Earthquake-triggered landslide interpretation model of high resolution remote sensing imageries based on bag of visual word

Traditional visual interpretation is often inefficient due to its excessively workload professional knowledge and strong subjectivity.Therefore,building an automatic interpretation model on high spatial resolution remote sensing images is the key to the quick and efficient interpretation of earthquake-triggered landslides.Aiming at addressing this problem,a landslide interpretation model of high-resolution images based on bag of visual word(BoVW)feature was proposed.The high-resolution images were pre-processed,and then BoVW feature and support vector machine(SVM)was adopted to establish an automatic landslide interpretation model.This model was further compared with the currently widely used Histogram of Oriented Gradient(HoG)feature extraction model.In order to test the effectiveness of the method,typical landslide images were selected to construct a landslide sample library,which was subsequently utilized as the foundation for conducting an experimental study.The results show that the accuracy of landslide extraction using this method reaches as high as 89%,indicating that the method can be used for the automatic interpretation of landslides in disaster-prone areas,and has high practical value for regional disaster prevention and damage reduction.

Preparation of earthquake-triggered landslide inventory maps using remote sensing and GIS technologies:Principles and case studies

Inventory maps of earthquake-triggered landslides can be constructed using several methods,which are often subject to obvious differences due to lack of commonly accepted criteria or principles,To solve this problem,the author describes the principles for preparing inventory maps of earthquake-triggered landslides,focusing on varied methods and their criteria,The principles include the following key points：all landslides should be mapped as long as they can be recognized from images;both the boundary and source area position of landslides should be mapped;spatial distribution pattern of earthquake-triggered landslides should be continuous;complex landslides should be divided into distinct groups;three types of errors such as precision of the location and boundary of landslides,false positive errors,and false negative errors of earthquake-triggered landslide inventories should be controlled and reduced;and inventories of co-seismic landslides should be constructed by the visual interpretation method rather than automatic extraction of satellite images or/and aerial photographs,In addition,selection of remote sensing images and creation of landslides attribute database are also discussed in this paper,Then the author applies these principles to produce inventory maps of four events：the 12 May 2008 Wenchuan,China M_w 7.9,14 April 2010 Yushu,China M_w 6,9,12 January 2010 Haiti M_w7.0,and 2007 Aysen Fjord,Chile M_w 6.2,The results show obvious differences in comparison with previous studies by other researchers,which again attest to the necessity of establishment of unified principles for preparation of inventory maps of earthquake-triggered landslides.

Complex Seismic Focus Structure and Earthquake-Triggered Landslide Distribution:Analysis of the 2014 Ludian M_w6.1 Earthquake in Yunnan

Objective The 2014 Ludian Mw6.1 earthquake in Yunnan occurred in a mountainous area with complex tectonics and topography, which caused serious damage as well as co-seismic landslides of an unusual large scale. Because the suspected seismogenic faults on the surface, distribution of aftershocks and focal mechanism solutions are not consistent, it remains difficult to determine what is the real causal fault or seismogenic structure for this event. Actually, it may imply the complicity of the seismic source at depth. In addition, the distribution of the co- seismic landslides also exhibits some diffusion that is different from general eases, likely associated with the seismic focus structure.

The Calculation Method for the Volume of Landslides and Dammed Lake Sediments Triggered by a Strong Historical Earthquake in the Loess Plateau:A Case Study of Qiuzigou, Gansu Province, Northwest China

The quantitative calculation of the volume of large earthquake-triggered landslides and related dammed lake sediments is of great significance in the study of secondary disasters and focal parameters of strong historical earthquakes.In this study,the dammed lake induced by Qishan M7 earthquake(Lingtai County,Gansu Province,Northwest China)is selected as the research object.Based on the information collected from the 4 boreholes in the dammed lake area,we further take advantage of the lowlevel Unmanned Aerial Vehicle(UAV)photogrammetry and the morphology recovery method,to calculate the volume of the dammed lake and landslides,respectively.Finally,major conclusions are obtained as follows:①the AMS-14C age at the bottom of the Qiuzigou Dammed Lake sediments is 2890±30 BP,which coincides with the 780 BC Qishan earthquake;furthermore,the Qiuzigou Landslides seem to have been triggered by the earthquake,forming an enclosed dammed lake deposition environment after the upstream sediments accumulate;②the Qiuzigou landslides are opposite-sliding landslides that have blocked the river valley;in detail,landslide volumes at the right and left banks are 235×104 m3 and 229×104 m3,respectively.The length of the dammed lake is 2.6 km,with a thickness of approximately 43 m near the landslides,and the total sedimentary volume is 573×104 m3;③the erosion rate of Qiuzigou Landslide Dammed Lake is 0.44 mm/a,the accumulation rate is 15.05 mm/a,and the soil erosion modulus is 593 t/(km2/a),characterized as slight erosion.Quantitative research on the formation of landslides and dammed lakes from strong historical earthquakes is vital for increasing our understanding of the vibrational characteristics and surface action processes of these types of earthquakes.

Relationship between the Landslides Triggered by the Tongwei M71/2 Earthquake in 1718 AD and the Disappearance of Yongning Ancient Town

Although China’s historical earthquake documentation is relatively rich,it is not all based on scientific records.Therefore,the verification of the seismic information in historical records can effectively avoid exaggerating or underestimating the damage they produced.In this paper,we analyze the detailed information of the 1718 AD Tongwei M7⅟􀄟earthquake through field surveys,document sorting,and manual visual interpretation of UAV images.Major conclusions are listed as follows:①The low-level terraces of Weihe River between Gangu and Wushan are fully developed with flatted surface,and the residents here are mostly killed by house collapses.In addition,the disappearance of Yongning Ancient Town is not directly related to the earthquakeinduced landslides;②In fact,“Yongning Town is entirely buried by the earthquake”in the historical records describes the phenomenon that loess dust has pervaded the entire Weihe Valley.These dust grains are produced by the sliding of earthquakeinduced landslides.Thus,there is no possibility that large-scale landslides have slipped over Weihe bedrock and buried Yongning Town;③After the earthquake,survivors abandoned the ruins and selected a new site to live.They built a new town named“Pan’an(means always peaceful in the future)”.earthquake-induced landslides may be the reason of burying the residential areas on valley-side slopes,while those locations inside the valley are associated with the amplification effect of ground vibration.On the basis of compiling historical seismic data,scientific methods are used to explore the real meaning of these documents,for the purpose of providing basic data for the risk evaluation of strong historical earthquakes.

Spatiotemporal variation of the loose deposits in Baisha River Basin after the 2008 Wenchuan earthquake

Extensive landslides were triggered by the Wenchuan earthquake(in China)on 12 May 2008,causing a tremendous amount of loose material suspended on the hillslopes,likely to be eroded and transported by rain.It is of great significance to study the long-term variation of the quantity and spatiotemporal distribution of loose deposits after a great earthquake in order to understand the restoration process,to assess the risk of future soil erosion,including geological hazards,and to further develop ecological governance strategy.In this study,information about the multi-temporal loose deposits on the ranges of the Baisha River Basin,an alpine valley near the epicenter,was extracted by objectoriented remote sensing interpretation,and analysis on their spatiotemporal variation showed that the earthquake in 2008 resulted in loose deposits covering an additional area of 81.09 km2,with a volume estimated at 0.357 billion m3.Within five years after the earthquake,the vegetation had recovered rapidly,and the extent of the deposits was significantly less.From 5 to 13 years after the earthquake,the vegetation continued to recover but at a slower rate and a decreasing trend in newly formed deposits was evident.The total area,volume and quantity of the loose deposits gradually stabilized.The geometric mean center of the loose deposits gradually moved back towards the upstream area after the earthquake and,by 2021,the mean center was about 4 km away from its pre-earthquake position.Taking the mean center of the loose deposits in 2008 as the center,the deposits in the downstream area shrank significantly after the earthquake,but collapses and landslides of deposits persisted in the upstream area to the west-southwest,west-northwest and northnorthwest from the mean center,indicating future source areas of new loose deposits in the basin.

Landslide Susceptibility Mapping in Terms of the Slope-Unit or Raster-Unit,Which is Better?

Choice of appropriate mapping units is important in landslide susceptibility mapping(LSM).There are various possible units for this choice,while it remains unclear which one is better in performance.The purpose of this study is to make a quantitative comparison of two commonly-used units:slope-unit(SU)and raster-unit(RU)based on the landslides triggered by the 2013 Minxian,Gansu,China Mw5.9 earthquake.Ten landslide influencing factors were considered in this analysis.For each type of mapping units,the 70%samples were randomly selected and trained 20 times on the LR model,yielding 20 susceptibility maps,and the remaining 30%samples were tested for the accuracy of the modeling outcome.Different metrics,including the mean probability,model uncertainty,and model prediction skills,were used to evaluate the quality of the susceptibility maps.The results show that the resultant probability maps using two mapping units can largely predict the distribution of actual landslides,on which the high susceptibility area corresponds to the landslide-prone area.The AUC(area under curve)values,ranging from 0.8 to 0.86,show that the prediction ability of two mapping units is roughly the same.While comparing with the RU,the use of SU can lower the model uncertainties caused by the variation of training sets.We converted the RU-based assessment results into SU-based scheme.The results show that two assessment results are well fitted with good linear relationship,which implies that it is feasible to convert the RU-based landslide susceptibility mapping into the SU-based scheme.This analysis indicates that compared with the RU,the SU cannot improve the performance and accuracy of seismic landslide susceptibility mapping.

A Heterogeneous Sampling Strategy to Model Earthquake‑Triggered Landslides

Regional modeling of landslide hazards is an essential tool for the assessment and management of risk in mountain environments.Previous studies that have focused on modeling earthquake-triggered landslides report high prediction accuracies.However,it is common to use a validation strategy with an equal number of landslide and non-landslide samples,scattered homogeneously across the study area.Consequently,there are overestimations in the epicenter area,and the spatial pattern of modeled locations does not agree well with real events.In order to improve landslide hazard mapping,we proposed a spatially heterogeneous non-landslide sampling strategy by considering local ratios of landslide to non-landslide area.Coseismic landslides triggered by the 2008 Wenchuan Earthquake on the eastern Tibetan Plateau were used as an example.To assess the performance of the new strategy,we trained two random forest models that shared the same hyperparameters.The frst was trained using samples from the new heterogeneous strategy,and the second used the traditional approach.In each case the spatial match between modeled and measured(interpreted)landslides was examined by scatterplot,with a 2 km-by-2 km fshnet.Although the traditional approach achieved higher AUC_(ROC)(0.95)accuracy than the proposed one(0.85),the coefcient of determination(R^(2))for the new strategy(0.88)was much higher than for the traditional strategy(0.55).Our results indicate that the proposed strategy outperforms the traditional one when comparing against landslide inventory data.Our work demonstrates that higher prediction accuracies in landslide hazard modeling may be deceptive,and validation of the modeled spatial pattern should be prioritized.The proposed method may also be used to improve the mapping of precipitation-induced landslides.Application of the proposed strategy could beneft precise assessment of landslide risks in mountain environments.