Refine
Language
- English (22)
Is part of the Bibliography
- yes (22)
Keywords
- Chile (2)
- Germany (2)
- damage (2)
- earthquake (2)
- insurance (2)
- prediction (2)
- preparedness (2)
- probability (2)
- recovery (2)
- transferability (2)
Institute
- Institut für Umweltwissenschaften und Geographie (22) (remove)
Probabilistic models to inform landslide early warning systems often rely on rainfall totals observed during past events with landslides. However, these models are generally developed for broad regions using large catalogs, with dozens, hundreds, or even thousands of landslide occurrences. This study evaluates strategies for training landslide forecasting models with a scanty record of landslide-triggering events, which is a typical limitation in remote, sparsely populated regions. We evaluate 136 statistical models trained on a precipitation dataset with five landslide-triggering precipitation events recorded near Sitka, Alaska, USA, as well as 6000 d of non-triggering rainfall (2002–2020). We also conduct extensive statistical evaluation for three primary purposes: (1) to select the best-fitting models, (2) to evaluate performance of the preferred models, and (3) to select and evaluate warning thresholds. We use Akaike, Bayesian, and leave-one-out information criteria to compare the 136 models, which are trained on different cumulative precipitation variables at time intervals ranging from 1 h to 2 weeks, using both frequentist and Bayesian methods to estimate the daily probability and intensity of potential landslide occurrence (logistic regression and Poisson regression). We evaluate the best-fit models using leave-one-out validation as well as by testing a subset of the data. Despite this sparse landslide inventory, we find that probabilistic models can effectively distinguish days with landslides from days without slide activity. Our statistical analyses show that 3 h precipitation totals are the best predictor of elevated landslide hazard, and adding antecedent precipitation (days to weeks) did not improve model performance. This relatively short timescale of precipitation combined with the limited role of antecedent conditions likely reflects the rapid draining of porous colluvial soils on the very steep hillslopes around Sitka. Although frequentist and Bayesian inferences produce similar estimates of landslide hazard, they do have different implications for use and interpretation: frequentist models are familiar and easy to implement, but Bayesian models capture the rare-events problem more explicitly and allow for better understanding of parameter uncertainty given the available data. We use the resulting estimates of daily landslide probability to establish two decision boundaries that define three levels of warning. With these decision boundaries, the frequentist logistic regression model incorporates National Weather Service quantitative precipitation forecasts into a real-time landslide early warning “dashboard” system (https://sitkalandslide.org/, last access: 9 October 2023). This dashboard provides accessible and data-driven situational awareness for community members and emergency managers.
Bayesian geomorphology
(2020)
The rapidly growing amount and diversity of data are confronting us more than ever with the need to make informed predictions under uncertainty. The adverse impacts of climate change and natural hazards also motivate our search for reliable predictions. The range of statistical techniques that geomorphologists use to tackle this challenge has been growing, but rarely involves Bayesian methods. Instead, many geomorphic models rely on estimated averages that largely miss out on the variability of form and process. Yet seemingly fixed estimates of channel heads, sediment rating curves or glacier equilibrium lines, for example, are all prone to uncertainties. Neighbouring scientific disciplines such as physics, hydrology or ecology have readily embraced Bayesian methods to fully capture and better explain such uncertainties, as the necessary computational tools have advanced greatly. The aim of this article is to introduce the Bayesian toolkit to scientists concerned with Earth surface processes and landforms, and to show how geomorphic models might benefit from probabilistic concepts. I briefly review the use of Bayesian reasoning in geomorphology, and outline the corresponding variants of regression and classification in several worked examples.
Dense tree stands and high wind speeds characterize the temperate rainforests of southern Chilean Patagonia, where landslides frequently strip hillslopes of soils, rock, and biomass. Assuming that wind loads on trees promote slope instability, we explore the role of forest cover and wind speed in predicting landslides with a hierarchical Bayesian logistic regression. We find that higher crown openness and wind speeds credibly predict higher probabilities of detecting landslides regardless of topographic location, though much better in low-order channels and on midslope locations than on open slopes. Wind speed has less predictive power in areas that were impacted by tephra fall from recent volcanic eruptions, while the influence of forest cover in terms of crown openness remains. <br /> Plain Language Summary Chilean Patagonia hosts some of Earth's largest swaths of temperate rainforests, where frequent landslides erode soil, rock, and vegetation. We explore the role of forest cover and wind disturbances in promoting such landslides with a model that predicts from crown openness and wind speed the probability of detecting landslide terrain. We find that both forest cover and wind speed play important, yet previously underappreciated, roles in this context, especially when grouped by landform types and previous volcanic disturbance, which may override the comparable modest control of wind on landslides. Our study is the first of its kind in one of the windiest spots on Earth and encourages a more discerning approach to landslide prediction.
Seasonal variations in landslide activity remain understudied compared to recent advances in landslide early warning at hourly to daily timescales.
Here, we learn the seasonal pattern of monthly landslide activity in the Pacific Northwest from five heterogeneous landslide inventories with differing spatial and temporal coverage and reporting protocols combined in a Bayesian multi-level model.
We find that landslide activity is distinctly seasonal, with credible increases in landslide intensity, inter-annual variability, and probability marking the onset of the landslide season in November.
Peaks in landslide probability in January and intensity in February lag the annual peak in mean monthly precipitation and landslide activity is more variable in winter than in summer, when landslides are rare.
For a given monthly rainfall, landslide intensity at the season peak in February is up to 10 times higher than at the onset in November, underlining the importance of antecedent seasonal hillslope conditions.
Large earthquakes initiate chains of surface processes that last much longer than the brief moments of strong shaking. Most moderate‐ and large‐magnitude earthquakes trigger landslides, ranging from small failures in the soil cover to massive, devastating rock avalanches. Some landslides dam rivers and impound lakes, which can collapse days to centuries later, and flood mountain valleys for hundreds of kilometers downstream. Landslide deposits on slopes can remobilize during heavy rainfall and evolve into debris flows. Cracks and fractures can form and widen on mountain crests and flanks, promoting increased frequency of landslides that lasts for decades. More gradual impacts involve the flushing of excess debris downstream by rivers, which can generate bank erosion and floodplain accretion as well as channel avulsions that affect flooding frequency, settlements, ecosystems, and infrastructure. Ultimately, earthquake sequences and their geomorphic consequences alter mountain landscapes over both human and geologic time scales. Two recent events have attracted intense research into earthquake‐induced landslides and their consequences: the magnitude M 7.6 Chi‐Chi, Taiwan earthquake of 1999, and the M 7.9 Wenchuan, China earthquake of 2008. Using data and insights from these and several other earthquakes, we analyze how such events initiate processes that change mountain landscapes, highlight research gaps, and suggest pathways toward a more complete understanding of the seismic effects on the Earth's surface.
Models for the predictions of monetary losses from floods mainly blend data deemed to represent a single flood type and region. Moreover, these approaches largely ignore indicators of preparedness and how predictors may vary between regions and events, challenging the transferability of flood loss models. We use a flood loss database of 1812 German flood-affected households to explore how Bayesian multilevel models can estimate normalised flood damage stratified by event, region, or flood process type. Multilevel models acknowledge natural groups in the data and allow each group to learn from others. We obtain posterior estimates that differ between flood types, with credibly varying influences of water depth, contamination, duration, implementation of property-level precautionary measures, insurance, and previous flood experience; these influences overlap across most events or regions, however. We infer that the underlying damaging processes of distinct flood types deserve further attention. Each reported flood loss and affected region involved mixed flood types, likely explaining the uncertainty in the coefficients. Our results emphasise the need to consider flood types as an important step towards applying flood loss models elsewhere. We argue that failing to do so may unduly generalise the model and systematically bias loss estimations from empirical data.
The propagation of a seismic rupture on a fault introduces spatial variations in the seismic wave field surrounding the fault. This directivity effect results in larger shaking amplitudes in the rupture propagation direction. Its seismic radiation pattern also causes amplitude variations between the strike-normal and strike-parallel components of horizontal ground motion. We investigated the landslide response to these effects during the 2016 Kumamoto earthquake (M-w 7.1) in central Kyushu (Japan). Although the distribution of some 1500 earthquake-triggered landslides as a function of rupture distance is consistent with the observed Arias intensity, the landslides were more concentrated to the northeast of the southwest-northeast striking rupture. We examined several landslide susceptibility factors: hillslope inclination, the median amplification factor (MAF) of ground shaking, lithology, land cover, and topographic wetness. None of these factors sufficiently explains the landslide distribution or orientation (aspect), although the landslide head scarps have an elevated hillslope inclination and MAF. We propose a new physics-based ground-motion model (GMM) that accounts for the seismic rupture effects, and we demonstrate that the low-frequency seismic radiation pattern is consistent with the overall landslide distribution. Its spatial pattern is influenced by the rupture directivity effect, whereas landslide aspect is influenced by amplitude variations between the fault-normal and fault-parallel motion at frequencies < 2 Hz. This azimuth dependence implies that comparable landslide concentrations can occur at different distances from the rupture. This quantitative link between the prevalent landslide aspect and the low-frequency seismic radiation pattern can improve coseismic landslide hazard assessment.
Rock glaciers are permafrost or glacial landforms of debris and ice that deform under the influence of gravity. Recent estimates hold that, in the semiarid Chilean Andes for example, active rock glaciers store more water than glaciers. However, little is known about how many rock glaciers might decay because of global warming and how much this decay might contribute to water and sediment release. We investigated an inventory of >6500 rock glaciers in the Argentinian Andes, spanning the climatic gradient from the Desert Andes to cold-temperate Tierra del Fuego. We used active rock glaciers as a diagnostic of permafrost, assuming that the toes mark the 0 degrees C isotherm in climate scenarios for the twenty-first century and their impact on freezing conditions near the rock glacier toes. We find that, under future worst case warming, up to 95% of rock glaciers in the southern Desert Andes and in the Central Andes will rest in areas above 0 degrees C and that this freezing level might move up more than twice as much (similar to 500 m) as during the entire Holocene (similar to 200 m). Many active rock glaciers are already well below the current freezing level and exemplify how local controls may confound regional prognoses. A Bayesian Multifactor Analysis of Variance further shows that only in the Central Andes are the toes of active rock glaciers credibly higher than those of inactive ones. Elsewhere in the Andes, active and inactive rock glaciers occupy indistinguishable elevation bands, regardless of aspect, the formation mechanism, or shape of rock glaciers. The state of rock glacier activity predicts differences in elevations of toes to 140 m at best so that regional inference of the distribution of discontinuous permafrost from rock-glacier toes cannot be more accurate than this in the Argentinian Andes. We conclude that the Central Andes-where rock glaciers are largest, cover the most area, and have a greater density than glaciers-is likely to experience the most widespread disturbance to the thermal regime of the twenty-first century. (C) 2018 Elsevier B.V. All rights reserved.
Pokhara (ca. 850 m a.s.l.), Nepal's second-largest city, lies at the foot of the Higher Himalayas and has more than tripled its population in the past 3 decades. Construction materials are in high demand in rapidly expanding built-up areas, and several informal settlements cater to unregulated sand and gravel mining in the Pokhara Valley's main river, the Seti Khola. This river is fed by the Sabche glacier below Annapurna III (7555 m a.s.l.), some 35 km upstream of the city, and traverses one of the steepest topographic gradients in the Himalayas. In May 2012 a sudden flood caused >70 fatalities and intense damage along this river and rekindled concerns about flood risk management. We estimate the flow dynamics and inundation depths of flood scenarios using the hydrodynamic model HEC-RAS (Hydrologic Engineering Center’s River Analysis System). We simulate the potential impacts of peak discharges from 1000 to 10 000 m3 s−1 on land cover based on high-resolution Maxar satellite imagery and OpenStreetMap data (buildings and road network). We also trace the dynamics of two informal settlements near Kaseri and Yamdi with high potential flood impact from RapidEye, PlanetScope, and Google Earth imagery of the past 2 decades. Our hydrodynamic simulations highlight several sites of potential hydraulic ponding that would largely affect these informal settlements and sites of sand and gravel mining. These built-up areas grew between 3- and 20-fold, thus likely raising local flood exposure well beyond changes in flood hazard. Besides these drastic local changes, about 1 % of Pokhara's built-up urban area and essential rural road network is in the highest-hazard zones highlighted by our flood simulations. Our results stress the need to adapt early-warning strategies for locally differing hydrological and geomorphic conditions in this rapidly growing urban watershed.