Accounting for covariate distributions in slope-unit-based landslide susceptibility models. A case study in the alpine environment

Amato, Gabriele; Eisank, Clemens; Castro‐Camilo, Daniela; Lombardo, Luigi

doi:10.1016/j.enggeo.2019.105237

Cited by 71 publications

(47 citation statements)

References 85 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The most common choice corresponds to a regular squared lattice or grid (e.g., Cama et al 2016;Hussin et al 2016;Reichenbach et al 2018). However, this mapping unit is sensible to mapping errors (Steger et al 2016) and the assignment of the instability status is inherently uncertain as it is often subjectively chosen between the centroid of the landslide body (e.g., Hussin et al 2016;Castro Camilo et al 2017) or the highest point along the landslide polygon (e.g., Amato et al 2019;Lombardo et al 2014).…”

Section: Mapping Unitmentioning

confidence: 99%

“…In fact, at the SU level or catchment or any large mapping unit, one needs to approximate the distribution of properties that vary at small spatial scales (e.g., Dreyfus et al 2013). For a continuous factor such as Slope or any other terrain derivative, this comes relatively easy by computing some descriptive statistics such as the mean and standard deviation (same as we did here) or a much finer description into quantiles (e.g., Amato et al 2019). However, for a geological map or a bedding measurement, representing these two properties at the SU level is more complex.…”

Section: Reference Model Interpretationmentioning

confidence: 99%

See 1 more Smart Citation

From scenario-based seismic hazard to scenario-based landslide hazard: rewinding to the past via statistical simulations

Luo

Lombardo

Westen

et al. 2021

Stoch Environ Res Risk Assess

Self Cite

View full text Add to dashboard Cite

The vast majority of statistically-based landslide susceptibility studies assumes the slope instability process to be time-invariant under the definition that “the past and present are keys to the future”. This assumption may generally be valid. However, the trigger, be it a rainfall or an earthquake event, clearly varies over time. And yet, the temporal component of the trigger is rarely included in landslide susceptibility studies and only confined to hazard assessment. In this work, we investigate a population of landslides triggered in response to the 2017 Jiuzhaigou earthquake ($$M_w = 6.5$$ M w = 6.5 ) including the associated ground motion in the analyses, these being carried out at the Slope Unit (SU) level. We do this by implementing a Bayesian version of a Generalized Additive Model and assuming that the slope instability across the SUs in the study area behaves according to a Bernoulli probability distribution. This procedure would generally produce a susceptibility map reflecting the spatial pattern of the specific trigger and therefore of limited use for land use planning. However, we implement this first analytical step to reliably estimate the ground motion effect, and its distribution, on unstable SUs. We then assume the effect of the ground motion to be time-invariant, enabling statistical simulations for any ground motion scenario that occurred in the area from 1933 to 2017. As a result, we obtain the full spectrum of potential coseismic susceptibility patterns over the last century and compress this information into a hazard model/map representative of all the possible ground motion patterns since 1933. This backward statistical simulations can also be further exploited in the opposite direction where, by accounting for scenario-based ground motion, one can also use it in a forward direction to estimate future unstable slopes.

show abstract

Section: Mapping Unitmentioning

confidence: 99%

Section: Reference Model Interpretationmentioning

confidence: 99%

From scenario-based seismic hazard to scenario-based landslide hazard: rewinding to the past via statistical simulations

Luo

Lombardo

Westen

et al. 2021

Stoch Environ Res Risk Assess

Self Cite

View full text Add to dashboard Cite

show abstract

“…Traditional heuristic, general statistical, and machine learning methods used for LSM tasks are mostly pixel-based without considering surrounding pixels. A few other methods considering slopes as a single unit instead of a single pixel usually lead to a mapping result less smooth than that by pixel-based methods [76][77][78] because it is not always possible to divide different slope units. CNN models make it possible to mine the spatial structure information from surrounding pixels by use of convolution or pooling.…”

Section: Pixel Itself or Surrounding Pixels For Lsm Taskmentioning

confidence: 99%

Mapping Post-Earthquake Landslide Susceptibility: A U-Net Like Approach

et al. 2020

View full text Add to dashboard Cite

A serious earthquake could trigger thousands of landslides and produce some slopes more sensitive to slide in future. Landslides could threaten human’s lives and properties, and thus mapping the post-earthquake landslide susceptibility is very valuable for a rapid response to landslide disasters in terms of relief resource allocation and posterior earthquake reconstruction. Previous researchers have proposed many methods to map landslide susceptibility but seldom considered the spatial structure information of the factors that influence a slide. In this study, we first developed a U-net like model suitable for mapping post-earthquake landslide susceptibility. The post-earthquake high spatial airborne images were used for producing a landslide inventory. Pre-earthquake Landsat TM (Thematic Mapper) images and the influencing factors such as digital elevation model (DEM), slope, aspect, multi-scale topographic position index (mTPI), lithology, fault, road network, streams network, and macroseismic intensity (MI) were prepared as the input layers of the model. Application of the model to the heavy-hit area of the destructive 2008 Wenchuan earthquake resulted in a high validation accuracy (precision 0.77, recall 0.90, F1 score 0.83, and AUC 0.90). The performance of this U-net like model was also compared with those of traditional logistic regression (LR) and support vector machine (SVM) models on both the model area and independent testing area with the former being stronger than the two traditional models. The U-net like model introduced in this paper provides us the inspiration that balancing the environmental influence of a pixel itself and its surrounding pixels to perform a better landslide susceptibility mapping (LSM) task is useful and feasible when using remote sensing and GIS technology.

show abstract

“…Slope steepness (Slope) (Zevenbergen and Thorne, 1987) 8. Slope Unit area (SU area) (e.g., Amato et al, 2019) 9. Topographic Position Index (TPI) (Guisan et al, 1999) 10.…”

Section: Covariatesmentioning

confidence: 99%

Advanced non-destructive techniques for the diagnosis of historic buildings: The Loka-Hteik-Pan temple in Bagan

Costamagna

Quintero

Bianchini

et al. 2020

Journal of Cultural Heritage

View full text Add to dashboard Cite

Cultural heritage is the foundation upon which global and historical values are based on. It connects us to the legacy left by our ancestors and identifies who we are as part of the modern society. Globally and specifically in the northeastern Romania, the landscape where cultural heritage sites were built on is constantly evolving due to mass wasting processes. Among these processes, landslide and gullies can disrupt the gravitational equilibrium directly or around these sites, threatening their very existence and our capacity to pass them on to future generations. Because landsliding and gullying are stochastic processes, the use of spatial statistics has often been employed to map locations at risk. In this work, we make use of advanced spatial Bayesian statistics to model landslide and gully erosion susceptibilities, separately. And, we ultimately combine these two outputs into a unified multi-hazard susceptibility model which we cross with the known cultural heritage sites in a study area close to the city of Iaşi, in Romania. Specifically, we implement a Bayesian version of a Generalized Additive Model (GAM) which assumes that the two separate landslide and gully presence/absence distributions to behave according to a Bernoulli probability distribution. Contrary to common practices in the literature, the two susceptibility models both feature fixed and random effects, including covariates acting at a latent level. We do this to also capture the unexplained but spatially coherent distribution of properties not directly included in the model. As for the properties directly expressed as covariates, our GAM features terrain attributes obtained from a LIDAR survey, in addition to land use and soil layers.The two single models outstandingly perform (AUC > 0.9) both during the calibration and validation phases. This modeling procedure ensures that the probability of occurrence of the two mass wasting processes under consideration is well estimated and therefore can be used to reliably plan conservation practices for local cultural heritage sites deemed at risk.

show abstract

Accounting for covariate distributions in slope-unit-based landslide susceptibility models. A case study in the alpine environment

Cited by 71 publications

References 85 publications

From scenario-based seismic hazard to scenario-based landslide hazard: rewinding to the past via statistical simulations

From scenario-based seismic hazard to scenario-based landslide hazard: rewinding to the past via statistical simulations

Mapping Post-Earthquake Landslide Susceptibility: A U-Net Like Approach

Advanced non-destructive techniques for the diagnosis of historic buildings: The Loka-Hteik-Pan temple in Bagan

Contact Info

Product

Resources

About