software used in their work.…”