Space-time landslide susceptibility modelling in Taiwan

Abstract

Portraying spatiotemporal variations in landslide susceptibility patterns is crucial for landslide prevention and management. In this study, we implement a space-time modeling approach to predict the landslide susceptibility on a yearly basis across the main island of Taiwan, from 2004 to 2018. We use a Bayesian version of a binomial generalized additive model, which assumes that landslide occurrences follow a Bernoulli distribution. We generate 46,074 slope units to partition the island of Taiwan and divided the time domain into 14 annual units. The binary landslide label assigned to each slope unit and their temporal replicates come from an available landslide database, that contains an inventory for every year. We only consider new landslides or reactivations of previous mass movements in the yearly inventories. This information and its absence counterpart are regressed against a set of static and dynamic covariates. Our modeling strategy features an initial explanatory model to test the goodness-of-fit and interpret the effect of covariates. Then, five cross-validation (CV) schemes are tested to provide a full spectrum of the predictive capacity of our model. Specifically, we implement a fully randomized 10-fold CV, a spatially constrained CV, two temporal CV (a leave one year out and a sequential temporal aggregation), together with a spatio-temporal CV. We summarize the performance in each of these tests, through their pure numerical expression as well as their residual representation in space and time. Overall, our space-time model produces excellent and interpretable results. We consider this type of dynamic prediction the new direction to take to finally move away from the static view provided by traditional susceptibility models. And, we consider such analyses just a stepping stone for further improvements, the most natural of which would lead to statistical simulations for future scenarios.

A data-driven framework for landslide size space-time modelling

“…We implemented the space-time landslide size modelling in the same study area as our previous study (Fang et al, 2022). The study area is located in the main island of Taiwan (Fig.…”

“…Therefore, to isolate the contribution of new occurrences and/or reactivated failures, we calculated the difference of two subsequent yearly inventories to derive landslide expansion areas for each year under consideration. This preprocessing procedure for landslide maps is the same as our previous study, and further details are provided in Fang et al (2022). As a result, we obtained 14 yearly landslide inventory maps (Fig.…”

“…For this reason, we prepared a set of static and dynamic factors to build our space-time model. Specifically, we derived eight static terrain attributes from SRTM DEM data, five of which have already been employed in Fang et al (2022): slope, plan curvature, profile curvature, northness, and eastness. In the present study we also obtained three relief-related factors (intensity, range, and variance) to represent the gravitational potential energy across the terrain (Stepinski and Jasiewicz, 2011).…”

“…As for the dynamic ones, we opted to include rainfall, normalized difference vegetation index (NDVI), and a yearly function of the timesteps between subsequent inventories. Our previous study showed that maximum daily rainfall is an appropriate dynamic factor in a yearly space-time landslide susceptibility model for Taiwan (Fang et al, 2022). Therefore, we collected the rainfall estimates from 188 meteorological stations and interpolated the yearly daily maximum rainfall via a cokriging routine, which used elevation as a parameter to represent the orographic effect on the precipitation patterns.…”

“…We tested this model with a dataset of the main island of Taiwan for the period from 2004 to 2018, during which tropical cyclones triggered many landslides. Earlier, we have implemented a space-time landslide susceptibility model for the same study area (Fang et al, 2022), which focused on the landslide space-time prediction. This present study aims at estimating probabilistically "how frequent" and "how large" landslides are expected within mapping units.…”