Rainfall-triggered Landslides Research Articles

A study of rainfall thresholds for landslides in Badung Regency using satellite-derived rainfall grid datasets

Integrating field rainfall data with satellite data improves data accuracy and overcomes rainfall data limitations for rain thresholds. Integration can involve field rainfall data, satellite rainfall data, or a different satellite dataset. Merging these rainfall data sources provides more spatial coverage of satellite data. To determine how well rainfall thresholds predict rainfall-triggered landslides, the threshold model must be validated. This study will evaluate satellite rainfall data before and after integration in developing a rainfall threshold model for landslide prediction in Badung Regency. To do so, the study used a cumulative rainfall threshold over 3, 7, 15, and 30 days and two rainfall satellite products (integrated merged multi-satellite retrievals (IMERG) and precipitation estimation from remotely sensed information using artificial neural networks (PERSIANN)). Median, first, and third quartiles were used to set thresholds. The area under the curve (AUC) was calculated to validate rainfall threshold outcomes using receiver operating characteristic (ROC) curves. Analysis showed that integrating satellite rainfall data into the rainfall threshold model for landslide prediction yields better results than other methods. An AUC value of 0.903 (90.3%) for the 30-day cumulative rainfall thresholds supports this claim. This model could be a good input for a landslide early warning system in Badung Regency.

Hydromechanical modeling of evolving post-wildfire regional-scale landslide susceptibility

Post-wildfire mass wasting is a major problem throughout many regions worldwide. Recent dramatic increases in global wildfire activities coupled with a shift in wildfire-prone elevation to higher altitudes raise the need to better predict post-fire rainfall-triggered landslides. Despite its importance, only a limited number of studies have investigated landslide susceptibility in areas hit by wildfires using hydromechanical models. However, most of these studies follow either qualitative or semi-quantitative approaches without explicitly considering the fire's effects on the impacted area's physical behavior. This study aims to develop and employ a physics-based framework to generate susceptibility maps of rainfall-triggered shallow landslides in areas disturbed by wildfire. A coupled hydromechanical model considering unsaturated flow and root reinforcement is integrated into an infinite slope stability model to simulate the triggering of shallow landslides from rainfall. The impact of fire is considered through its effects on soil and land cover properties, near-surface processes, and canopy interception. The developed model is then integrated into a geographic information system (GIS) to characterize the regional distribution of landslide potential and its variability considering topography, geology, land cover, and burn severity. The proposed framework was tested for a study site in Southern California. The site was burned in the San Gabriel Complex Fire in June 2016 and experienced widespread landsliding almost three years later following an extreme rainstorm in January 2019. The proposed framework could successfully model the location of observed shallow landslides. The model also revealed a significantly higher likelihood for slope failure in areas burned at moderate to high severities as opposed to unburned and low-burn severity areas. The findings of this study can be employed to predict the timing and general locations of rainfall-triggered shallow landslides following wildfires.

Brief report of fatal rainfall-triggered landslides from record-breaking 2023 storms in Auckland, New Zealand

AbstractThe aim of this paper is to report on fatal landslides resulting from two extraordinary storm events in January and February 2023, in Auckland, New Zealand, including the characteristics of the terrain, the site geology, and the storms. The January 27 Auckland Anniversary storm was from an “atmospheric river,” dumping Summer’s worth of rain (265 mm) in one day. This was the highest 24-h total on record, estimated a 1 in 200-year event, with a peak 2-min rainfall of 4.2 mm, coinciding with Friday’s evening “rush hour.” This led to widespread landsliding throughout Auckland, including a fatal landslide in Parnell, central Auckland. Notably, this occurred only meters from a similarly destructive landslide in 1997. With January 2023 Auckland’s wettest January on record, and with the region still in a state of emergency, Tropical Cyclone Gabrielle crossed the region on February 13–14. This also caused widespread landsliding, including a fatal landslide at Muriwai on the west Auckland coast. Fatal landslides from the same escarpment had occurred in 1965, close to the February 13, 2023, fatal landslide. Taken together, both storm events caused > 140,000 landslides across the North Island, and flooding, making hundreds of people homeless, with damage estimates of > US$8.6 billion (3.4% of NZ’s GDP). For future disaster risk reduction, attention should be paid to revising legislation and local planning, which currently allows people to legally build (and re-build) houses in landslide-prone areas of New Zealand.

PyLandslide: A Python tool for landslide susceptibility mapping and uncertainty analysis

Mitigating the impacts of landslides and planning resilient infrastructure necessitates assessing the exposure to this hazard through, for example, susceptibility mapping involving the spatial integration of various contributing factors. Here, we introduce PyLandslide, an open-source Python tool that leverages machine learning and sensitivity analysis to quantify the weights of various contributing factors, estimate the associated uncertainties, and generate susceptibility maps. We apply PyLandslide to the case of rainfall-triggered landslides in Italy driven by historical precipitation data (1981–2023) and nine climate projections for the mid-century (2041–2050). Results highlight distance to roads as the most influential factor in determining landslide susceptibility in Italy, followed by slope. Our findings reveal an overall reduction in susceptibility in the mid-century compared to the historical period; however, the directional changes vary spatially. Uncertainty analysis should play a central role in decision-making on landslides, where weights are intricately linked to investments.

Land Cover Trajectories and Their Impacts on Rainfall-Triggered Landslide Occurrence in a Cultivated Mountainous Region of Western Japan

Land cover changes in mountainous regions are potential precursors to landslide disasters. However, the effects of past long-term land cover changes on the characteristics of recent landslides remains underexplored. We studied land cover evolution over a 56-year period on Omishima Island in western Japan to examine the spatial relations of landslides in the July 2018 storm event based on rainfall, land cover trajectories, and topography. We generated land cover maps for 1962, 1981, and 2018 by aerial photo interpretation. We also identified 512 new landslides. Based on 47-year precipitation records, we estimated the return periods of 1- to 264-h rainfalls during the storm using the generalised extreme value (GEV) distributions. Return periods showed wide variation when the derived GEV distributions were applied to 1-km grid rainfall distributions. Despite such pronounced spatial variations in rainfall, we did not observe a clear correlation between rainfall intensity and landslide distribution. In contrast, land cover trajectories had a pronounced effect on landslide occurrence. Landslides were more concentrated on slopes that experienced land cover changes after 1962. A comparison of slopes on farmland developed between 1962 and 1981 (mainly citrus orchards) indicated that landslide density and area ratio were significantly lower on slopes that had reverted to forests than on those remaining as farmland. However, the values of the reforested slopes exceeded those of forests and farmlands that remained since before 1962. Our geospatial analysis revealed that even if the field had shifted to forests, the effects of reduced slope stability due to orchard development had remained for at least 37 years. This suggested that the impacts of converting forests to orchards last longer than harvesting in managed plantation forests.

Machine Learning and IoT-Based Approaches to Detect and Predict Rainfall-Triggered Landslides

Machine Learning and IoT-Based Approaches to Detect and Predict Rainfall-Triggered Landslides

Establishing the landslide-triggering rainfall thresholds for the Kashmir Himalaya

Establishing the landslide-triggering rainfall thresholds for the Kashmir Himalaya

Exploring the Use of Pattern Classification Approaches for the Recognition of Landslide-Triggering Rainfalls

Rainfall-triggered landslides are well-known natural hazards that pose significant risks, and lot of effort has been invested to reduce the risk associated with this type of phenomenon. One approach to reduce such risk is the establishment of landslide early warning systems (LEWSs). LEWSs are designated to proactively identify conditions favorable to the initiation of landslides. When dealing with regional scale works, LEWSs are usually based on statistical methodologies to determine the minimum amount of rainfall required to trigger a landslide. This amount is often expressed in terms of minimum intensity or cumulative rainfall in a given time period. This research explores the use of artificial intelligence, specifically Long Short-Term Memory (LSTM) networks to analyze rainfall time series as either likely or not likely to result in a landslide. Various lengths of time series and different configurations of the model were tested to identify the best setting of the model. To develop the research, the selected test site was the Emilia-Romagna region in Italy, which has a robust landslide inventory, with assessed accuracy. Model performances were evaluated using several statistical indicators, including sensitivity (0.9), specificity (0.8), positive prediction power (0.82), negative prediction power (0.89), Efficiency (0.85) and misclassification rate (0.15). These results showed that the defined model correctly identified the rainfall conditions associated with landslide initiation with a high degree of accuracy and a low rate of false positives. In summary, this research demonstrates the potential of artificial intelligence, particularly LSTM networks, in improving the accuracy of LEWSs by analyzing rainfall time series data, ultimately enhancing our ability to predict and mitigate the risks of rainfall-triggered landslides.

Dynamic rainfall thresholds for landslide early warning in Progo Catchment, Java, Indonesia

This study aims to derive and evaluate new empirical rainfall thresholds as the basis for landslide early warning in Progo Catchment, Indonesia, using high-resolution rainfall datasets. Although attempts have been made to determine such thresholds for regions in Indonesia, they used coarse-resolution data and fixed rainfall duration that might not reflect the characteristics of rainfall events that induced the landslides. Therefore, we evaluated gauge-adjusted global satellite mapping of precipitation (GSMaP-GNRT) and bias-corrected climate prediction center morphing method (CMORPH-CRT) hourly rainfall estimates against measurements at rainfall stations. Based on this evaluation, a minimum rainfall of 0.2 mm/h was used to identify rain events, in addition to a minimum of 24 h of consecutive no-rain to separate two rainfall events. Rainfall thresholds were determined at various levels of non-exceedance probability, using accumulated and duration of rainfall events corresponding to 213 landslide occurrences from 2012 to 2021 compiled in this study. Receiver operating characteristics (ROC) analysis showed that thresholds based on rainfall station data, GSMaP-GNRT, and CMORPH-CRT resulted in area under ROC curve values of 0.72, 0.73, and 0.64, respectively. This result indicates that the performance of high-resolution satellite-derived data is comparable to that of ground observations in the Progo Catchment. However, GSMaP-GNRT outperformed CMORPH-CRT in discriminating the occurrence/non-occurrence of landslide-triggering rainfall events. For early warning purposes, the rainfall threshold is selected based on the probability exlevel at which the threshold maximizes the true skill score, i.e., at 10% if based on station data, or at 20% if based on GSMaP-GNRT.

Applying rainfall threshold estimates and frequency ratio model for landslide hazard assessment in the coastal mountain setting of South Asia

Landslides pose a serious risk to life and property in the mountainous regions around the globe. Understanding the interplay of landslide conditioning and triggering factors is essential for lessening the impacts caused by the hazard. Cox's Bazar — a coastal mountainous district in Bangladesh is recurrently affected by rainfall-triggered landslides. Based on analysis of 14 experiential landslides and combination of gauged and satellite rainfall estimates for the period from 2003 to 2019, the present study determines three landslide-triggering rainfall thresholds for the Cox's Bazar District (CBD): 1. Intensity-Duration (ID) threshold derived in this study revealed that any rainfall event with an intensity of ≥4.04 ​mm/h if prolonging for ≥12h can cause slope failures; 2. Event-Duration (ED) threshold suggested that a normalized cumulative event rainfall (EMAP) of 0.15 for one day is expected to trigger landslides; and 3. threshold calculated using randomly chosen antecedent rainfall expressed best distinction on 30-day rainfall and the equation of the threshold came out as Rth ​= ​64–0.02 Ra30. The recurrence probability of the derived antecedent rainfall threshold and likely landslides was determined through the Poisson distribution. Moreover, we assess the landslide susceptibility of the district with a coupled use of Frequency Ratio (FR) statistical measure and Geographic Information System (GIS). Considering the combined role of selected conditioning factors, the landslide susceptibility status of the CBD was quantified and classified into probability intervals. The accuracy of the susceptibility maps was assessed through the Relative Landslide Density Index (R-Index) that used a field landslide inventory, comprising well distributed 891 events. Moreover, gridded population data was superimposed on the derived susceptibility maps to understand the risk levels of people. The derivation of landslide-triggering rainfall thresholds and spatial susceptibility assessment has been useful to propose a low-cost Landslide Early Warning System (LEWS) which can contribute in alleviating the adverse effects of landslide hazard in the CBD.

I–D Threshold Analysis of Rainfall-Triggered Landslides Based on TRMM Precipitation Data in Wudu, China

This study explored the applicability of TRMM, TRMM nonlinear downscaling, and ANUSPLIN (ANU) interpolation of three different types of precipitation data to define regional-scale rainfall-triggered landslide thresholds. The spatial resolution of TRMM precipitation data was downscaled from 0.25° to 500 m by the downscaling model considering the relationship between humidity, NDVI, and numerous topographic factors and precipitation. The rainfall threshold was calculated using the rainfall intensity–duration threshold model. The calculation showed that TRMM downscaled precipitation data have better detection capability for extreme precipitation events than the other two, the TRMM downscaling threshold was better than the ANU interpolation, and the cumulative effective rainfall of TRMM downscaling was preferred as the macroscopic critical rainfall-triggered landslide threshold for the early warning of the Wudu. The predictive performance of the rainfall threshold of 50% was better than the other two (10% and 90%). When the probability of landslide occurrence was 50%, the TRMM downscaled threshold curve was given by I50=21.03×D−1.004. The authors also analyzed the influence of factors such as topography landform and soil type on the rainfall threshold of landslides in the study area. The rainfall intensity of small undulating mountains was higher than that of medium and large undulating mountains, and the rainfall intensity of landslides peaks at high altitude mountains of 3500–5000 m.

The Probabilistic Risk Assessment Imperative in Philippine Climate Change Adaptation-Disaster Risk Reduction (CCA-DRR) Efforts

The Philippines is ranked first in the World Risk Index 2022 according to the Germany-based Bündnis Entwicklung Hilft and the Institute for International Law of Peace and Armed Conflict (IFHV). The country will likely remain at the top rank of the global risk index over the next few decades, mainly due to the Philippines being in the Pacific Typhoon Belt, where the adverse impacts of climate change are predicted to get worse. To solve this predicament, intensive preparations through development planning are needed, and Probabilistic Risk Assessment (PRA) is required to capture and anticipate future hazards bigger than the historical record. The probabilistic approach is recognized in the Philippine Development Plan 2017-2022 and 2023-2028, the 2019 National Climate Risk Management Framework Policy of the Climate Change Commission (CCC), and the United Nations Disaster Risk Reduction (UNDRR) document entitled, “Why Invest in Probabilistic Risk Assessment?” as necessary for effective climate change adaptation actions. However, the Philippines has yet to realize the nationwide-scale implementation of PRAs in our CCA-DRR efforts, despite it being a strategy recommended by the national government to effectively address worsening natural hazard impacts that severely hamper national development. This paper highlights the importance of PRA in the context of Philippine Climate Change Adaptation and Disaster Risk Reduction (CCA-DRR) efforts to lower the cost of damage from floods, rainfall-triggered landslides, and storm surges and help communities attain their sustainable development goals. We recommend that these policies be systematically implemented without any more delay.

Effects of Tectonic Setting and Hydraulic Properties on Silent Large-Scale Landslides: A Case Study of the Zhaobishan Landslide, China

Unlike strong earthquake-triggered or heavy rainfall-triggered landslides, silent large-scale landslides (SLL) occur without significant triggering factors and cause unexpected significant disaster risks and mass casualties. Understanding the initiation mechanism of SLLs is crucial for risk reduction. In this study, the mechanism of the Zhaobishan SLL was investigated, and the SLL was jointly controlled by weak-soil (fractured rock mass) and strong-water (abundant water replenishment) conditions under the impact of active tectonism and complex hydraulic properties. Strong tectonic uplift, high fault density, and historical earthquakes led to weak-soil conditions conducive to the Zhaobishan SLL. The combined effect of unique lithology, antiform, and cultivated land contributed to the water replenishment characteristics of extensive runoff confluence (3.16 times that of the landslide body) and supported long-distance groundwater replenishment, thereby forming strong-water conditions for the landslide. The amplified seepage amount caused the strength of the soil mass on the sliding surface to decrease to 0.4 times its initial strength, eventually triggering the Zhaobishan SLL, which occurred 4.6 days after the peak rainfall. Moreover, the landslide deposits have accumulated on the semi-diagenetic clay rock, thereby controlling the subsequent recurring debris flows in the Lengzi Gully. To reduce disaster risk of SLL in vulnerable mountainous regions, the water confluence area behind the main scarp of the landslides and the hysteresis characteristics between landslides and peak rainfall should be further considered, and recurring debris flows following massive landslides also should be focused.

Exploring extreme rainfall-triggered landslides using 3D unsaturated flow, antecedent moisture and spatially distributed soil depth

The susceptibility to shallow landslides is affected by several factors. Rainfall intensity, soil thickness distribution, and antecedent moisture conditions are components that influence the spatio-temporal prediction of landslides at the watershed scale. Yet, the combined impact of these factors on the susceptibility to landslides has been overlooked. The purpose of this paper is to investigate landslides triggered by extreme precipitation in the Tijuca Massif, Rio de Janeiro, southeastern Brazil. Our approach couples a 3D variably saturated flow solver with the infinite slope stability method to calculate the statistical distributions of the safety factor and the pore pressure at the soil–bedrock interface. Numerical simulations were performed for 6 scenarios considering 1-year spin-up time for soil moisture and different spatially distributed soil depths. The results show that during extreme precipitation events, soil depth and initial moisture conditions may not have much influence on the safety factor, as the intensity and duration of rainfall will be the triggering agents. Comparison of simulated unstable zones and field data from mapped landslide scars further supported this conclusion. We conclude that simple soil depth models are feasible options for regional studies of landslide susceptibility. Our findings are relevant to understanding shallow landslides induced by extreme rainfall in Rio de Janeiro and other regions with similar geological and climate settings.

Anthropogenic influence on extreme temperature and precipitation in Central Asia

We investigate the contribution of anthropogenic forcing to the extreme temperature and precipitation events in Central Asia (CA) during the last 60 years. We bias-adjust and downscale two Inter-Sectoral Impact Model Intercomparison Project (ISIMIP) ensemble outputs, with natural (labelled as hist-nat, driven only by solar and volcanic forcing) and natural plus anthropogenic forcing (labelled as hist, driven by all-forcings), to 0.25^{circ } times 0.25 ^{circ } spatial resolution. Each ensemble contains six models from ISIMIP, based on the Coupled Model Inter-comparison Project phase 6 (CMIP6). The presented downscaling methodology is necessary to create a reliable climate state for regional climate impact studies. Our analysis shows a higher risk of extreme heat events (factor 4 in signal-to-noise ratio) over large parts of CA due to anthropogenic influence. Furthermore, a higher likelihood of extreme precipitation over CA, especially over Kyrgyzstan and Tajikistan, can be attributed to anthropogenic forcing (over 100% changes in intensity and 20% in frequency). Given that these regions show a high risk of rainfall-triggered landslides and floods during historical times, we report that human-induced climate warming can contribute to extreme precipitation events over vulnerable areas of CA. Our high-resolution data set can be used in impact studies focusing on the attribution of extreme events in CA and is freely available to the scientific community.

Predicting Earthquake-Induced Landslides by Using a Stochastic Modeling Approach: A Case Study of the 2001 El Salvador Coseismic Landslides

In January and February 2001, El Salvador was hit by two strong earthquakes that triggered thousands of landslides, causing 1259 fatalities and extensive damage. The analysis of aerial and SPOT-4 satellite images allowed us to map 6491 coseismic landslides, mainly debris slides and flows that occurred in volcanic epiclastites and pyroclastites. Four different multivariate adaptive regression splines (MARS) models were produced using different predictors and landslide inventories which contain slope failures triggered by an extreme rainfall event in 2009 and those induced by the earthquakes of 2001. In a predictive analysis, three validation scenarios were employed: the first and the second included 25% and 95% of the landslides, respectively, while the third was based on a k-fold spatial cross-validation. The results of our analysis revealed that: (i) the MARS algorithm provides reliable predictions of coseismic landslides; (ii) a better ability to predict coseismic slope failures was observed when including susceptibility to rainfall-triggered landslides as an independent variable; (iii) the best accuracy is achieved by models trained with both preparatory and trigger variables; (iv) an incomplete inventory of coseismic slope failures built just after the earthquake event can be used to identify potential locations of yet unreported landslides.

Failure Mechanism of a Rainfall-Triggered Landslide in Clay Slopes

In December 2013, a portion of a large and deep ancient landslide on the southern slope of the Montescaglioso town (Basilicata, Southern Italy) was abruptly reactivated, as a consequence of exceptional rainfall events, causing relevant damages to structures and infrastructures. The sliding surface is supposed to be located within a thick deposit of Pleistocene stiff clays overlain by dislocated blocks of calcarenites and cemented conglomerates. This paper discusses the research carried out to investigate the failure mechanism that occurred during the landslide event and the factors that controlled the reactivation. To this purpose, geological and geomorphological analyses were first proposed, followed by a back-analysis of the landslide process, performed via limit equilibrium calculations implementing time-dependent pore water pressure distributions derived from transient seepage finite element analyses. Furthermore, the overall landslide mechanism was investigated through a three-dimensional finite element analysis, built using the monitoring campaign carried out in the post-failure stage and calibrated according to the in situ failure mechanism evidence. Both the limit equilibrium and finite element analyses provide results in good agreement with the geomorphological evidence, further allowing us to recognize the effects of rainfall infiltration in the increase of pore water pressure along the sliding surface and the variation of the stress–strain state leading to failure occurrence.

DEWS: A QGIS tool pack for the automatic selection of reference rain gauges for landslide-triggering rainfall thresholds

DEWS: A QGIS tool pack for the automatic selection of reference rain gauges for landslide-triggering rainfall thresholds

Landslides forecasting using satellite rainfall estimations and machine learning in the Colombian Andean region

Due to its high topographic gradients and climatic conditions, the Colombian Andean Region (CAR) has been associated with human and economic losses caused by rainfall-triggered landslides over time; for this reason, the definition of the rainfall conditions that trigger landslides has received significant research interest. However, numerous areas that constitute this region do not have ground-rainfall gauge networks to correlate rainfall and landslides. In this sense, the information provided by Satellite Rainfall Estimations (SREs) can help address this deficiency. Therefore, this work aimed to propose a methodology to spatialise the rainfall-triggered landslide hazard over the Andean Region using conditioning variables, SREs (CHIRPSv2 and MSWEPv2.6), and commonly used ground-rainfall gauges. Thus, the performance of the selected SREs is assessed using the modified Kling Gupta efficiency Metric (KGE’) and its components. Subsequently, the Random Forest Model (RFM) was trained for the 1981–2019 period, and its performance was assessed in terms of different skill scores and the Area Under the Receiver Operating Characteristic Curve (ROC-AUC). The model's predictive capacity for rainfall-triggered landslides was also tested using input data from recent events outside the training period (2020–2021). The KGE's results showed a good correlation between rainfall estimates (CHIRPSv2 and MSWEPv2.6) and observations (ground-rainfall gauges) for accumulated rainfall scales greater than three days, being better at the monthly and annual scales (KGE'>0.5). In addition, the performance of the RFM was good in all evaluated cases (ROC-AUC>0.78), with the ground-rainfall database outperforming the satellite databases. The results indicated that CHIRPSv2 is the best-performing SRE for landslide prediction in the Colombian Andean Region. Finally, the landslide forecast model results proved that the RFM could predict rain-triggered landslides using only rainfall variables, providing a suitable option for Landslides Early Warning Systems (LEWS). Lastly, the model results indicated an accurate probability of occurrence using CHIRPSv2 estimations, suggesting that the model would also yield good results in areas with a scarcity of ground-rainfall data.

Hazard assessment for regional typhoon-triggered landslides by using physically-based model – a case study from southeastern China

ABSTRACT Landslide hazard assessment is an important component of risk management and land-use planning. This study aims to investigate the application of a physically-based model named after the fast shallow landslide assessment model (FSLAM) to rainfall-triggered landslide hazard assessment. In August 2015, a total of 123 landslides induced by Typhoon Soudelor in Wenzhou City, southeastern China, was taken as an example. Five input raster files (elevation, soil types, vegetation, antecedent rainfall, event rainfall) and two parameter files regarding soil properties and vegetation were determined. Considering the randomness and uncertainty of soil and vegetation parameters on the regional scale, FSLAM model computes the probability of failure (PoF) by using random parameters inputs. Finally, the landslide hazard map was generated for the study area to reflect the landslide risk. The results showed that FSLAM could accurately capture the effect of rainfall on PoF of slopes, and more than 70% of the landslide were identified in very high/high hazard zones. The accuracy of the receiver operating characteristic (ROC) reached 0.720, which was higher than that of the Transient Rainfall Infiltration and Grid-Based Regional Slope-Stability (TRIGRS) model (0.620). Regarding the computational time, FSLAM had better efficiency, and the consuming time was 1/25 compared with TRIGRS.