Frequency Area Distribution Research Articles

Correction: The performance of landslides frequency-area distribution analyses using a newly developed fully automatic tool

Correction: The performance of landslides frequency-area distribution analyses using a newly developed fully automatic tool

The performance of landslides frequency-area distribution analyses using a newly developed fully automatic tool

The performance of landslides frequency-area distribution analyses using a newly developed fully automatic tool

Coastal rocky slopes instability analysis and landslide frequency-area distribution alongside the road network in west Mediterranean context (northern of Morocco)

Coastal rocky slopes instability analysis and landslide frequency-area distribution alongside the road network in west Mediterranean context (northern of Morocco)

Generating multi-temporal landslide inventories through a general deep transfer learning strategy using HR EO data

Mapping of landslides over space has seen an increasing attention and good results in the last decade. While current methods are chiefly applied to generate event-inventories, whereas multi-temporal (MT) inventories are rare, even using manual landslide mapping. Here, we present an innovative deep learning strategy which employs transfer learning that allows for the Attention Deep Supervision Multi-Scale U-Net model to be adapted for landslide detection tasks in new areas. The method also provides the flexibility of re-training a pretrained model to detect both rainfall- and earthquake-triggered landslides on new target areas. For the mapping, we used archived Planet Lab remote sensing images spanning a period between 2009 till 2021 with spatial resolution of 3–5 m to systematically generate MT landslide inventories. When we examined all cases, our approach provided an average F1 score of 0.8 indicating that we successfully identified the spatiotemporal occurrences of landslides. To examine the size distribution of mapped landslides we compared the frequency-area distributions of predicted co-seismic landslides with manually mapped products from the literature. Results showed a good match between calculated power-law exponents where the difference ranges between 0.04 and 0.21. Overall, this study showed that the proposed algorithm could be applied to large areas to generate polygon-based MT landslide inventories.

Investigating the effects of landslides inventory completeness on susceptibility mapping and frequency-area distributions: Case of Taounate province, Northern Morocco

In explanatory statistics, training data constitute the most significant input since they determine the reliability and accuracy of the output predictive models. In Landslide Susceptibility Mapping (LSM), such data is derived from landslides inventory maps (LIM) that feed the algorithms information regarding the location of past landslide occurrences. However, the impact of LIM size and completeness is rarely investigated. Therefore, we attempt to quantify in this research the amount by which LIM size variability influences the results of LSM modelling in two areas with contrasted geomorphological settings. The latter zones are located in the geomorphologically heterogenous Intra and Mesorif domains (zone A) and the homogenous Prerif region (zone B). To prepare training data, we first extracted 33%, and 66% from the original dataset, which contained an excess of 10000 landslides. Then, we investigated the effect of information loss on frequency area distribution analysis (FAD), which is a useful tool for quantifying the completeness of a given LIM. Then 18 LSMs were generated using Frequency Ratio (FR), Logistic Regression (LR) and Artificial Neural Networks (ANN) models. Our findings suggest that in heterogenous and complex geological settings such as zone A, the LSM and FAD distributions seem to be significantly impacted, while in a homogenous and monotonous study area (zone B), little to no variance is observed in the output models. However, the variability in susceptibility assessment results is also influenced by the algorithm used in LSM preparation, with LR being the most stable algorithm and Artificial Neural Networks presenting the most sensitive model with the most accurate results when using the full LIM.

Investigating the effects of landscape characteristics on landslide susceptibility and frequency-area distributions: the case of Taounate province, Northern Morocco

The geomorphological control on landslides susceptibility maps (LSM) and frequency-area distributions (FAD) is under-investigated. Therefore, we investigate LSMs and FAD distributions in two contrasted landscapes of Taounate province using explanatory statistical techniques. In the geologically heterogenous context, our findings show that LSM spatial distributions for deep landslides are different to those of shallow processes, regardless of the LSM technique used. The Predictive factors also differ between both types of mass movements in this rugged and heterogenous zone. However, the opposite is observed in the geologically homogenous South, where shallow and deep landslides seem to be controlled by the same factors and tend to produce LSMs that are similarly distributed. This is further emphasized by the Area Under Curve (AUC) results which yield good to very good values in the South regardless of the types of landslides used for validation, with average to bad values in the North when shallow landslides are used to validate deep LSMs and vice versa. The same can be said about FADs, which are also sensitive to landscape variations.

Could road constructions be more hazardous than an earthquake in terms of mass movement?

Roads can have a significant impact on the frequency of mass wasting events in mountainous areas. However, characterizing the extent and pervasiveness of mass movements over time has rarely been documented due to limitations in available data sources to consistently map such events. We monitored the evolution of a road network and assessed its effect on mass movements for a 11-year window in Arhavi, Turkey. The main road construction projects run in the area are associated with a hydroelectric power plant as well as other road extension works and are clearly associated with the vast majority (90.1%) of mass movements in the area. We also notice that the overall number and size of the mass movements are much larger than in the naturally occurring comparison area. This means that the sediment load originating from the anthropogenically induced mass movements is larger than its counterpart associated with naturally occurring landslides. Notably, this extra sediment load could cause river channel aggregation, reduce accommodation space and as a consequence, it could lead to an increase in the probability and severity of flooding along the river channel. This marks a strong and negative effect of human activities on the natural course of earth surface processes. We also compare frequency-area distributions of human-induced mass movements mapped in this study and co-seismic landslide inventories from the literature. By doing so, we aim to better understand the consequences of human effects on mass movements in a comparative manner. Our findings show that the damage generated by the road construction in terms of sediment loads to river channels is compatible with the possible effect of a theoretical earthquake with a magnitude greater than Mw = 6.0.

Investigating the Effects of Landslides Inventory Completeness on Susceptibility Mapping and Frequency-Area Distributions: Case of Taounate Province, Northern Morocco

Investigating the Effects of Landslides Inventory Completeness on Susceptibility Mapping and Frequency-Area Distributions: Case of Taounate Province, Northern Morocco

Refined burned-area mapping protocol using Sentinel-2 data increases estimate of 2019 Indonesian burning

Abstract. Many nations are challenged by landscape fires. A confident knowledge of the area and distribution of burning is crucial to monitor these fires and to assess how they might best be reduced. Given the differences that arise using different detection approaches, and the uncertainties surrounding burned-area estimates, their relative merits require evaluation. Here we propose, illustrate, and examine one promising approach for Indonesia where recurring forest and peatland fires have become an international crisis. Drawing on Sentinel-2 satellite time-series analysis, we present and validate new 2019 burned-area estimates for Indonesia. The corresponding burned-area map is available at https://doi.org/10.5281/zenodo.4551243 (Gaveau et al., 2021a). We show that >3.11 million hectares (Mha) burned in 2019. This burned-area extent is double the Landsat-derived official estimate of 1.64 Mha from the Indonesian Ministry of Environment and Forestry and 50 % more that the MODIS MCD64A1 burned-area estimate of 2.03 Mha. Though we observed proportionally less peatland burning (31 % vs. 39 % and 40 % for the official and MCD64A1 products, respectively), in absolute terms we still observed a greater area of peatland affected (0.96 Mha) than the official estimate (0.64 Mha). This new burned-area dataset has greater reliability than these alternatives, attaining a user accuracy of 97.9 % (CI: 97.1 %–98.8 %) compared to 95.1 % (CI: 93.5 %–96.7 %) and 76 % (CI: 73.3 %–78.7 %), respectively. It omits fewer burned areas, particularly smaller- (<100 ha) to intermediate-sized (100–1000 ha) burns, attaining a producer accuracy of 75.6 % (CI: 68.3 %–83.0 %) compared to 49.5 % (CI: 42.5 %–56.6 %) and 53.1 % (CI: 45.8 %–60.5 %), respectively. The frequency–area distribution of the Sentinel-2 burn scars follows the apparent fractal-like power law or Pareto pattern often reported in other fire studies, suggesting good detection over several magnitudes of scale. Our relatively accurate estimates have important implications for carbon-emission calculations from forest and peatland fires in Indonesia.

Landslide Inventory along a National Highway Corridor in the Hissar-Allay Mountains, Central Tajikistan

An increasing amount of landslides leading to significant human and economic consequences is a primary concern for the government of Tajikistan and local authorities. Based on the Committee on Emergency Situations data, from 1996 to 2018, there were 3460 emergencies and more than 1000 fatalities because of earthquake-triggered and rainfall-induced landslides in the region. In addition, landslides caused severe damage to houses and infrastructure facilities due to the population’s lack of landslide hazard knowledge. Therefore, current research focuses on developing a regional-scale landslide inventory map in the Hissar–Allay region, central Tajikistan, where the population density is much higher than at other mountainous territories. In recent decades, the enhancements in geographic information systems, the open access to high-resolution remote sensing data, and an extensive field survey allowed us to identify 922 landslides possible along the highway corridor in the Hissar–Allay region. Based on Varnes’s system, these landslides are classified into four categories: debris flows, rockfalls, shallow landslides, and complex (deep-seated) landslides, considering landslides morphology, geology, deformation of slopes, degree and aspect of slopes, and weathered and disintegrated zones on slopes in the study area. The results show that 8.24% of the total study area is affected by landslides. Along the highway corridor in the Hissar–Allay region there are 96 bodies of deep-seated landslides and 216 rockfall catchments, 273 debris flow catchments, and 313 shallow landslides. Thus, shallow landslides are the most frequent type of movement. In addition, landslide frequency-area distribution analysis shows that shallow landslides are frequent with an area of 1.88E+04 m2; most frequent debris flow channels have a place of 5.58E+05 m2; rockfalls, for its part, are rife with an area of 1.50E+05 m2, and frequent complex landslides have an area of 4.70E+06 m2. Furthermore, it was found out that slopes consist of Silurian formation comprise shales, pebbles, sands, loams, and limestones, metamorphic clays are exposed to landslides more than other geological formations because of the layered structure and their broad spatial distribution in the study area. As the first applied research to compile a landslide inventory map in the Hissar–Allay region on the regional scale, our study provides a sound basis for future explorations of landslide susceptibility, hazard, and risk assessment for this region.

Frequency–magnitude of landslides affected by the 27–29 November 2017 Tropical Cyclone Cempaka in Pacitan, East Java

Tropical cyclone (TC) Cempaka which occurred on 27–29 November 2017 has caused floods, landslides, and strong winds in certain areas of Java Island. Pacitan Regency was the most severely affected by TC Cempaka. The landslide frequency–area distribution curve of event inventory i.e. TC Cempaka can help to understand landslide susceptibility, hazard, vulnerability, and risk. Landslides were identified by using a local government database and by comparing pre- and post-event high-resolution satellite imageries. Field investigation was carried out in March 2018 to November 2018 to verify the landslide location and update the information. Power law, inverse gamma, and double Pareto model were employed to describe the frequency–magnitude of landslide (mLS) triggered by TC Cempaka. The exponent β values of power law, inverse gamma, and double Pareto were 2.6±0.28 (fitted for 8.5% of dataset), 2.2±0.08 (fitted for 83% of dataset), and 2.3±0.09 (best fitted for dataset), respectively. The P-values were 0.51, 0.67, and 0.91 for power law, inverse gamma, and double Pareto, respectively. This study revealed that rollover occurred at 200 and 300 m2 for double Pareto and inverse gamma, respectively. The cutoff points totaled 1096.49 ± 236.44 and 7235.4 ± 1896.7 m2 for double Pareto and power law, respectively. Rollover phenomenon was real and existed in the dataset because it was far from the minimum resolvable size of the landslide that the authors can delineate from the satellite images. mLS for Pacitan was distributed at around 2 to 4. The magnitude of large landslides was 3.2, that of medium landslides was less than 3, and that of small landslides was almost 4. Numerical estimation calculated a fixed mLS=3.01. Comparison analysis of β values obtained from several landslide inventories triggered by heavy rainfall suggests that the variability of β is related to the intensity and duration of rainfall. Triggering events, such as intensity and duration of rainfall, affect the proportion of large landslides that occur in an area. More complete landslide inventories and rainfall data or other landslide triggering factors from other areas are required for further relationship analysis between the β value and landslide triggering factors.

Comparison of Earthquake-Triggered Landslide Inventories: A Case Study of the 2015 Gorkha Earthquake, Nepal

Despite landslide inventories being compiled throughout the world every year at different scales, limited efforts have been made to critically compare them using various techniques or by different investigators. Event-based landslide inventories indicate the location, distribution, and detected boundaries of landslides caused by a single event, such as an earthquake or a rainstorm. Event-based landslide inventories are essential for landslide susceptibility mapping, hazard modeling, and further management of risk mitigation. In Nepal, there were several attempts to map landslides in detail after the Gorkha earthquake. Particularly after the main event on 25 April 2015, researchers around the world mapped the landslides induced by this earthquake. In this research, we compared four of these published inventories qualitatively and quantitatively using different techniques. Two principal methodologies, namely the cartographical degree of matching and frequency area distribution (FAD), were optimized and applied to evaluate inventory maps. We also showed the impact of using satellite imagery with different spatial resolutions on the landslide inventory generation by analyzing matches and mismatches between the inventories. The results of our work give an overview of the impact of methodology selection and outline the limitations and advantages of different remote sensing and mapping techniques for landslide inventorying.

UAV-Based Slope Failure Detection Using Deep-Learning Convolutional Neural Networks

Slope failures occur when parts of a slope collapse abruptly under the influence of gravity, often triggered by a rainfall event or earthquake. The resulting slope failures often cause problems in mountainous or hilly regions, and the detection of slope failure is therefore an important topic for research. Most of the methods currently used for mapping and modelling slope failures rely on classification algorithms or feature extraction, but the spatial complexity of slope failures, the uncertainties inherent in expert knowledge, and problems in transferability, all combine to inhibit slope failure detection. In an attempt to overcome some of these problems we have analyzed the potential of deep learning convolutional neural networks (CNNs) for slope failure detection, in an area along a road section in the northern Himalayas, India. We used optical data from unmanned aerial vehicles (UAVs) over two separate study areas. Different CNN designs were used to produce eight different slope failure distribution maps, which were then compared with manually extracted slope failure polygons using different accuracy assessment metrics such as the precision, F-score, and mean intersection-over-union (mIOU). A slope failure inventory data set was produced for each of the study areas using a frequency-area distribution (FAD). The CNN approach that was found to perform best (precision accuracy assessment of almost 90% precision, F-score 85%, mIOU 74%) was one that used a window size of 64 × 64 pixels for the sample patches, and included slope data as an additional input layer. The additional information from the slope data helped to discriminate between slope failure areas and roads, which had similar spectral characteristics in the optical imagery. We concluded that the effectiveness of CNNs for slope failure detection was strongly dependent on their design (i.e., the window size selected for the sample patch, the data used, and the training strategies), but that CNNs are currently only designed by trial and error. While CNNs can be powerful tools, such trial and error strategies make it difficult to explain why a particular pooling or layer numbering works better than any other.

Large landslides and deep-seated gravitational slope deformations in the Czech Flysch Carpathians: New LiDAR-based inventory

Moderate-relief landscapes, such as the Czech Flysch Outer Western Carpathians (COWC), in comparison with alpine regions are rarely subject to extensive landslide inventory mapping. An understanding of landslides in such a landscape is needed, because densely populated hilly landscapes in temperate zones are usually of major socio-economic importance. In this study, we performed the first LiDAR-based landslide mapping for the entire COWC area (~7539 km2), one of the most landslide-prone regions in Europe. By calculating various landscape and landslide metrics, we infer the distribution, frequency-area relationships, kinematics and controls of mass movements with special attention on large landslides (≥0.1 km2) and deep-seated gravitational slope deformations (DSGSDs). We mapped a total of 13,611 landslides, of which 1357 failures are large landslides and DSGSDs. Whereas the lower and more subdued areas in the southern part of COWC are hotspots in terms of the total number of landslides, the higher and more topographically pronounced areas in the northeast are affected predominantly by large landslides and DSGSDs. However, landslides with ≥0.1 km2 are widespread throughout the whole territory of COWC. A discrepancy also exists in the spatial distribution of different types of landslides. Rock slides and DSGSDs are dominant in the north-east, while flow-type landslides are dominant in the southern lower topographic relief with claystone-dominated flysch. We conclude that 1) distinct geological units (nappes) produce landslide populations with different frequency-area distributions; 2) stratigraphic composition alongside the tectonic style of flysch formations control the type of landslides; 3) DSGSDs affect mainly slopes formed by rigid rocks sitting atop soft formations; and 4) geological conditions, rather than topography, control distribution of large landslides and DSGSDs in COWC.

How size and trigger matter: analyzing rainfall- and earthquake-triggered landslide inventories and their causal relation in the Koshi River basin, central Himalaya

Abstract. Inventories of landslides caused by different triggering mechanisms, such as earthquakes, extreme rainfall events or anthropogenic activities, may show different characteristics in terms of distribution, contributing factors and frequency–area relationships. The aim of this research is to study such differences in landslide inventories and the effect they have on landslide susceptibility assessment. The study area is the watershed of the transboundary Koshi River in the central Himalaya, shared by China, Nepal and India. Detailed landslide inventories were generated based on visual interpretation of remote-sensing images and field investigation for different time periods and triggering mechanisms. Maps and images from the period 1992 to 2015 were used to map 5858 rainfall-triggered landslides, and after the 2015 Gorkha earthquake, an additional 14 127 coseismic landslides were mapped. A set of topographic, geological and land cover factors were employed to analyze their correlation with different types and sizes of landslides. The frequency–area distributions of rainfall- and earthquake-triggered landslides (ETLs) have a similar cutoff value and power-law exponent, although the ETLs might have a larger frequency of a smaller one. In addition, topographic factors varied considerably for the two triggering events, with both altitude and slope angle showing significantly different patterns for rainfall-triggered and earthquake-triggered landslides. Landslides were classified into two size groups, in combination with the main triggering mechanism (rainfall- or earthquake-triggered). Susceptibility maps for different combinations of landslide size and triggering mechanism were generated using logistic regression analysis. The different triggers and sizes of landslide data were used to validate the models. The results showed that susceptible areas for small- and large-size rainfall- and earthquake-triggered landslides differed substantially.

Exploring landslide erosion volume–area scaling relationships by slip depth using changes in DTMs for basin sediment volume estimation

Landslides produce large quantities of sediment deposits and reduce reservoir life. This study investigated landslides at the Shihmen Reservoir basin in Taiwan that were induced by Typhoon Sinlaku and Typhoon Jangmi in 2008. We formulate scaling relationships between landslide erosion volume and area and conclude that sediment budget can be estimated based on the easier-todetermine landslide erosion area. The methodologies applied for the investigation were geomorphological analysis through 5 m × 5 m digital terrain models (DTMs) of the basin created before and after the landslide events and spatial analysis through a geographic information system. The erosion area and volume of landslides were measured through the subtraction of DTMs produced before and after the events. Statistical analysis revealed that the landslide erosion frequency–magnitude distribution exhibited power-law behaviors with a scaling exponent of 2.15 for the frequency–area distribution and 1.66 for the frequency–volume distribution. This paper proposes different scaling relationships for different moving depths, and landslide erosion volumes were estimated on the basis of depth; thus, landslides of different scales can be distinguished to avoid errors in volume estimation. Two different scaling exponents are proposed: 1.21 for landslide erosions with depths of less than 2 m and 1.01 for landslide erosions with depths of more than 2 m. The proposed scaling relationships are practical for landslide erosion volume estimation by different depths according to the landslide area, and they can provide preliminary results for sediment budget planning in a reservoir basin.

VERIFICATION OF HISTORICAL LANDSLIDE INVENTORY MAPS FOR THE PODSLJEME AREA IN THE CITY OF ZAGREB USING LiDAR-BASED LANDSLIDE INVENTORY

Data on landslide locations in the Podsljeme area can be found in three historical landslide inventory maps. Inventory maps from 1979 and 2007 are made based on geomorphological field mapping and historical records while an inventory map from 2017, compiled for the study area (21 km2), is based on LiDAR data acquired in December 2013. A comparison of three landslide inventory maps was performed with three tests based on landslide statistics, frequency-area distribution, geographical discrepancy of landslides and landslide density maps. The results show significant differences in the number of identified landslides as well as in the size and distribution of landslides. A comparison of inventories also showed the unreliability of the existing historical inventories and usefulness of LiDAR data for preparation of complete landslide inventory maps. LiDAR-based inventory for the Podsljeme area could be a valuable tool for a wide range of users i.e., decision-makers, land developers and environmental and civil defense agencies. It is also necessary for landslide susceptibility and hazard mapping, which is a prerequisite for landslide risk reduction.

Factors controlling landslide frequency–area distributions

AbstractA power‐law relation for the frequency–area distribution (FAD) of medium and large landslides (e.g. tens to millions of square meters) has been observed by numerous authors. But the FAD of small landslides diverges from the power‐law distribution, with a rollover point below which frequencies decrease for smaller landslides. Some studies conclude that this divergence is an artifact of unmapped small landslides due to lack of spatial or temporal resolution; others posit that it is caused by the change in the underlying failure process. An explanation for this dilemma is essential both to evaluate the factors controlling FADs of landslides and power‐law scaling, which is a crucial factor regarding both landscape evolution and landslide hazard assessment. This study examines the FADs of 45 earthquake‐induced landslide inventories from around the world in the context of the proposed explanations. We show that each inventory probably involves some combination of the proposed explanations, though not all explanations contribute to each case. We propose an alternative explanation to understand the reason for the divergence from a power‐law. We suggest that the geometry of a landslide at the time of mapping reflects not just one single movement but many, including the propagation of numerous smaller landslides before and after the main failure. Because only the resulting combination of these landslides can be observed due to a lack of temporal resolution, many smaller landslides are not taken into account in the inventory. This reveals that the divergence from the power‐law is not necessarily attributed to the incompleteness of an inventory. This conceptual model will need to be validated by ongoing observation and analysis. Also, we show that because of the subjectivity of mapping procedures, the total number of landslides and total landslide areas in inventories differ significantly, and therefore the shapes of FADs also differ considerably. © 2018 The Authors. Earth Surface Processes and Landforms published by John Wiley & Sons Ltd.

Characterization of landslide distribution and sediment yield in the TsengWen River Watershed, Taiwan

Sediment yield is one of the primary variables of the global denudation system and is often adopted as a measure of terrestrial erosion rates in watersheds. In general watersheds, sediment supply and transport are two of the major mechanisms that may alternate sediment yield. The ordinary approach to conduct sediment prediction is to assume that erosion rates and average annual sediment yields are affected by mean annual precipitation. However, this assumption cannot fully reflect the behavior of sediment yield in regions with relatively high temporal variability. In this study, the TsengWen River Watershed (TWRW) was investigated to characterize the landslide distribution and the associated sediment yield. The results show that lithology can influence the frequency-area distribution of landslides and hence can control the volumetric sediment on hillslopes that contributes to fluvial systems. Short-term sediment yield in downstream reaches of channels correlated with the rainfall erosivity and the water discharge during rainstorms. The amount of supplied sediment and storm runoff on hillslopes as well as the transport capacity of river channels could create limitations in watershed systems. The large amount of sediment transported by rivers is attributed to the additional supply of sediment materials from upland sources such as hillslopes. Subsequent storm events could change the relationships of the sediment supply with respect to water discharge. Hence, large rainfall events are crucial for controlling sediment production, storm runoff on hillslopes, and the transport capacity of river channels in medium watersheds. In addition, anthropogenic forces may also create corresponding impact on the relationships of sediment yield, supply, and transport processes.

An updated method for estimating landslide‐event magnitude

AbstractSummary statistics derived from the frequency–area distribution (FAD) of inventories of triggered landslides allows for direct comparison of landslides triggered by one event (e.g. earthquake, rainstorm) with another. Such comparisons are vital to understand links between the landslide‐event and the environmental characteristics of the area affected. This could lead to methods for rapid estimation of landslide‐event magnitude, which in turn could lead to estimates of the total triggered landslide area. Previous studies proposed that the FAD of landslides follows an inverse power‐law, which provides the basis to model the size distribution of landslides and to estimate landslide‐event magnitude (mLS), which quantifies the severity of the event. In this study, we use a much larger collection of earthquake‐induced landslide (EQIL) inventories (n=45) than previous studies to show that size distributions are much more variable than previously assumed. We present an updated model and propose a method for estimating mLS and its uncertainty that better fits the observations and is more reproducible, robust, and consistent than existing methods. We validate our model by computing mLS for all of the inventories in our dataset and comparing that with the total landslide areas of the inventories. We show that our method is able to estimate the total landslide area of the events in this larger inventory dataset more successfully than the existing methods. © 2018 The Authors. Earth Surface Processes and Landforms published by John Wiley & Sons Ltd.